This action might not be possible to undo. Are you sure you want to continue?
Meighan I. Dillon
Mathematics Department
Southern Polytechnic State University
1100 S. Marietta Pkwy
Marietta, GA 30060
mdillon@spsu.edu
December 11, 2006
1 Introduction
Geometry literally means earth measure. When geometry was developed by
the ancients of various cultures, it was probably for the purposes of surveying,
i.e., measuring the earth. In mathematics, geometry usually refers to a more
general study of curves and surfaces. Diﬀerent branches of geometry— diﬀer
ential geometry, algebraic geometry, geometric analysis— are distinguished
in part by the objects of interest and in part by the diﬀerent sets of tools
brought to bear. In diﬀerential geometry, for example, the objects are curves
and surfaces in complex space and the tools arise largely from diﬀerential
calculus. Algebraic geometry is the study of objects that can be described
using rational functions. The tools are typically, but not always, of an alge
braic nature. Some of them are quite elaborate and rarely encountered by
undergraduates, and only encountered by graduate students who will spe
cialize in algebraic geometry or in an area that relies on algebraic geometry,
for example computer science.
Mathematicians who study geometry usually do not talk much about
similar triangles or alternate interior angles or angles in a circle or most of
the other things you probably think of when you think “geometry.” A part of
Euclidean geometry that does come up in many branches of geometry studied
1
today is projective geometry. Although this is not a course in projective
geometry, we will study certain projective planes here. In that sense, this
course forms a bridge from high school geometry to more advanced studies.
The high school geometry course you probably had was primarily a course
in synthetic Euclidean geometry. Synthetic geometry starts from ﬁrst princi
ples: deﬁnitions, and axioms or postulates, which are truths held to be self
evident. The heart of the subject is logical synthesis applied to the axioms
in a rigorous way to uncover the facts (theorems) which relate the objects
of the geometry. This remains an attractive subject for students of any age
with any background because all assumptions are, ideally, clearly stated at
the beginning of the program. The objects are familiar and the relations
among them rich. No previous knowledge is necessary.
Planar Euclidean geometry is the geometry of straight edge and compass.
The straight edge is unmarked, so cannot measure length, and the compass
does not stay open once you lift it from the page. The objects and all rela
tions among them must be produced using only a straight edge and compass
applied to a ﬂat surface. One approach to studying Euclidean geometry is
constructive, and usually a portion of the high school course deals with ac
tual constructions. You can think of the synthetic approach as supplying
detailed instructions for allowed constructions.
The book that brought synthetic geometry to us in the West was The
Elements of Euclid, a masterwork dating from around the year 300 B.C. Not
only has The Elements been studied more or less continuously since it was
written, it has remained the most important source for the subject since that
time. High school students usually study textbooks based on The Elements,
not The Elements itself. These textbooks are, more or less, the products of
interpretations, corrections, and “improvements” to The Elements— many
of questionable merit — that have been incorporated into the subject since
the time of Euclid. One view of these sorts of textbooks is that they attempt
to provide the elusive “royal road to geometry,”
1
ﬁt for schoolchildren.[2], p.
vi.
Our course will start with The Elements itself, in a translation that dates
to the beginning of the last century. Using this as a departure, we explore
1
There is a legend that Ptolemy, the king of Egypt, asked Euclid, his tutor, if there
was not a shorter way to learn geometry. Must one go through The Elements? Euclid’s
response was that there was “no royal road to geometry.” Similar stories are attributed
to other mathematicians in response to complaints from other kings trying to learn math
ematics. See [2], p. 1.
2
some developments in Euclidean geometry since the glory days of Greece.
Since a great deal has happened over the last few centuries, what you see
here represents a very small sampling of topics taken from what is available
for study under the heading ‘geometry.’
Well before the twentieth century, mathematicians recognized that The
Elements had gaps and inaccuracies. There were many signiﬁcant contribu
tions to rectifying The Elements but David Hilbert’s Grundlagen der Geome
trie (Foundations of Geometry), dating from the early 1900s, became the de
ﬁnitive correction and essentially closed the matter once and for all. Hilbert
was able to control some of the diﬃculties with The Elements that had been
recognized but had remained unresolved since antiquity. One of the more
signiﬁcant contributions of Hilbert was his recognition of the necessity of
leaving certain terms undeﬁned in an axiomatic development. Aristotle, who
predated Euclid, thought a great deal about deﬁning basic geometric objects
and noted that “ ‘the deﬁnitions’ of point, line, and surface ... all deﬁne the
prior by means of the posterior” [2], p. 155. In other words, for predeccesors
of Euclid through the dawn of the 20th century, it was common but uneasy
practice to work with deﬁnitions where, for instance, you might use lines to
deﬁne points and points to deﬁne lines. Moreover, it was wellknown that this
was a logical subversion. Hilbert recognized that the solution to the problem
was to jettison some deﬁnitions altogether. One must start with a minimal
set of basic terms that remain undeﬁned: their properties are then detailed
as postulates. For example, we do not deﬁne ‘point’, ‘line’, or ‘plane’ but we
understand the nature of these objects through the postulates, for example,
two distinct points determine a unique line, and a line together with a point
not on that line determine a unique plane. All other deﬁnitions, for example
‘angle’, are then carefully crafted using this basic set of undeﬁned terms.
Our foray into synthetic geometry includes contributions from the Renais
sance that amounted to the initiation of the study of the projective plane
through the addition of ideal points to the Euclidean plane. Our study of
Euclid and Hilbert culminates with the nine point circle.
If Euclidean planar geometry derives from straightedge and compass con
structions, aﬃne geometry derives from straightedge constructions alone.
This sounds poorer than Euclidean geometry but we come into a wealth of
ideas when we use linear algebra to develop a model for aﬃne geometry.
This is an extension of the familiar ideas of coordinate geometry and leads
naturally to the three dimensional vector space model for a projective plane,
an important tool in use today.
3
Projective geometry dates to the 16th century, thus predates many of the
fundamental ideas that we think of as typical of the modern era in mathemat
ics, such as sets and functions. Nonetheless, projective geometry has a strong
ﬂavor of modernity as it provides a setting in which curves and surfaces have
few or no exceptional conﬁgurations. This gives projective geometry tremen
dous power which we exhibit with a brief study of plane algebraic curves.
Algebraic curves can be described parametrically with polynomials. Plane
curves, that is, those that lie in R
2
or C
2
, necessitate the introduction of cal
culations and tools that are manageable but rich enough to give the reader
an appreciation of some of the ideas handled in modern algebraic geometry.
Our goal in this ﬁnal section of the course is Bezout’s Theorem, which is a
description of the intersection of two curves in a plane.
2 Synthetic Geometry
2.1 Background on Euclid and The Elements
There are no contemporary accounts of Euclid’s life but there are some details
about Euclid that scholars have been able to cobble together indirectly and
about which there is little controversy.
It is generally held that Euclid founded a school of mathematics in Alexan
dria, Egypt.
2
In particular, Archimedes, 287212 B.C., sometimes described
as the greatest mathematician who ever lived, studied in Alexandria at
Euclid’s school and cited Euclid’s work in his own writing. It is clear then
that Euclid predated Archimedes. On the other hand, The Elements has de
tailed references to the works of Eudoxus and Theaetetus. To have learned
their work, Euclid would have to have gone to Plato’s Academy.[2], p. 2.
The Academy was established outside Athens around 387 B.C.
3
There is
general agreement that Euclid was too young to have studied with Plato at
the Academy, and too old to have taught Archimedes in Alexandria. This
helps narrow down the dates for Euclid which are currently accepted as about
325265 B.C. [6]
As a student at the Academy, Euclid would have been the product of
2
Alexandria is named for Alexander the Great, who, as a child, studied with Aristotle.
3
This is the origin of the word academic. Academy is actually the name of the place
where Plato set up his institution. The Academy remained in use until 526, another
astonishingly longlived force in the intellectual world.
4
a rich tradition of careful thought, schooled in the writings and teachings
of Plato, Aristotle, and the other Greek geometers. Aristotle, a student of
Plato himself, suggested in his own writings that his students had sources
which codiﬁed the principles of mathematics and science, including geometry,
that were accepted at that time. These sources presumably would have been
available to Euclid as well. In other words, Euclid’s was not the ﬁrst geom
etry text, even if we restrict attention to the West. This does not diminish
its greatness but it is important to maintain perspective. One commonly
accepted view of The Elements is that it pulled together what was known in
geometry at the time, with the goal of proving that the ﬁve platonic solids
are the only solutions to the problem of constructing a regular solid. There is
room for doubt there, though, as several books of The Elements have nothing
whatever to do with the construction of the platonic solids.[2], p. 2.
The Elements, which is in thirteen books, starts with a set of 23 deﬁni
tions. We quote from [2], p. 153154.
1. A point is that which has no part.
2. A line is a breadthless length.
3. The extremities of a line are points.
4. A straight line is a line which lies evenly with the points on itself.
5. A surface is that which has length and breadth only.
6. The extremities of a surface are lines.
7. A plane surface is a surface which lies evenly with the straight lines
on itself.
8. A plane angle is the inclination to one another of two lines in a plane
which meet one another and do not lie in a straight line.
9. When the lines containing the angle are straight, the angle is called
rectilineal.
10. When a straight line set up on a straight line makes the adjacent angles
equal to one another, each of the equal angles is right, and the straight
line standing on the other is called a perpendicular to that on which
is stands.
5
11. An obtuse angle is an angle greater than a right angle.
12. An acute angle is an angle less than a right angle.
13. A boundary is that which is an extremity of anything.
14. A ﬁgure is that which is contained by any boundary or boundaries.
15. A circle is a plane ﬁgure contained by one line such that all the straight
lines falling upon it from one point among those lying within the ﬁgure
are equal to one another.
16. And the point is called the center of the circle.
17. A diameter of the circle is any straight line drawn through the center
and terminated in both directions by the circumference of the circle,
and such a straight line also bisects the circle.
18. A semicircle is the ﬁgure contained by the diameter and the circum
ference cut oﬀ by it. And the center of the semicircle is that of the
circle.
19. Rectilineal ﬁgures are those which are contained by straight lines,
trilateral ﬁgures being those contained by three, quadrilateral those
contained by four, and multilateral those contained by more than
four straight lines.
20. Of trilateral ﬁgures, an equilateral triangle is that which has its three
sides equal, an isosceles triangle, that which has two sides equal, and
a scalene triangle that which has its three sides unequal.
21. Further, of trilateral ﬁgures, a rightangled triangle is that which
has a right angle, and obtuseangled triangle that which has an
obtuse angle, and an acuteangled triangle, that which has three
angles acute.
22. Of quadrilateral ﬁgures, a square is that which is both equilateral and
rightangled; an oblong that which is not equilateral but rightangled;
a rhombus that which is equilateral but not rightangled; and a rhom
boid that which has its opposite sides and angles equal to one another
but is neither equilateral nor rightangled. All other quadrilaterals are
called trapezia.
6
23. Parallel straight lines are straight lines which, being in the same
plane and being produced indeﬁnitely in both directions, do not meet
one another in either direction.
Mathematicians and philosophers succeeding Euclid left a wealth of com
mentary on the deﬁnitions, especially those of point, line, and plane. The
deﬁnition of a line as the shortest distance between two points, for instance,
is due to Legendre,
4
and dates to 1794.
The second and fourth deﬁnitions indicate a distinction between the no
tions of line and straight line. What Euclid calls a line, we usually call a
curve. Euclid’s deﬁnitions thus contain a classiﬁcation of curves into three
types: lines, circles, and everything else. That may not be so strange but
it is worthwhile to note that more detailed classiﬁcations of curves were of
great interest to geometers through the centuries, Euclid’s predecessors and
contemporaries included.
The deﬁnition of angle refers to lines, not necessarily straight lines. We are
used to thinking of an angle as a “rectilineal angle,” in Euclid’s terminology.
When we talk about the angle at which curves intersect, we may get a little
nervous, and think something like, “The angle of intersection of curves at
a point would be deﬁned as the angle of intersection of their tangents at
that point.” Since we are raised from an early age on calculus, we may
feel awkward because we know that curves may or may not have deﬁned
tangents at a particular point. Euclid is not concerned with this problem as,
once again, he directs his energies towards lines and circles, which present
no such ambiguities.
Another note is that Euclid’s deﬁnition of a “rectilineal ﬁgure” at ﬁrst
seems to refer to what we would call a “polygon” but closer examination
reveals a distinction. In standard usage today, a triangle is three noncollinear
points and the line segments they determine. Euclid’s deﬁnition of triangle
refers to the space enclosed by what we call a triangle. The object Euclid
called a triangle is today called a 2simplex. What Euclid calls a circle, we
would call a disk. In spite of this nicely detailed dictionary, make sure that
you know what someone means when he says “circle” or “triangle.” Context
4
According to The MacTutor History of Mathematics Archive, http://www
history.mcs.standrews.ac.uk/history/index.html, Legendre’s reworking of The Elements,
“El´ements de g´eom´etrie ... was the leading elementary text on the topic for around 100
years” and actually formed the basis for most subsequent texts on Euclidean geometry in
Europe and the United States.
7
often makes it clear but in any case, it is not a dumb question.
The deﬁnition of a circle is not what we are used to.
Exercise
What is the usual deﬁnition of a circle? Show that Euclid’s deﬁnition is
equivalent to the more familiar deﬁnition.
Euclid’s ﬁve postulates follow the deﬁnitions. [2], p. 154155
1. A straight line can be drawn from any point to any point.
2. A ﬁnite straight line can be produced continuously in a straight line.
3. One may describe a circle with any center and any radius.
4. All right angles are equal to one another.
5. (The Parallel Postulate) Consider a straight line falling on two straight
lines. If the interior angles on the same side are less than two right
angles, then the two straight lines, if produced indeﬁnitely, meet on
that side on which the angles are less than the two right angles.
Much later in The Elements, Euclid reveals that he actually assumes
that there is a unique line determined by each pair of points. [2] p. 195.
The second postulate says that a line segment (ﬁnite straight line) can be
extended into a line. There is another hidden assumption there, as well, viz.,
that the line determined by a segment is unique.[2], p. 196.
The Parallel Postulate is not what we are used to seeing.
Exercise
What is the usual formulation for the existence of parallel lines in a plane?
Show that this follows from Euclid’s fourth and ﬁfth postulates. (See [2], p.
220 for other formulations of the parallel postulate and the associated
history.)
The ﬁfth postulate was controversial from Euclid’s time through succeed
ing centuries. The concern was whether it should really be a theorem, that
is, whether the ﬁfth postulate actually followed from the other four. A brief
history of attempts to prove it is given in [2], p. 202219. This is an impor
tant part of the history of mathematics partly because all the fussing over the
8
Parallel Postulate led to the development of noneuclidean geometries during
the 19th century. Noneuclidean geometry may seem like an interesting idea
that is of purely academic interest but nothing could be farther from the
truth. Consider, for example, the geometry of the surface of a sphere. You
realize how important this might be to humans when you think about tra
versing large distances across the surface of the earth. Other noneuclidean
geometries are routinely used in the study of space— as in the cosmos. (See
http://archive.ncsa.uiuc.edu/Cyberia/Cosmos/CosmosShape.html for exam
ple.)
The development of noneuclidean geometries by Lobachevsky in Russia
and Bolyai in Hungary proved that the parallel postulate is independent of
the other four. In other words, if you keep the other four postulates but excise
or change the parallel postulate, you still get a viable axiomatic system. That
was an enormous breakthrough, settling a controversy that persisted from
Euclid’s time, although it was many years before mathematicians accepted
and understood the signiﬁcance of noneuclidean geometries.
Euclid states ﬁve “common notions” in addition to the deﬁnitions and
postulates. These are often referred to as “axioms” in the literature. (See p.
221 in [2].) We quote from [2], p. 155:
1. Things which are equal to the same thing are also equal to one another.
2. If equals be added to equals, the wholes are equal.
3. If equals be subtracted from equals, the remainders are equal.
4. Things which coincide with one another are equal to one another.
5. The whole is greater than the part.
The common notions look a bit strange because of where we stand in the
history of the subject. To discuss them further, we should pause to discuss
equivalence relations, an idea that is relatively modern.
Let S be a set. A relation ∼ on S is called an equivalence relation provided
1. ∼ is reﬂexive, i.e., a ∼ a for all a in S;
2. ∼ is symmetric, i.e., a ∼ b implies b ∼ a;
3. ∼ is transitive, i.e., if a ∼ b and b ∼ c, then a ∼ c.
9
Equality is the archetype for equivalence relations. Isomorphism is an
algebraic equivalence relation. Homeomorphism is a topological equivalence
relation. Congruence is a geometric equivalence relation. We are so used
to thinking about congruence that it seems strange to have to say much
about it but remember that equivalence relations per se were not familiar to
Euclid and remember that he was working the subject from the most basic
principles.
The ﬁrst common notion says that equality is transitive. Beware that
in the Heath edition of Euclid, the word equal is used to mean
congruent. In modern English usage, we say angles α and β are equal only
if α and β are diﬀerent names for the same angle. Diﬀerent angles that have
the same measure are not equal, but congruent. This may seem like splitting
hairs but in some sense it underlies aspects of The Elements that bothered
Euclid himself.
The evaluation of Euclid’s postulate system was to whether the deﬁned
terms, postulates, and common notions form a set of complete, consistent,
and independent basic notions. We have mentioned Legendre, who was the
most important commentator of his time and possibly for the next hundred
years. The late nineteenth and early twentieth centuries saw major develop
ments in Euclidean geometry, most notably with the works of Pasch, in 1882;
Veronese, in 1891; and David Hilbert, in 1903 [4]. By the time later editions
of Hilbert’s Grundlagen der Geometrie were published, the stage was set,
if not for a genuine revolution, then for a jump in the evolution of mathe
matics. The importance of axiomatization and abstraction in mathematics
was explored more deeply than ever before. Hilbert’s work was certainly an
important impetus in this direction. This movement was manifested most ob
viously in The Principia Mathematica of Russell and Whitehead, who sought
to reduce all of mathematics to formal logic and set theory.[9] Bourbaki, the
name for a group of mathematicians (with changing membership) based in
Paris from the 1930s through the present, initiated an ambitious project of
putting all of mathematics on a ﬁrm axiomatic foundation. Whether or not
one prefers this approach to mathematics— some argue that examples rather
than axiomatization should drive the subject and that a focus on abstrac
tion leads one astray of applications, a level of ‘purity’ that is fundamentally
misguided— it was and remains a profound inﬂuence. Regardless of how you
respond to the role of abstraction in mathematics, it has been hugely eﬀective
both in dispatching certain problems that remained intractable through the
ages (for example, Fermat’s Last Theorem) and in bringing certain subjects—
10
notably, algebraic geometry— to a better place for further development. (See
[6] for more information about Bourbaki and their project. Texts that re
sulted from their work are still very much in use.)
We consider some of the results (propositions) Euclid proves in The Ele
ments and some possible problems with the proofs.
Proposition 1, its proof, and the accompanying ﬁgure, we quote from [2],
p. 241.
Proposition 1. On a given ﬁnite straight line to construct an equilateral triangle.
Let AB be the given ﬁnite straight line.
Thus it is required to construct an equilateral triangle on the straight line AB.
With center A and distance AB let the circle BCD be described; again, with
center B and distance BA let the circle ACE be described; and from the point
C, in which the circles cut one another, to the points A, B let the straight lines
CA, CB be joined.
Now, since the point A is the center of the circle CDB, AC is equal to AB.
Again, since the point B is the center of the circle CAE, BC is equal to BA.
But CA was also proved equal to AB; therefore, each of the straight lines CA,
CB is equal to AB.
And things which are equal to the same thing are also equal to one another;
therefore, CA is also equal to CB.
Therefore the three straight lines CA, AB, BC are equal to one another.
Therefore the triangle ABC is equilateral; and it has been constructed on the
given ﬁnite straight line AB.
(Being) what it was required to do.
Early commentators noticed that there is an unarticulated assumption
used in this proof, namely that the circles must intersect. Extensive com
ments elaborating on this point are in [2], p. 2423. The correction involves
an assumption about the continuity of a line or curve.
We continue on to Proposition 2, again quoting statement, proof, and
ﬁgure from [2], p. 244.
Proposition 2. To place at a given point (as an extremity) a straight line equal
to a given straight line.
Let A be the given point, and BC the given straight line.
Thus it is required to place at the point A (as an extremity) a straight line equal
to the given straight line BC.
11
A B
C
D E
Figure 1: Proposition 1
From the point A to the point B let the straight line AB be joined; and on it let
the equilateral triangle DAB be constructed.
Let the straight lines AE, BF be produced in a straight line with DA, DB; with
center B and distance BC let the circle CGH be described; and again, with
center D and distance DG let the circle GKL be described.
Then, since the point B is the center of the circle CGH, BC is equal to BG.
Again, since the point D is the center of the circle GKL, DL is equal to DG.
And in these DA is equal to DB; therefore the remainder AL is equal to the
remainder BG.
But BC was also proved equal to BG; therefore each of the straight lines AL,
BC is equal to BG.
And things which are equal to the same thing are also equal to one another;
therefore AL is also equal to BC.
Therefore at the given point A the straight line AL is placed equal to the given
straight line BC.
(Being) what it was required to do.
The statement of the proposition is that, given a point A and a line
segment BC, we can construct a line segment AL which is congruent to BC.
Euclid’s ﬁrst step is to employ Proposition 1 to construct an equilateral
triangle DAB. The points E and F are respectively points on the lines
determined respectively by the segments DA and DB. Euclid is invoking
12
A
B
D
E
F
C
G
K
H
L
Figure 2: Proposition 2
the second postulate here, which says, in our language, that a line segment
is part of a line.
Next, Euclid constructs two circles, the ﬁrst with center B, radius BC.
G and H are points on this circle, but note that while H is arbitrary, G must
be the point where this circle intersects the line BD, as in the picture. The
second circle has center D, radius DG, and contains points K and where
is the intersection point of the line AD and the circle. The circle centered
at B we now call CGH; the circle centered at D we now call GKL. As
radii of the same circle, BC
∼
= BG. Similarly, DL
∼
= DG. By construction,
DA
∼
= DB. The next step invokes Common Notion 3, which allows us to
subtract line segments, essentially, to get AL
∼
= BC.
Try to answer the following questions. Does a correct proof of Propo
sition 2 depend on the relative positions of the point A and the segment
BC? For instance, if A is on the line determined by BC, does the proof still
apply? Does it change depending on whether A is between B and C? What
if A is an endpoint of the segment? In this case, we ought to be able to
construct a segment extending oﬀ the given segment with the same length.
Does Euclid’s proof apply in that case? If the proof needs modiﬁcation, do
you think Euclid missed something or do you think he made a deliberate
choice?
If you think about constructing a segment of a given length, using a
straight edge and compass, you would probably mark the endpoints of the
given segment with the compass, move the compass to a new location, mark
13
A
B
C
D
E
F
Figure 3: Proposition 3
the endpoints of the new segment, and use the straight edge to connect
the dots. This violates the stricture that the compass used for Euclidean
geometry does not stay open when we lift it oﬀ the page. Proposition 2
proves that even with a ﬂoppy compass, you can transfer information about
length from one location to another. Proposition 3 is a corollary that says we
can cut a given line segment to the match the length of a smaller segment.
We quote from [2], p. 246.
Proposition 3. Given two unequal straight lines, to cut oﬀ from the greater a
straight line equal to the smaller.
Let AB, C be the two given unequal straight lines and let AB be the greater of
them. Thus it is required to cut oﬀ from AB the greater a straight line equal to C
the less. At the point A let AD be placed equal to the straight line C; and with
center A and distance AD the circle DEF be described. Now, since the point A
is the center of the circle DEF, AE is equal to AD. But C is also equal to AD.
Therefore each of the straight lines AE, C is equal to AD; so that AE is also
equal to C. Therefore, given the two striaght lines AB, C, from AB the greater
AE has been cut oﬀ equal to C the less. (Being) what it was required to do.
Again, look for implicit assumptions about the relative positions of ob
jects and ask if they make a material diﬀerence to the proof.
The next two propositions are the ﬁrst substantial theorems about trian
gles. Each has spawned much comment through the centuries. Proposition 4
is the familiar sideangleside criterion for triangle congruence. We quote
14
from [2], p. 2478. This time, we omit Euclid’s ﬁgure. You should supply
one as an aid to follow the argument.
Proposition 4. If two triangles have the two sides equal to two sides respectively,
and have the angles contained by the equal straight lines equal, they will also have
the base equal to the base, the triangle will be equal to the triangle, and the remain
ing angles will be equal to the remaining angles respectively, namely those which
the equal sides subtend.
Let ABC, DEF be two triangles having the two sides AB, AC equal to the two
sides DE, DF respectively, namely AB to DE and AC to DF, and the angle
BAC equal to the angle EDF.
I say that the base BC is also equal to the base EF, the triangle ABC will be
equal to the triangle DEF, and remaining angles will be equal to the remaining
angles respectively, namely those which the equal sides subtend, that is, the
angle ABC to the angle DEF, and the angle ACB to the angle DFE.
For, if the triangle ABC be applied to the triangle DEF, and if the point A be
placed on the point D and the straight line AB on DE, then the point B will
also coincide with E, because AB is equal to DE.
Again, AB coinciding with DE, the straight line AC will also coincied with DF,
because the angle BAC is equal to the angle EDF; hence the point C will also
coincide with the point F, because AC is again equal to DF.
But B also coincided with E; hence the base BC will coincide with the base EF.
[For if, when B coincides with E and C with F, the base BC does not coincide
with the base EF, two straight lines will enclose a space: which is impossible.
Therefore the base BC will coincide with EF] and will be equal to it.
And the remaining angles will also coincide with the remaining angles and will be
equal to them, the angle ABC to the angle DEF, and the angle ACB to the
angle DFE.
Therefore etc.
(Being) what it was required to prove.
The controversy about the proof is to do with the socalled “method of
superposition,” that is, the business of picking up one triangle and dropping
it down onto another for comparison. Heath et. al. aver that Euclid himself
was uncomfortable about employing this argument [2], p. 249. In fact, Heath
argues that Common Notion 4 and the bracketed material in the proof were
added by later commentators to justify this step in Euclid’s proof. [2], p.
225.
15
Why is superposition problematic? The argument, which goes back to
the 16th century, is that if superposition is indeed a legitimate method of
proof, one could use it to prove many of the propositions in The Elements
with little trouble. Consider the pains Euclid took to prove Proposition 2.
If superposition were allowed, we could use it there and dispatch the proof
in one line. Using it in this proof and not the others belies a conﬂict in
the underlying ideas. Hilbert resolves the conﬂict with a dramatic ﬂourish,
which we take up in the next section. (See [2], pp. 249250 for more details
regarding this and other diﬃculties with the proof of Proposition 4.)
We will grant Proposition 4 and proceed to consider Proposition 5, known
famously as pons asinorum, the asses’ bridge.[1], p. 6. The most common
stories behind the name are (1) that it refers to those who cannot follow the
proof , or (2) that it refers to the picture Euclid supplied for his proof. The
actual origin and meaning of the name is lost, but people still enjoy puzzling
over it. [2], p. 415
We quote the statement, proof, and ﬁgure from [2], p. 251.
Proposition 5. In isosceles triangles the angles at the base are equal to one an
other, and, if the equal straight lines be produced further, the angles under the base
will be equal to one another.
Let ABC be an isosceles triangle having the side AB equal to the side AC; and
let the straight lines BD, CE be produced further in a straight line with AB,
AC.
I say that the angle ABC is equal to the angle ACB, and the angle CBD to the
angle BCE.
Let a point F be taken at random on BD; from AE the greater let AG be cut oﬀ
equal to AF the less; and let the straight lines FC, GB be joined.
Then, since AF is equal to AG and AB to AC, the two sides FA, AC are equal
to the two sides GA, AB respectively; and they contain a common angle, the
angle FAG.
Therefore the base FC is equal to the base GB, and the triangle AFC is equal
to the triangle AGB, and the remaining angles will be equal to the remaining
angles respectively, namely those which the equal sides subtend, that is, the
angle ACF to the angle ABG, and the angle AFC to the angle AGB. And,
since the whole AF is equal to the whole AG, and in these AB is equal to AC,
the remainder BF is equal to the remainder CG.
But FCwas also proved equal to GB; therefore the two sides BF, FC are equal
to the two sides CG, GB respectively; and the angle BFC is equal to the angle
CGB, while the base BC is common to them; therefore the triangle BFC is also
16
A
B
C
F G
D
E
Figure 4: Pons Asinorum
equal to the triangle CGB, and the remaining angles will be equal to the
remaining angles respectively, namely those which the equal sides subtend;
therefore the angle FBC is equal to the angle GCB, and the angle BCF to the
angle CBG.
Accordingly, since the whole angle ABG was proved equal to the angle ACF,
and in these the angle CBG is equal to the angle BCF, the remaining angle
ABC is equal to the remaining angle ACB; and they are at the base of the
triangle ABC. But the angle FBC was also proved equal to the angle GCB; and
they are under the base.
Therefore etc.
Q.E.D.
Pons asinorum is signiﬁcant for several reasons. First, it can be argued
that this is the board from which a great many of the fundamental theorems
of Euclidean geometry spring. This is, in fact, Coxeter’s approach to teaching
geometry in [1] (cf. Section 1.3.) Second, references in Aristotle indicate
that geometers who preceded Euclid knew pons asinorum but had a diﬀerent
proof, and indeed, quite a diﬀerent approach to the subject altogether. This
helps us understand the role of The Elements in history. Euclid was, to a
large extent, pulling together known results but from early on in the program,
he forges his own way through the thicket. See [2], pp. 252254. Finally,
17
Proclus, a prominent commentator who came around 700 years after Euclid,
relates the proof that Pappus described, which is quite possibly what you
saw in high school. (Compare also [1], Section 1.3.) Here is Pappus’s proof,
as quoted from [2], p. 254. Note that this does not cover the part of the
theorem describing the angles under the base of the triangle.
Let ABC be an isosceles triangle, and AB equal to AC. Let us conceive this one
triangle as two triangles, and let us argue in this way. Since AB is equal to AC,
and AC to AB, the two sides AB, AC are equal to the two side AC, AB. And
the angle BAC is equal to the angle CAB for it is the same. Therefore all the
corresponding parts (in the triangles) are equal, namely BC to BC, the triangle
ABC to the triangle ABC (i.e. ACB), the angle ABC to the angle ACB, and
the angle ACB to the angle ABC, for these are the angles subtended by the
equal sides AB, AC. Therefore in isosceles triangles the angles at the base are
equal.
Pappus’s proof is often presented without attribution and with modiﬁ
cation, namely, that the triangle is picked up from the page and reﬂected
across a median or angle or base bisector (none of which have been deﬁned)
and dropped down again upon itself...or upon a trace it has left on the page
after it has been picked up.... In light of previous remarks, this approach
is at best unsettling and Pappus’s original approach, to “conceive this one
triangle as two triangles,” while faintly suspect, seems elegant in comparison.
Before leaving this topic, we invoke a rebuke to commentators who proposed
to improve Euclid’s proof by means such as these. This is a quote of C. L.
Dodgson (a.k.a. Lewis Carroll) from Euclid and His Modern Rivals, as cited
in [1], p. 6: “Minos: It is proposed to prove [pons asinorum] by taking up
the isosceles Triangle, turning it over, and then laying it down again upon
itself.
“Euclid: Surely that has too much of the Irish Bull about it, and reminds
one a little too vividly of the man who walked down his own throat, to deserve
a place in a stricly philosophical treatise?”
Book I of Euclid’s Elements ends with Proposition 48, the theorem that
says that if the squares of two sides of a triangle add up to the square of
the third side, then it is a right triangle. Proposition 47 is the Pythagorean
Theorem. Note that a square here is a geometric ﬁgure. In these books of
The Elements, Euclid does not work with numbers. Book II is concerned
with parallelograms and Book III takes up circles. Our goal for this part of
the course is the nine point circle but for now we turn away from Euclid to
18
his most successful modern rival, Hilbert, so that we can proceed to our goal
on a more secure foundation.
2.2 Hilbert’s Axioms
Hilbert’s goal in studying The Elements was to sort out the objects Euclid
studied, the assumptions about them, and the relationships among them. His
goal was a description of the foundation of Euclidean geometry that would
yield all the familiar theorems while meeting modern standards of rigor. In
his time, Euclid likely set the standard for rigor but standards change and
Hilbert wanted to bring Euclid’s work up to a modern level of acceptability.
Several of Euclid’s deﬁnitions had to be jettisoned. By Hilbert’s time, it
had long been recognized that is was impossible to explain something like a
straight line “by any regular deﬁnition which does not introduce words al
ready containing in themselves, by implication, the notion to be deﬁned.” [2],
p. 168 (cf. Aristotle’s remark about deﬁning the prior with the posterior.)
Predecessors of Hilbert started with material approximations to terms like
point and line. Hilbert was the one responsible for acknowledging the neces
sity of starting with terms that were and would remain, frankly, undeﬁned.
This may seem like a triviality but it was an important idea in the early
development of logic and set theory and, in some sense, goes back to Kant,
whom Hilbert quotes before the introduction of Foundations of Geometry.
[4], p. 2.
All human knowledge thus begins with intuitions, proceeds thence to concepts
and ends with ideas.
Kant, Critique of Pure Reason, “Elements of Transcendentalism,” Second Part,
II.
Hilbert singled out the following undeﬁned terms referring to objects:
points, denoted A, B, C, etc., lines, denoted a, b, c, etc., planes, denoted
α, β, γ, etc, and space. The points are called the elements of line geometry.
Points and lines are the elements of plane geometry. Points, lines and planes
are the elements of space geometry. The undeﬁned terms referring to relations
among points, lines, planes, and spaces, are lies on, between, and congruent.
Hilbert’s convention, which we follow, is that when we refer to points and
lines with language such as, “A and B are points” it is to be understood that
A and B are distinct.
19
Hilbert’s axioms are organized into ﬁve subsets: Axioms of Incidence,
Axioms of Order, Axioms of Congruence, Axiom of Parallels, and Axioms of
Continuity. These are not verbatim but are only slightly modiﬁed from the
translation of Grundlagen [4] that we are using.
I. Axioms of Incidence
1. Two points A and B determine a line a that contains them.
2. The line determined by two points is unique.
We denote by AB the unique line determined by points A and B.
3. Every line contains at least two points. There exist three points that
do not lie on a single line.
This axiom is a guarantee against triviality, in other words, that points,
lines, and planes are all diﬀerent sorts of objects. Points that lie on a single
line are collinear. Points that do not lie on a single line are noncollinear.
Hilbert did not use these words in his axioms but did later on in the text.
4. Every set of three noncollinear points is contained in a plane. Every
plane contains at least one point.
5. The plane determined by the three noncollinear points is unique.
6. If two points of a line a lie in a plane α, then all of the points of a lie
in α.
7. If two planes have a point in common, then they have at least one more
point in common.
8. There exist at least four points which do not lie in a plane.
Points that lie in a single plane are coplanar. Points that are not in a
single plane are noncoplanar.
Modern set theory dates from the late 19th century with the work of
Georg Cantor. Hilbert knew set theory but did not use its language and
conventions in the Grundlagen. Although we will not deal with it in a deep
way, we are so used to its basic ideas and notation that it seems natural to
use them here. The underlying idea is that lines and planes are sets. The
20
elements of these sets are points. If A is a point lying on the line a or the
plane α, we write A ∈ a or A ∈ α. If the line a lies in the plane α, we think
of a as a subset of α and write a ⊂ α. If the point A belongs to the lines
a and b, the line a and plane α, or the planes α and β, write A ∈ a ∩ b,
A ∈ a ∩ α, or A ∈ α ∩ β.
Exercises
Using only Hilbert’s axioms as stated above, prove the following theorems.
1. The intersection of two lines in a plane is a single point or empty.
2. The intersection of two planes is empty or exactly one line and no
other points.
3. Let α be a plane and a a line that does not lie in α. The intersection
of a and α is empty or a single point.
4. Let a be a line and A a point not on a. There is a exactly one plane
containing a and A.
5. Let a and b be lines with a single point of intersection. There is
exactly one plane containing both a and b.
II. Axioms of Order
Line segments and angles are not undeﬁned terms but the notion of be
tweenness is and it underlies a great many of the familiar ideas from Euclidean
geometry, in particular, segments and angles. Hilbert attributes this set of
axioms to Pasch, a German mathematician of the late 19th centery. [4]
When the point B lies between points A and C, we write A ∗ B ∗ C.
Designate the line determined by points A and B by AB.
1. If A∗ B ∗ C then A, B, C are collinear and C ∗ B ∗ A.
2. If A and C are two points there exists a point B ∈ AC with A∗ C ∗ B.
Hilbert deﬁnes the line segment AB or BA to be the point A and B
together with the points between A and B. When A∗ C ∗ B, C is inside AB
otherwise C is outside AB. A and B are the endpoints of AB. Note that
Hilbert does not use the notion of sensed or directed line segments.
21
3. Of any three points on a line, there is no more than one that lies
between the other two.
4. Let A, B, C be noncollinear points and let a be a line in the plane
ABC that does not contain any of A, B, or C but that passes through
AB. Then a must also pass through AC or BC.
The last axiom says that in a plane, the sides of a triangle are like doors
to its interior and if a line enters through one door, it must exit through
another.
Exercise
Use Axioms of Incidence and Betweenness to prove that for any points A
and B there is a point C on AB with A ∗ C ∗ B. The argument makes use
of Axiom II.4 and is not as easy as one might expect.
The axioms to this point allow us to prove several things that seem intu
itively obvious when we talk about space. For us, space typically means R
n
,
but remember that neither Euclid nor Hilbert was discussing R
n
. Hilbert
in fact remarked famously that his axiom system should be as applicable to
chairs, tables, and bottles of beer, as to points, lines, and planes. [8]
Omitting proofs, we state some of the consequences of the Axioms of
Order.
Theorem 1. Given any ﬁnite number of points on a line, it is possible to
order them A, B, C, . . . , K so that A∗ B ∗ C, B ∗ C ∗ D, . . . , I ∗ J ∗ K.
Theorem 2. Between any two points on a line there exist an inﬁnite number
of points.
Theorem 3. Let a be a line and α a plane containing a. The points of α
that do not belong to a fall into two collections, designated as sides of the
line. For points A and B on the same side of a, the segment AB does not
intersect a. For points A and C on opposite sides of the line, AC intersects
a.
As a line divides a plane into two sets, a point divides a line into two sets,
and a plane divides 3space into two sets. The Axioms of Order are what
allow us to view polygons as having insides and outsides, for example. They
also give us the notion of a ray.
22
Deﬁnition 1. Let A, A
, O, B be four points of the line a such that O lies
between A and B but not between A and A
. The points A, A
are then said
to lie on the line a on one and the same side of O and the points A, B are
said to lie on the line a on diﬀerent sides of the point O. The totality of
points of a that lie on one and same side of O is called a ray emanating from
O. Thus every point of a line partitions it into two rays.
If we are to have triangles, we need angles and some axioms guiding us
in how to think about the notion of congruence.
III. Axioms of Congruence
When discussing segments and angles we use the words congruent and
equal interchangeably.
The ﬁrst axiom of congruence allows the possibility of constructing con
gruent segments. Recall that this was the second proposition in Book I of
The Elements.
1. Given a segment AB on a line a and a line a
with a point A
∈ a
, we
can ﬁnd a point B
on either side of A
so that AB
∼
= A
B
.
The second axiom is a form of the ﬁrst of Euclid’s Common Notions.
2. If AB
∼
= CD and EF
∼
= CD then AB
∼
= EF.
Exercise
Prove that every segment is congruent to itself, i.e., that segment
congruence is reﬂexive. Prove that segment congruence is also symmetric
and transitive.
3. On the line a let AB and BC be segments which, except for B, have
no point in common. On the same or another line a
let A
B
and B
C
be two segments which except for B
also have no point in common. In
that case, if AB
∼
= A
B
and BC
∼
= B
C
then AC
∼
= A
C
.
This axiom allows the addition of segments, as did Euclid’s Common
Notion 3.
Deﬁnition 2. Let α be a plane and h, k any two distinct rays emanating
from O in α and lying on distinct lines. The pair of rays h, k is called an
angle and is denoted ∠(h, k) or ∠(k, h).
23
Straight angles and angles that exceed straight angles are excluded by
this deﬁnition.
5
When necessary, we use a to distinguish a line a from a ray
a contained in a.
An angle partions the points of a plane into the interior of the angle and
the exterior of the angle.
Exercises
1. Give a precise deﬁnition of the interior and the exterior of an angle.
Show that neither set is empty.
2. If R is any region of a plane with the property that whenever points
A and B are in R, then the segment AB is entirely contained in R,
then R is convex. Show that the interior of an angle is convex. Is the
exterior convex as well?
3. Let H be a point of ray h and K be a point of the ray k where h and
k form an angle ∠(h, k). Show that HK lies in the interior of ∠(h, k).
4. A ray emanating from O lies either entirely inside or entirely outside
the angle. A ray that lies in the interior intersects HK.
We continue with axioms of congruence.
4. Every angle in a given plane can be constructed on a given side of
a given ray in a uniquely determined way. Moreover, every angle is
congruent to itself.
As with segments, we are not treating angles as having sense or direction.
Thus ∠(h, k) and ∠(k, h) denote the same angle.
5. If ΔABC and ΔA
B
C
are triangles with AB
∼
= A
B
, AC
∼
= A
C
,
and∠BAC
∼
= ∠B
A
C
, then the congruence ∠ABC
∼
= ∠A
B
C
is also
satisﬁed.
Notice that Axiom III.5 also gives us ∠BCA
∼
= B
C
A
.
5
In [4], it says that “obtuse angles” are excluded by this deﬁnition. This may be an
error in the translation.
24
Exercise
Prove the uniqueness of segment construction, using uniqueness of angle
construction and Axiom III.5. Start with the assumption that a segment
congruent to AB is constructed two ways on a ray emanating from A
to B
and to B
.
Deﬁnition 3. Two angles with a common vertex and a common side are
supplementary provided their other sides form a line. A right angle is an
angle congruent to one of its supplements.
Recall that Euclid deﬁned a right angle in terms of lines impinging on one
another. A bit later in Book I, there is a proposition that says a right angle
is congruent to its supplement. Once again, then, we see Hilbert taking one
of Euclid’s propositions as an axiom.
Axiom III.5 does not quite give us the sideangleside (SAS) criterion
for triangle congruence but it is very close. Before settling the congruence
theorems for triangles, we revisit pons asinorum.
Theorem 4. The base angles of an isosceles triangle are congruent.
Proof. Given isosceles ΔABC with AB
∼
= BC, we can write AB
∼
= BC,
∠B
∼
= ∠B (III.4), BC
∼
= AB. By III.5, ∠A
∼
= ∠C.
Theorem 5 (SAS or ﬁrst congruence theorem for triangles). If ΔABC
and ΔA
B
C
satisfy AB
∼
= A
B
, ∠ABC
∼
= ∠A
B
C
, BC
∼
= B
C
, then
ΔABC
∼
= ΔA
B
C
.
Proof. Axiom III.5 implies ∠BAC
∼
= B
A
C
and ∠ACB
∼
= A
B
C
. It
remains to show AC
∼
= A
C
.
Choose C
on A
C
so that either A
∗C
∗C
or A
∗C
∗C
and A
C
∼
= AC.
We have shown that C
is uniquely determined.
Note then that we have SAS for triangles ΔABC and ΔA
B
C
since
AC
∼
= A
C
, ∠A
∼
= ∠A
, and AB
∼
= A
B
. Axiom III.5 implies that ∠ABC
∼
=
A
B
C
. We also have ∠ABC
∼
= ∠A
B
C”. This violates uniqueness of angle
construction as given in Axiom III.4, unless C
= C
.
The next theorem is commonly called anglesideangle (ASA).
Theorem 6 (ASA or second congruence theorem for triangles). If ∠A
∼
=
∠A
, AB
∼
= A
B
and ∠B
∼
= ∠B
, then ΔABC
∼
= ΔA
B
C
.
25
A B
C
A B
C
C
Figure 5: SAS for triangles
Proof. The picture above applies here as well. Choose C
exactly as in
the last proof and note that this gives us SAS for ΔABC and ΔA
B
C
as AC
∼
= A
C
, ∠A
∼
= ∠A
and AB
∼
= A
B
. The last theorem gives us
ΔABC
∼
= ΔA
B
C
which in turn implies that ∠ABC
∼
= ∠A
B
C
. By
assumption, ∠ABC
∼
= ∠A
B
C
so that, again, unless C
= C
, uniqueness
of angle construction is violated. We conclude that AC
∼
= A
C
, thus that
ΔABC
∼
= ΔA
B
C
by SAS.
Theorem 7. Supplements of congruent angles are congruent.
Proof. Suppose ∠ABC
∼
= ∠A
B
C
. Without loss of generality, we can as
sume AB
∼
= A
B
and BC
∼
= B
C
. By SAS, ΔABC
∼
= A
B
C
.
Now choose points D, D
with A ∗ B ∗ D, A
∗ B
∗ D
, BD
∼
= B
D
.
Again by SAS, ΔADC
∼
= ΔA
D
C
. This gives us ∠ADC
∼
= ∠A
D
C
, in
turn implying that CD
∼
= C
D
, thus SAS for ΔCBD and C
B
D
.
We conclude that ∠DBC
∼
= ∠D
B
C
, as desired.
Corollary 1. Vertical angles are congruent.
Proof. Consider the angles in Figure 7. Each angle marked with double arcs
is supplementary to the one marked with a single arc. The latter is congruent
to itself by Axiom III.4. By the previous theorem, the supplements are
thus congruent. Now notice that we can treat any pair of vertical angles as
supplements to a given angle which is congruent to itself.
The next lemma allows us to add and subtract congruent angles, thus to
prove the sidesideside (SSS) criterion for triangle congruence. The reader
26
A B
C
D A B
C
D
Figure 6: Supplements of congruent angles are congruent
Figure 7: Vertical angles are congruent.
27
should supply a picture to help understand the hypotheses of the result as
well as the proof.
Lemma 1. Let h, k, be rays emanating from a point O. Let h
, k
,
be
rays emanating from a point O
. Suppose that h, k are on the same side
(respectively, diﬀerent sides) of and that h
, k
are on the same side (re
spectively, diﬀerent sides) of
. If ∠(h, k)
∼
= ∠(h
, k
) and ∠(k, )
∼
= ∠(k
,
),
then ∠(h, )
∼
= ∠(h
,
).
Proof. We complete the proof for the case where h, k are on the same side
of and h
, k
, the same side of
. Using congruence of supplements of
congruent angles, the reader can supply details of the other case.
Changing labels if necessary, we can suppose h is interior to ∠(k, ). Pick
K ∈ k, L ∈ and choose K
∈ k
, L
∈
so that OK
∼
= O
K
, OL
∼
= O
L
.
Since h is interior to ∠(k, ), KL ∩ h = H, a unique point. Choose H
∈ h
to satisfy OH
∼
= O
H
. We claim that H
∈ K
L
.
Note that we have SAS for ΔOKL
∼
= ΔO
K
L
and for ΔOHL
∼
= ΔO
H
L
.
Following through, we get ∠OLK
∼
= ∠O
L
K
and ∠OLH = ∠OLK
∼
=
∠O
L
H
. By uniqueness of congruent angles on a given side of a ray,
∠O
L
K
∼
= ∠O
L
H
implying that L
K
= L
H
, i.e., H
∈ K
L
as claimed.
Since KL
∼
= K
L
, HL
∼
= H
L
, it follows that KH
∼
= K
H
.
Using supplements of ∠OHL
∼
= ∠O
H
L
, KH
∼
= K
H
and ∠OKH
∼
=
∠O
K
H
, we get ASA for ΔOKH
∼
= ΔO
K
H
, implying that ∠(h, k)
∼
=
∠(h
, k
).
Recall that we deﬁned a right angle as one that was congruent to its
supplement. We now have a procedure that allows us to construct a right
angle on a given line.
Theorem 8. We can construct a right angle.
Proof. Fix a point O and a ray h emanating from O. Choose a point B oﬀ the
line h and let k be the ray emanating from O passing through B. Construct
an angle congruent to ∠(h, k), by taking a ray k
emanating from O on the
side of h that does not contain B. Choose B
on k
so that OB
∼
= OB
. Let
A be the point where h intersects BB
. Consider two cases.
Case 1: A = O. Here ∠(h, k) and ∠(h, k
) are supplements and congruent
so already are right angles.
Case 2: If A ,= O then we have SAS for ΔAOB and ΔAOB
since
AO
∼
= AO, ∠AOB
∼
= ∠AOB
by construction, and OB
∼
= OB
, also by
28
O
h
k
k
B
A
Figure 8: Construct a right angle
construction. The implication is that ∠OAB
∼
= ∠OAB
. Notice that these
are supplements, thus must be right angles.
The next result brings us closer to SSS as well as transitivity for angle
congruence.
Lemma 2. Let a be a line through points A and B. Let C
1
and C
2
be
on opposite sides of a and suppose AC
1
∼
= AC
2
and BC
1
∼
= BC
2
. Then
∠ABC
1
∼
= ∠ABC
2
.
Proof.
6
Figure 9 shows one arrangement of points that satisﬁes the hypothe
ses of the lemma. Using pons asinorum, we get congruent angles as marked.
We invoke Lemma 1 to subtract congruent angles so that ∠BC
1
A
∼
= ∠BC
2
A.
Now SAS applies to ΔBC
1
A
∼
= ΔBC
2
A. As corresponding angles, ∠ABC
1
∼
=
∠ABC
2
.
The reader should complete the proof by addressing cases in which C
1
C
2
intersects AB, at an endpoint or otherwise.
We must work around the fact that we do not have transitivity of angle
congruence in the next proof.
6
I am indebted to Wilson Stothers for pointing out errors in a previous version of this
proof.
29
A B
C1
C2
Figure 9: This is one conﬁguration possible given the hypotheses of Lemma 2.
Theorem 9 (SSS or the third congruence theorem for triangles). If
in triangles ΔABC, ΔA
B
C
, corresponding sides are congruent, then cor
responding angles are congruent as well so that the triangles themselves are
congruent.
Proof. Construct one angle congruent to ∠BAC at A
on each side of the ray
emanating from A
through C
. B
is on same side of A
C
as one of the rays
making these angles. On that ray, take the point B
1
so that AB
∼
= A
B
1
.
Take B
2
on the other ray so that AB
∼
= A
B
2
. We have CA
∼
= C
A
by
assumption, ∠BAC
∼
= ∠B
1
A
C
and ∠BAC
∼
= ∠B
2
A
C
by construction,
and AB
∼
= A
B
1
∼
= A
B
2
, also by construction. This gives us SAS for
ΔABC
∼
= ΔA
B
1
C
and ΔABC
∼
= ΔA
B
2
C
. It follows from there that
CB
∼
= C
B
1
and CB
∼
= C
B
2
implying C
B
1
∼
= C
B
2
. By Lemma 1, we
have ∠C
A
B
1
∼
= ∠C
A
B
2
. Lemma 1 also applies to A
B
C
and A
B
2
C
so that ∠C
A
B
∼
= ∠C
A
B
2
. Uniqueness of angle construction implies that
B
= B
1
. This gives us ΔABC
∼
= ΔA
B
1
C
= ΔA
B
C
, as desired.
Note that since segment congruence is symmetric, we can also say ΔA
B
C
∼
=
ΔABC.
The SSS criterion for triangle congruence implies that angle congruence
is an equivalence relation. The ﬁrst result in this direction corresponds to
Axiom III.2.
Theorem 10. If ∠(h, k)
∼
= ∠(h
, k
) and ∠(h
, k
)
∼
= ∠(h
, k
) then ∠(h, k)
∼
=
∠(h
, k
).
30
A
B
C A
B
C
B1
B2
Figure 10: SSS for triangle congruence
Proof. Label the vertices of the angles O, O
, O
respectively. On h, choose
a point A. Choose A
and A
on h
, h
respectively so that OA
∼
= O
A
∼
=
O
A
. Choose a point B on k and B
, B
on k
, k
respectively so that
OB
∼
= O
B
∼
= O
B
. This gives us SAS for ΔAOB
∼
= ΔA
O
B
and
ΔA
O
B
∼
= ΔA
O
B
, thus SSS. SSS allows us to say ΔAOB
∼
= ΔA
O
B
,
implying ∠(h, k)
∼
= ∠(h
, k
) as desired.
Corollary 2. Angle congruence is an equivalence relation.
Proof. Since ∠(h
, k
)
∼
= ∠(h
, k
), ∠(h, k)
∼
= ∠(h
, k
) implies by the theorem
that ∠(h
, k
)
∼
= ∠(h, k). Transitivity is now immediate by the theorem.
Our next goal is to establish means for comparing angles. We need two
lemmas.
Lemma 3. Suppose ∠(h, k)
∼
= ∠(h
, k
), with ∠(h, k) emanating from vertex
O, ∠(h
k
) emanating from O
. Let emanate from O interior to ∠(h, k).
The unique ray
emanating from O
on the k
side of h
that satisﬁes
∠(h, )
∼
= ∠(h
,
) and ∠(, k)
∼
= ∠(
, k
) is interior to ∠(h, k).
Proof. Let
be the unique ray emanating from O
on the k
side of h
with
∠(h, )
∼
= ∠(h
,
). To show
is interior to ∠(h
, k
), we need prove that if
H
,= O
is a point on h
and K
,= O
is a point on k
, then H
K
intersects
.
Let H
∈ h
, K
∈ k
be diﬀerent from O
and choose H ∈ h and K ∈ k
so that OH
∼
= O
H
and OK
∼
= O
K
. This gives us SAS for ΔOHK
∼
=
31
ΔO
H
K
. Since is interior to ∠(h, k), we can let L = HK ∩. Pick L
∈
so that OL
∼
= O
L
. This gets us SAS for ΔOHL
∼
= ΔO
H
L
. Note now
that ∠OHL = ∠OHK
∼
= ∠O
H
K
, by the ﬁrst triangle congruence and
∠OHK = ∠OHL
∼
= ∠O
H
L
by the second. By uniqueness, ∠O
H
K
=
∠O
H
L
. We conclude that L
∈ H
K
, which proves the result.
Theorem 11. Let ∠(h, k) and ∠(h
,
) be given, ∠(h, k) emanating from O,
∠(h
,
) emanating from O
. Suppose is the ray emanating from O on the
k side of h such that ∠(h, )
∼
= ∠(h
,
). Suppose k
emanates from O
on
the
side of h
such that ∠(h, k)
∼
= ∠(h
, k
). Then is interior to ∠(h, k)
if and only if k
is exterior to ∠(h
,
).
Proof. Suppose everything is as hypothesized, with interior to ∠(h, k) but
say that k
is interior to ∠(h
,
). By the previous lemma, there is a unique
ray k
emanating from O, interior to ∠(h, ) with ∠(h, k
)
∼
= ∠(h
, k
). Since
∠(h
, k
)
∼
= ∠(h, k), and both k and k
are on the side of h, that violates
uniqueness of angle construction. We conclude that interior implies k
exterior. Changing labels, we get the result in the other direction as well.
Let ∠(h, k) and ∠(h
,
) be as in the theorem. We write ∠(h, k) < ∠(h
,
)
provided the construction of an angle congruent to ∠(h, k) using h
and a ray
emanating from O
on the
side of h
yields a ray interior to ∠(h
,
). It is
clear then that for any two angles α and β, one and only one of the following
statements is true: α < β, α
∼
= β, α > β.
Theorem 12. All right angles are congruent.
Proof. A right angle is deﬁned as one congruent to its supplement. Let
∠(h, k), emanating from O, and ∠(h
, k
) emanating from O
, be right an
gles with supplements ∠(h, ) and ∠(h
,
) respectively. Suppose ∠(h, k) <
∠(h
, k
). Let k
emanate from O on the k
side of h
so that ∠(h, k)
∼
=
∠(h
, k
). Note that k
is interior to ∠(h
, k
) and exterior to its supplement
∠(h
,
). This gives us
∠(h, )
∼
= ∠(h
, k
) < ∠(h
, k
)
∼
= ∠(h
,
)
as well as
∠(h, )
∼
= ∠(h
, k
) > ∠(h
,
)
which is a contradiction.
32
h
k
h
k
k
l l
Figure 11: Right angles are congruent
Deﬁnition 4. An angle greater than its supplement is obtuse. An angle
smaller than its supplement is acute.
Exterior angles play an important role in The Elements.
Deﬁnition 5. Let ΔABC be a triangle. The interior angles of the triangle
are ∠ABC, ∠BCA, and ∠BAC. The exterior angles of the triangle are the
supplements of the interior angles.
Theorem 13. The exterior angle of a triangle is greater than either of the
interior angles that are not adjacent to it.
Proof. Given ΔABC, choose D ∈ AB with A ∗ B ∗ D∗ and AD
∼
= CB.
Suppose ∠CAD
∼
= ∠ACB. Then ΔACD
∼
= ΔACB by SAS, the implication
being that ∠ACD
∼
= ∠CAB. By congruence of supplements, ∠ACD would
be congruent to the supplement of ∠ACB, implying that D ∈ BC. Since
D ∈ AB, it follows that D = B which is absurd. We conclude that ∠CAD ,
∼
=
∠ACB.
Suppose next that ∠CAD < ∠ACB. If we construct an angle congruent
to ∠CAD at C on the AB side of C, we get a ray interior to ∠ACB that
intersects AB at a point B
. This gives us a triangle ΔAB
C, though, with
exterior ∠CAD congruent to interior angle ∠ACB
, which, as shown above,
is impossible.
We conclude that ∠CAD > ∠ACB.
To show that ∠CAD > ∠ABC, use the fact that ∠CAD is congruent to
its vertical.
Exercises Prove the following corollaries to Theorem 13.
33
1. In every triangle, the greater angle lies opposite the greater side.
(Hint: Start by picking a point on the greater side that cuts oﬀ a
segment with the length of a smaller side.)
2. A triangle with two equal angles is isosceles.
3. If AB
∼
= A
B
, ∠BAC
∼
= B
A
C
, and ∠BCA
∼
= B
C
A
, then the
triangles are congruent. (This is the angleangleside criterion, or
AAS.)
Theorem 14. Every segment can be bisected.
Proof. Let AB be given and let C be a point oﬀ AB. Construct an angle
congruent to ∠CAB with vertex B, taking a ray, k, that extends on the
side of AB that does not contain C. Pick the point D on k that satisﬁes
AC
∼
= BD.
Since C and C
lie on opposite sides of AB, CC
has a nonempty inter
section with AB. Call that intersection point E. We claim that E cannot
coincide with A or B. If it did coincide with A, say, then ∠BAC is exterior
to ΔBAD. But ∠ABD, which is interior to ΔBAD, was constructed to be
congruent to ∠BAC.
Note now that ΔACE
∼
= ΔBDE by AAS: ∠AEC
∼
= BED, since they
are vertical angles; ∠EAC
∼
= ∠EBD by construction; and AC
∼
= BD by
construction. We conclude that AE
∼
= EB, thus, that CD bisects AB.
Corollary 3. Every angle can be bisected.
Proof. Let ∠(h, k) emanate from O. Pick A ∈ h and B ∈ k so that OA =
OB. Note that ΔAOB is isosceles. Let AB have midpoint C. Triangles
ΔACO and ΔBCO have SAS by pons asinorum. It follows that ∠AOC
∼
=
∠BOC, hence, that OC bisects ∠(h, k).
Axiom IV: Axiom of Parallels
Deﬁnition 6. Two lines are parallel provided they lie in the same plane and
do not intersect.
When we deﬁne something, it is important to establish that the deﬁnition
is not vacuous, that is, that we deﬁne something that exists.
34
A
B
C
D
E
Figure 12: Segments can be bisected
Consider a, a line in a plane, α. Pick two points on a, A and H, and let
B be a point of α not on a. Let c = AB. Using B as the vertex and the side
of c containing H, construct the angle congruent to ∠HAB. Let k be the
constructed ray and take a point K ∈ k.
If a and k had a point of intersection, say D, then A, B, and D would
form the vertices of a triangle. One of its interior angles would be ∠HAB
and the constructed angle at B would be exterior. Since we constructed this
angle congruent to ∠HAB, this is impossible. Thus a and k are parallel.
Can there be more than one parallel to a through B? The Parallel Pos
tulate says no. It is important to realize that there is nothing in Hilbert’s
system that forces uniqueness.
IV. (Euclid’s Axiom) Let a be any line and A a point not on it.
There there is at most one line in the plane, determined by a and
A, that passes through A and does not intersect a.
The construction we described above thus yields the unique parallel to a
through B.
If the line a is parallel to the line b write ab. It is easy to see that this
relation is symmetric.
Exercise Show that the parallel relation is transitive.
35
A
B
K
k
a
c
H
Figure 13: Construct a parallel to a given line
Theorem 15. If two parallels are interesected by a third line then the cor
responding and alternate angles are congruent and conversely, congruence of
the corresponding or alternate angles implies the lines are parallel.
Proof. Let a and b be parallel lines intersected at points A and B respectively
by a third line, c. Let C be a point on c. Take points A
∈ a, B
∈ b on the
same side of c. Consider the corresponding angles ∠A
AC and ∠B
BC.
By the discussion above, we can construct a line through B parallel to A
such that the angle at B corresponding to ∠A
AC is congruent to ∠A
AC.
By uniqueness of parallels, b must be that line. Thus ∠B
BC
∼
= ∠A
AC.
To get congruence of the other sets of corresponding angles and alternat
ing angles, use supplements and verticals.
Now suppose that a and b are lines intersected by the line c so that
corresponding and alternating angles are congruent. Let a∩c = A, b ∩c = B
and suppose a ∩ b = C, i.e., suppose a and b do intersect. We have ΔABC
then with interior angle ∠CAB congruent to the corresponding angle, which
is exterior to ΔABC, which is absurd.
Theorem 16. The angles of a triangle add up to two right angles.
Proof. Let ΔABC be given. Let c be the parallel to AB through C. Note
that we get congruent angles to ∠A, ∠B, ∠C as shown, the angles congruent
to ∠A and ∠B because of parallels, the one congruent to ∠C as a vertical
angle. The result follows since the constructed angles add up to a straight
angle, that is, two right angles.
36
a
b
A
B B
A
C
Figure 14: Alternate interior angles are congruent, as are corresponding an
gles.
A B
C
c
Figure 15: Angles in a triangle add up to a straight line.
37
It is only at this point that Hilbert introduces circles.
Deﬁnition 7. If M is any point in a plane α then the collection of all points
A in α for which the segments MA are congruent to each other is called a
circle. M is the center of the circle.
Besides the results we have about bisecting angles and segments, con
structing parallels, etc., there are two traditional construction that Hilbert
does not do explicitly that we will want in our arsenal while we pursue the
orems associated to circles.
Exercises
1. Theorem 8 shows how to construct a right angle to a given line. Give
a construction for dropping a perpendicular from a given point to a
given line.
2. Give a construction for drawing a perpendicular to a line at a given
point on that line.
The last set of Hilbert’s axioms are to do with the geometry of the line.
A thorough treatment of proportion, as in Book V of The Elements, calls for
these sorts of ideas. We state the continuity axioms here and invoke them
later, although we do not discuss the associated results in this part of the
Grundlagen.
V. Axioms of Continuity
1. (Axiom of Measure or Archimedes’s Axiom) If AB and CD are any
segments then there exists a number n such that n segments CD
constructed continguously (endtoend) from A, along the ray from
A through B, will pass beyond the point B.
2. (Axiom of Completeness) An extension of a set of points on a line with
its order and congruence relations that would preserve the relations
existing among the original elements as well as the fundamental prop
erties of line order and congruence that follows from Axioms IIII and
V.1 is impossible.
The Axiom of Completeness comes up when we construct the projective
plane by adding points to the Euclidean plane.
38
A B
C
D
E
F
Figure 16: The circumcircle of a triangle is centered where perpendicular
bisectors intersect
2.3 The Nine Point Circle
We use our work on Hilbert’s axioms to prove some of the theorems we need
to discuss the nine point circle, an interesting circle associated to a triangle.
Our main sources now are [1] and [2].
There are several circles associated to a given triangle. We start our dis
cussion by considering the circumcircle, the circle determined by the vertices
of the triangle.
Theorem 17. Three noncollinear points determine a circle.
Proof. Let A, B, C be noncollinear and consider ΔABC. Bisect AB and
AC and call the midpoints of those segments D and E respectively. Draw
the perpendicular to AB at D, the perpendicular to AC at E and let F be
the point where those two perpendiculars intersect. (Exercise: Why must
they intersect?)
We get SAS for ΔAFE
∼
= ΔCFE implying that CF
∼
= AF. We also
have SAS for ΔADF
∼
= ΔBDF so AF
∼
= BF. Since A, B, C are equidistant
from F, they determine the circle centered at F with radius AF.
Deﬁnition 8. A parallelogram is formed by 2 pairs of parallel lines. The
39
A
B
C D
E
Figure 17: A parallelogram
points where the lines intersect are the vertices of the parallelogram. Line
segments that connect nonadjacent vertices are diagonals.
Parallelograms come up in Book I of The Elements and prove an impor
tant tool for understanding plane ﬁgures in general. Proposition 34 from
Book I describes their basic features.
Theorem 18 (Prop. I.34). The opposite sides of a parallelogram are con
gruent. Opposite angles of a parallelogram are congruent. The diameters of
a parallelogram bisect one another.
Proof. Since AC cuts parallels AB and CD, alternate interior angles are
congruent giving us ∠CAB
∼
= ∠DCA. Also, AC cuts parallels AD and BC
so ∠ACB
∼
= ∠CAD. Adding angles we get ∠A
∼
= ∠C. Similarly we can get
∠B
∼
= ∠D, using the diagonal BD.
Now ΔABC
∼
= ΔADC by AAS, since AC is a common side. This gives
us AB
∼
= DC and AD
∼
= BC.
Let E be the point of intersection of AC and BD. (Why must they
intersect?) We have ASA for ΔAEB
∼
= ΔCED with corresponding sides
AE
∼
= CE. Similarly, DE
∼
= BE.
Exercise Show that adjacent angles in a parallelogram are supplements.
Deﬁnition 9. An altitude of a triangle is the perpendicular dropped from
a vertex to the line determined by the opposite side. The feet of a triangle
are the points on the lines determined by the sides of the triangle where
40
A B
C
F1
F2
F3
Figure 18: The orthic triangle
they intersect the altitudes. The orthic triangle of ΔABC is the triangle
determined by the feet.
The ninepoint circle, Feuerbach circle, or Euler circle of a triangle ΔABC
is the circumcircle of the orthic circle of ΔABC.
Deﬁnition 10. The line segment joining the vertex of a triangle to the mid
point of the opposite side is a median of the triangle.
You may remember from your high school geometry course that the me
dians of a triangle intersect in a single point, that is, they are concurrent.
To prove this, we need results about parallels and ratios, starting with a set
of propositions from Book I. In this part of The Elements, Euclid’s reference
to plane ﬁgures being equal means not that they are congruent, but that
they enclose equal areas. To avoid confusion, we will paraphrase Euclid, as
translated by Heath. Also, when we write ΔABC = ΔA
B
C
, it means that
the areas of the two triangles are equal.
Theorem 19 (Prop. I.35, 36, 37, 38, 39). Parallelograms (respectively tri
angles) on the same base or equal bases that lie on a line have equal areas if
and only if they are in the same parallels.
41
Figure 19: Rectangles in the same parallels
The notion of two parallelograms being “in the same parallels” makes
sense: bases form one line and opposite sides form another. Two triangles
are in the same parallels provided their bases form a line parallel to that
formed by the vertices of the opposite sides. Being in the same parallels
amounts to having the same height.
Euclid (Book V) and Hilbert (Axioms of Continuity) both set up the
tools for discussing comparitive lengths and areas. For example, we can
make sense of the idea of multiplying a line segment by 2 to get a new line
segment with double the length of the original. Then we can compare the
lengths of diﬀerent segments. If there are positive integers m and n such
that m AB
∼
= n CD then AB and CD are commensurate and we write
AB/CD = m/n. Area is trickier since it involves the notion of multiplying
line segments instead of adding them but without getting too involved, we can
accept the idea of comparing areas. For example, the area of a parallelogram
is twice the area of a triangle formed by adjacent sides and a diagonal.
When we write ΔABC/ΔEFG we mean the ratio of the areas of the
triangles and when we write AB/CD we mean the ratio of the lengths of line
segments.
We state Proposition 1 from Book VI and Proposition 39 from Book I
without proof.
Proposition 6 (Prop. VI.1). If triangles (respectively, parallelograms) have
the same height, the ratio of their areas is equal to the ratio of their bases.
Proposition 7 (Prop. VI.2). A line parallel to one side of a triangle cuts
the sides it intersects in equal proportions. Conversely, if a line cuts two
42
A
B C
D E
Figure 20: Parallel to the base cuts the other sides in equal proportions
sides of a triangle in equal proportions, it is parallel to the third side of the
triangle.
Proof. Consider ΔABC with DEBC, intersecting AB at D and AC at E.
Since ΔBDE and ΔCDE have coincident bases and are in the same paral
lels, ΔBDE = ΔCDE. It follows that ΔBDE/ΔADE = ΔCDE/ΔADE.
Next, view AD as the base of ΔADE and BD as the base of ΔBDE.
These two triangles have the same height as determined by the perpendicu
lar to AB from E. By Prop. VI.1, ΔBDE/ΔADE = BD/AD. Similarly,
ΔCDE/ΔADE = CE/EA. It follows that BD/AD = CE/AE, which
proves the ﬁrst part of the result.
Now let the sides AB, AC be cut so that BD/DA = CE/EA. We need
establish that DEBC.
We have ΔBDE/ΔADE = BD/DA. Similarly, ΔCDE/ΔADE =
CE/AE. It follows that ΔBDE/ΔADE = ΔCDE/ΔADE so ΔBDE =
ΔADE. As ΔBDE, ΔCDE have a common base, they must be in the same
parallels (Prop. I.39), which proves the result.
Exercise In reference to the last theorem, show that AB/AD = AC/AE.
This is the way we usually use the theorem.
Deﬁnition 11. Two triangles are similar provided corresponding sides are
in equal proportion.
Next is a theorem about similarity. Euclid’s proof uses a construction
in which the two triangles are arranged so that bases lie on a line and one
vertex is shared. The argument we use is maybe a variant on that preferred
43
A B
C
A'' =
B''
C''
Figure 21: Equiangular triangles have corresponding sides in equal propor
tions
by De Morgan (see [2], p. 202.) Here we use the results we developed in
Hilbert’s system.
Theorem 20 (Prop. VI.4). In equiangular triangles, corresponding sides
are in equal proportion.
Proof. Given equiangular triangles ΔABC, and ΔA
B
C
we can make a copy
of ΔA
B
C
∼
= ΔA
B
C
, as follows. Let A
= A, so that if we construct a
copy of ∠A
using AB for one ray and taking the other ray on the C side of
AB, then by uniqueness of angle construction, the second ray has to coincide
with AC. Let B
be on the B side of A
so that A
B
∼
= A
B
, and let C
on AC be chosen so that A
C
∼
= A
C
. Notice that by SAS, ∠B
∼
= ∠B and
∠C
∼
= ∠C.
Hilbert’s one theorem on parallels implies that B
C
BC. Now the exer
cise following Prop. VI.2 above gives us AB/A
B
= AC/A
C
. We can get
another pair of sides in the same ratio by constructing ΔA
B
C
starting
with B
= B, for example. This implies the result.
Exercise
1. Show that if corresponding sides of two triangles are in equal
proportions, then the triangles are equiangular.
2. Show that if two pairs of sides of two triangles are in equal
proportions, and the angles between are congruent, then the third
pair of sides of the two triangles are in the same portion as the others.
The proof for the following we get from [1], p. 10.
44
Theorem 21. The medians of a triangle are concurrent.
Proof. Consider ΔABC and let D be the point where two medians intersect,
say the ones determined by the midpoint E of AC and the midpoint F of
AB. Let G and H be the midpoints respectively of CD and BD. Since
EF cuts the sides of ΔABC into the same proportions, it is parallel to BC.
Similarly, we see GHBC, when we consider ΔBCD. This makes EFGH a
parallelogram. Since the diagonals of a parallelogram bisect one another, it
follows that the point D trisects segments EB, FC. Similarly, it must trisect
the third median, which implies the result.
There are several diﬀerent “centers” associated to a triangle. The point
where the medians intersect is the centroid, which is the center of gravity
of a triangular plate of uniform density. (That was proved by Archimedes.)
The center of the circumcircle is called the circumcenter. We did not prove
this, but it is true that the circle determined by the vertices of a triangle
is unique. This is another way of saying that diﬀerent circles can intersect
in one or two points, but no more. As a consequence, the circumcenter is
unique, i.e., the perpendicular bisectors of a triangle are concurrent. The
next result gives us one more center of a triangle.
Theorem 22. The altitudes of a triangle are concurrent.
The point of concurrency is the orthocenter of the triangle.
Before we prove the theorem, we consider the Euler line of a triangle.
(The discussion and proof are from [1], p. 17.)
Let O be the circumcenter and G the centroid of the triangle ΔABC.
(Recall that O is the point where perpendicular bisectors of sides of ΔABC
intersect.) If O = G, then the medians of ΔABC are perpendicular to the
sides. (See Figure 22.) This gives us congruent triangles ΔAA
C, ΔBA
A,
implying that AB
∼
= AC. Using a diﬀerent median, we can repeat the
argument to show that the triangle is equilateral.
If the triangle is not equilateral, the Euler line is OG.
Proof. Choose H ∈ OH with O ∗ G ∗ H and OH = 3OG, in other words,
GH = 2OG. Recall that the segment from vertex to centroid, for example
GA, is twice the segment from the midpoint of the opposite side to centroid,
A
G, that is GA = 2A
G. Consider ΔOGA
and ΔAHG.
Exercise Prove the following variant on Proposition 7, referring to
Figure 23. Show that OA/OA
= OB/OB
if and only if ABA
B
.
45
A
B C A
Figure 22: If medians and perpendicular bisectors coincide, the triangle is
equilateral.
O
A
B
B A
Figure 23: The triangles are similar if and only if ABA
B
.
46
Figure 24: The Euler line passes through the centroid G, the circumcenter,
O, and the orthocenter, H.
By the exercise, AHOA
. Moreover, OA
is the perpendicular bisector
of BC. Thus AH is perpendicular to BC. Similarly BH is perpendicular to
CA, and CH to AB. It follows that H is the intersection of the altitudes of
ΔABC.
Corollary 4. The circumcenter, centroid, and orthocenter of a triangle are
collinear.
We have not said much about circles in general and angles associated
with them.
Lemma 4. The circumcenter of a right triangle is the midpoint of its hy
potenuse so the hypotenuse is a diameter of the circumcircle.
Proof. Let ΔABC be a triangle with ∠C a right angle. The perpendicular
bisector to AB is parallel to CB so intersects AB at its midpoint, C
, by
Prop. VI.2. The perpendicular bisector to CB is parallel AC so intersects
AB at C
as well. It follows that C
is the circumcenter for ΔABC. Moreover,
the hypotenuse AB must be a diameter for the circumcircle.
Fix a point on a circle to be the vertex of an angle. When the rays
deﬁning the angle are determined by two other points on the circle, this is an
inscribed angle. If the two other points are opposite ends of a diameter, the
angle thus determined must be a right angle. We can see this by marking
the center of the diameter and joining it to the vertex of the angle. Now
use pons asinorum and supplementary angles to complete the proof. These
observations together with the last lemma prove the following.
47
A
B
C
A'
B'
C'
Figure 25: The circumcenter of a right triangle is the midpoint of its hy
potenuse.
Theorem 23. An angle inscribed in a circle is a right angle if and only if
its sides emanate from points at either end of a diameter.
This was an easy theorem to prove but is an instance of a wellknown
theorem that is a bit more work. We leave it as an exercise to prove in just
one case of the reader’s choosing.
Exercise Let α be an angle inscribed in a circle. Let β be the associated
central angle, i.e., the vertex of β is the center of the circle and its sides are
the radii determined by the points where the sides of α intersect the circle.
Prove that β = 2α. Choose a case to make the proof more accessible.
It is an easy step from there to see that a circle is the locus of points
determined by the ends of a diameter, together with the vertices of all the
right angles with sides passing through the endpoints of the diameter.
The proof of the next is from [1], p. 18.
Theorem 24 (The Nine Point Circle). The midpoints of the three sides
of a triangle, the midpoints of the three segments joining the vertices to the
orthocenter, and the feet of the three altitudes of the triangle, all lie on a
circle.
48
Figure 26: An inscribed angle is a right angle if and only if its sides are
determined by the endpoints of a diameter.
α
β
Figure 27: The inscribed angle is α. The central angle is β.
49
A
B C
H
A"
A'
B'
B" C"
C'
D
Figure 28: Proof of the nine point circle
Proof. Let ΔABC be given and let A
be the midpoint of BC, B
be the
midpoint of AC, and C
be the midpoint of AB. Let H be the orthocenter
and let A
be the midpoint of AH, B
the midpoint of BH, and C
the
midpoint of CH. Since C
B
cuts sides AB and AC in half, Prop. VI.2
implies C
B
BC. Next consider ΔBHC and note that since B
and C
cut BH and CH in half, B
C
BC as well. Considering ΔAHC, we see
that since B
cuts AC in half and C
cuts CH in half, B
C
AH. Similarly,
C
B
AH. This gives us a parallelogram, C
B
C
B
.
Since C
B
and B
C
are both parallel to AH which is perpendicular to
BC, which is parallel to B
C
, we see that C
B
C
B
is actually a rectangle.
We claim that in a rectangle, the diameters are equal. The claim follows
once we notice that ΔC
B
C
∼
= ΔB
C
B
, the right triangles formed by
the two diﬀerent diameters of rectangle C
B
C
B
.
Next, we turn our attention to the quadrilateral C
A
C
A
. An argument
similar to the one just advanced shows that this is a rectangle as well. It
shares a diameter with C
B
C
B
, viz., C
C
. The three diameters C
C
,
A
A
, B
B
, are concurrent so are diameters of a circle with center this point
of concurrency.
Finally, let D be the foot of the altitude from A to BC. Notice that
∠A
DA
is a right angle, so D must lie on the circle with diameter A
A
.
The same argument applies to the other three feet. This completes the
proof.
50
2.4 The Projective Plane
The material here is taken mostly from [3].
The title of this subsection is a bit misleading: there are many diﬀerent
types of projective planes. From our point of view right now, though, the
projective plane is what you get when you add ideal points to the Euclidean
plane, as developed via using Hilbert’s axioms. The eﬀect of adding these
points is to remove the special status of parallel lines. In other words, in
a projective plane, any two lines intersect. The formal construction is as
follows.
Allowing that a line in the Euclidean plane is parallel to itself, we make
parallelism an equivalence relation. The lines in the Euclidean plane are thus
partitioned into nonempty, nonintersecting equivalence classes, each class
determined by a direction. All the lines in given class then have the same
direction. Add a single point, called an ideal point, to the Euclidean plane for
each equivalence class of lines. (To distinguish them from the ideal points,
call the points of the original Euclidean plane ordinary points.) Decree that
all the lines in the class contain the associated ideal point. Further decree
that the collection of ideal points comprise a line, the ideal line or the line
at inﬁnity. The resulting collection of points and lines is the real projective
plane.
Theorem 25. Two distinct points determine a unique line.
Proof. If the two points are ordinary points, the theorem is just a restatement
of two of Hilbert’s incidence axioms. If the two points are ideal, they both lie
on the line at inﬁnity. Every ordinary line in the plane contains exactly one
ideal point, so the line at inﬁnity must be the only line containing a given
pair of ideal points. If one point is ordinary and the other ideal, the ideal
point identiﬁes a single parallel class, by construction. We claim that the
ordinary point belongs to exactly one line in that class.
To prove the claim, pick a line out of the parallel class associated to the
ideal point. If the ordinary point is on that line, we are done. If not, construct
a parallel to the chosen line through the ordinary point. By Hilbert’s Parallel
Axiom, there is only such parallel. This proves the claim and the theorem.
Theorem 26. Two distinct lines meet in one and only one point.
51
Proof. If the lines are both ordinary, then Hilbert’s Theorem 1 says that
they intersect in exactly one point or they do not intersect. If they do not
intersect, they belong to the same parallel class so intersect in a single ideal
point. If one line is ordinary and the other ideal, the ordinary line belongs
to a single parallel class so intersects the ideal line in the unique ideal point
associated to that parallel class.
The critical nature of these two theorems is that they establish once and
for all that we need not distinguish between ordinary and ideal points, nor
between ordinary lines and the line at inﬁnity. In the projective plane, all
points are created equal and all lines are created equal.
This is a good time to review Hilbert’s axioms with an eye towards which
might apply in this setting. We start with the last set, the axioms of conti
nuity.
V. Axioms of Continuity
1. (Axiom of Measure or Archimedes’s Axiom) If AB and CD are any
segments then there exists a number n such that n segments CD
constructed continguously (endtoend) from A, along the ray from
A through B, will pass beyond the point B.
2. (Axiom of Completeness) An extension of a set of points on a line with
its order and congruence relations that would preserve the relations
existing among the original elements as well as the fundamental prop
erties of line order and congruence that follows from Axioms IIII and
V.1 is impossible.
The Axiom of Completeness is particularly germane. It guarantees that
we cannot add a single point to a line in Euclidean space without losing
fundamental properties. By our choice of ideal points, we know that the
parallel axiom no longer applies. What about the others?
I. Hilbert’s Axioms of Incidence
1. Two points determine a line.
2. The line determined by two points is unique.
3. There exist at least two points on a line. There exist at least three
noncollinear points.
52
4. Three noncollinear points determine a plane. Every plane contains a
point.
5. The plane determined by three noncollinear points is unique.
6. If two points of a line lie in a plane, then every point of that line lies
in that plane.
7. Two planes with nonempty intersection have at least two points in
common.
8. There are at least four noncoplanar points.
Nothing here is violated by the addition of ideal points. We are safe in
assuming that Hilbert’s incidence axioms apply to the projective plane. The
Order Axioms are a diﬀerent matter.
II. Hilbert’s Axioms of Order
1. If A∗ B ∗ C then A, B, C are distinct points of a line and C ∗ B ∗ A.
2. Given A, B ∈ AB, there is C ∈ AB with A∗ B ∗ C.
3. Of any three points on a line there exists no more than one that lies
between the other two.
Putting these together with the axioms of continuity, we get that in a
Euclidean plane, there is a onetoone correspondence between the points on
a line and the real numbers. (There is actually quite a bit more to this than
we are explicating here. This is part of a course on the real line.) Then when
we write A ∗ B ∗ C, it either means x < y < z or z < y < x, where x, y,
z are the real numbers associated to the points A, B, C. If we add a single
point to a line, it cannot be associated to a number. We could think of an
ideal point as representing ∞ but for the purposes of order, we need to ﬁnd
a position for it on a line. Think of putting it inﬁnitely far to the right, for
example, on a line containing an ordinary point A. If we hang onto the order
axioms, then we must be able to ﬁnd B so that A∗∞∗B. This is a logistical
problem that does not seem easily resolved as B would presumably occupy a
position farther to the right than “inﬁnitely far.” A similar problem ensues if
we try planting ∞ inﬁnitely far to the left on an ordinary line. If we persist,
53
looking for another location on a line for an ideal point, the only positions
left are between ordinary points. Eﬀectively, those points are already spoken
for by real numbers.
Once we accept that the order axioms have to go, we are stuck with a
world that has no rays, no angles, no segments. The notion of congruence is
now gone.
The reader should be warned here: there are ways of deﬁning metric
systems on the projective plane, in particular, the FubiniStudy metric is a
commonly used device. But there is no natural extension of these concepts
as they apply in Euclidean geometry.
Axioms for the Real Projective Plane
1. There exists a line.
2. Each line has as least three points.
3. There exist four points, no three of which are collinear.
4. Two distinct points determine a unique line.
5. Two distinct lines determine a unique point.
6. There is a onetoone correspondence between the real numbers and all
but one point of a line.
The last axiom is what make this the real projective plane. Without it,
we get more general projective planes.
Exercise
Verify that the object in Figure 29 represents a projective plane, that is,
that it satisﬁes the ﬁrst ﬁve axioms above. Identify the points and the lines.
How many of each do we have? This is called the Fano plane.
2.4.1 Duality
The most striking feature of projective geometry is the principle of dual
ity. This can be illustrated in the projective plane by restating the axioms
switching the words “point” and “line,” and switching the phrases “lie on”
and “meet in.” When we do that we get:
54
Figure 29: The Fano Plane
Dual Axioms for the Projective Plane
1. There exists a point.
2. Each point belongs to at least three lines.
3. There are four lines, no three of which are concurrent.
4. Two distinct lines determine a unique point.
5. Two distince points determine a unique line.
6. There is a onetoone correspondence between the real numbers and all
but one line going through a point.
Note that these could serve equally well to deﬁne the real projective plane.
Exercises
1. Prove that the Axioms and Dual Axioms for the Projective Plane are
indeed equivalent.
2. Write the dual of each of the following statements in the projective
plane.
(a) The set of all points on a line
(b) Four points, no three of which are collinear
(c) Two lines, and a point on neither line
55
(d) The line determined by a given point and the point of intersection
of two given lines
(e) All lines in the projective plane
(f) Distinct concurrent lines have only one common point.
(g) Given a line and a point not on the line, distinct lines through the
given point meet the given line in distinct points.
Deﬁnition 12. A triangle is a set of three noncollinear points (vertices) and
the lines they determine (sides). A trilateral is a set of three nonconcurrent
lines and the points they determine.
It should be clear that triangles are selfdual, in other words, triangles
and trilaterals are the same objects.
Deﬁnition 13. A complete quadrangle is a conﬁguration in the projective
plane composed of a set of four points (vertices), no three of which are
collinear, together with the six lines (sides) they determine. Opposite sides
of a complete quadrangle are a pair of sides without a common vertex.
The dual of a complete quadrangle is a complete quadrilateral.
Exercise
Is a complete quadrangle selfdual? Explain.
Desargues’s theorem is major result of projective geometry having to
do with relationships of triangles in a plane or in space. The proof of the
theorem relies on embedding the projective plane in projective space. This is
not a mere device to facilitate the proof of the theorem. It is, in fact, central
to the validity of the theorem. In other words, there are projective planes
that cannot be embedded in space and for these, Desargues’s theorem is not
true. These planes are called nondesarguean. The real projective plane is
desarguean. To prove that, we need some terminology and some information
about the structure of projective space.
2.4.2 Projective Space
We construct real projective three space, RP
3
as follows. Start again by
thinking of the lines of Euclidean three space as partitioned into equivalence
classes determined by direction. Add a point to each parallel class and say
56
that point lies on all lines in the associated parallel class and on no line from
any another parallel class. As above, call the new points ideal points and the
points from Euclidean space ordinary points.
The collection of ideal points in a given plane comprises a line at inﬁnity
or an ideal line.
An ideal point P lies in a given plane α if and only if α contains a line
in the parallel class associated to P. If α
is another plane parallel to α, it
contains lines in the parallel class associated to P so contains P as well. It
follows that we are adding one collection of ideal points associated to each
parallel class of ordinary planes. We decree that this collection of points
comprise an ideal line. Since all the points of this ideal line lie in all the
planes belonging to a given parallel class, the ideal line also lies in all the
planes of the parallel class. Finally, the set of ideal points and lines together
are to comprise an ideal plane, which contains no other points or lines.
Theorem 27. Two distinct planes meet in a unique line.
Proof. Suppose the planes α
1
and α
2
are nonparallel and ordinary. By
Hilbert’s system, they intersect in a unique ordinary line, . To this, we
must append the ideal point P that identiﬁes the parallel class of . While it
is clear that α
1
∩ α
2
contains no other ordinary points, maybe it contains a
second ideal point, P
. If it does, let
1
∈ α
1
and
2
∈ α
2
be ordinary lines in
the parallel class determined by P
. Since P
,∈ , neither
1
nor
2
is parallel
to but
1

2
. Since each of
1
and
2
is coplanar with , each must meet
in an ordinary point. Say
1
∩ = P
1
, and
2
∩ = P
2
.
Note that
1
and
2
determine an ordinary plane, α. The unique ordinary
plane determined by and
1
is α
1
and the unique ordinary plane deter
mined by and
2
is α
2
so α cannot contain . It follows that the ordinary
intersection ∩ α is at most one point, implying that P
1
= P
2
, leading us to
a contradiction, namely, that
1
,
2
. We conclude that there cannot be a
second ideal point in the intersection of the two planes in this case.
Suppose α
1
and α
2
are ordinary and parallel. The result follows by our
construction of RP
3
in this case, as well as in the remaining case where α
1
is ordinary and α
2
is ideal.
Theorem 28. Two distinct points determine a unique line.
Proof. The proof of Theorem 25 is largely valid here. The case we must
speak to is when both points are ideal since there is more than one ideal line
in RP
3
.
57
Let P and Q be ideal points. By the deﬁnition of ideal points, it is clear
that there are no ordinary lines containing both P and Q. Take any ordinary
point O and consider the ordinary lines , determined by O and P, and m,
determined by O and Q. It is clear that and m are distinct thus determine
an ordinary plane α. Now P and Q belong to the ideal line in α. If they
belonged to a second ideal line, this line would have to belong to α as well,
contradicting the last result.
Theorem 29. Three noncollinear points determine a unique plane.
Proof. Let A, B, C be the three noncollinear points. If all are ideal, they
determine the unique ideal plane by construction. Suppose A is ordinary.
Let = AB and m = AC. Since they contain an ordinary point, A, and
m are ordinary lines and they are distinct since A, B, C are noncollinear
by assumption. Since two ordinary lines determine a unique ordinary plane,
this proves the result.
Theorem 30. Three distinct planes not containing a common line intersect
in a unique point.
Proof. The proof is left as an exercise.
Theorem 31. If two distinct lines meet in a point, they determine a unique
plane.
Proof. If both lines are ordinary, they determine a unique ordinary plane by
Hilbert. If both are ideal, they determine the unique ideal plane by con
struction. Suppose one is ordinary and one ideal. The ideal line determines
a parallel class of ordinary planes. Since the ordinary line intersects the ideal
line, the ordinary line must belong to one of these parallel planes. But it can
only belong to one so this is the unique plane the two lines determine.
Theorem 32. A line and a plane not containing the line meet in one and
only one point.
Proof. The ideal plane contains all ideal lines so line and plane cannot both be
ideal. If both are ordinary, then the set of ordinary points in their intersection
is either empty or a single point. If empty, then the line is parallel to a single
parallel class of lines in the plane so the line and plane share the single ideal
point associated to that class of parallels. If the intersection contains a single
ordinary point, the given line cannot be parallel to any lines in the plane so
58
there can be no more than the single ordinary point in the intersection of the
line and plane.
If the line is ordinary and the plane ideal, the line intersects the plane
in the unique ideal point associated to its parallel class. If the line is ideal
and the plane ordinary, then consider the parallel class of planes associated
to the line. The given plane is not in that parallel class so intersects any of
the planes in it in an ordinary line. This line belongs to a unique parallel
class and the associated ideal point must be the unique point of intersection
of the ideal line and the ordinary plane.
Theorem 33. A line and a point not on the line determine a unique plane.
Proof. If both are ordinary, the result follows by Hilbert. If both are ideal,
they determine the unique ideal plane. If the point P is ideal and the line
ordinary, consider the ideal line determined by P and the ideal point of
. This ideal line determines a parallel class of ordinary planes exactly one
of which contains . This is the unique plane determined by and P. Now
suppose P is ordinary and ideal. There is exactly one ordinary plane in the
parallel class determined by that contains P.
Exercise Prove that if two points lie in a plane, the line they determine
lies in that plane.
Now we see that we need not distinguish between ordinary and ideal
points, lines, and planes in projective 3space.
2.4.3 Duality in Projective 3Space
The duality principle applies to all projective spaces, not just real projective
spaces, that is, those based on R, like RP
2
and RP
3
. In projective 3space,
points and planes are dual, while lines are selfdual. If the dimension of
a point is 0, of a line is 1, a plane 2, etc. then duality in projective n
space allows us to switch points with n − 1dimensional objects, lines with
n−2dimensional objects, etc. In other words, the switch is allowed between
objects when the sum of their dimensions is one less than the ambient space.
Exercise
1. Arrange the theorems in the previous subsection in dual pairs.
59
2. Write the dual for each of the following, as statements or objects in
projective 3space.
(a) the set of points on a line
(b) the set of points on a plane
(c) the set of planes containing a given line
(d) the set of lines passing through a common point
(e) the set of lines lying on a given plane
(f) the set of planes in space
(g) the set of lines in space
(h) the set of all points in space
2.4.4 The Theorem of Desargues
Deﬁnition 14. Two triangles ΔABC, ΔA
B
C
are perspective from a point
O provided AA
, BB
, CC
are concurrent at O. In this case, O is a center of
perspectivity. The triangles are perspective from a line provided the points
AB ∩ A
B
= P
1
, AC ∩ A
C
= P
2
, and BC ∩ B
C
= P
3
are collinear. In
this case, is an axis of perspectivity.
Theorem 34 (Desargues). In projective 3space, two triangles are per
spective from a point if and only if they are perspective from a line.
Proof. For the ﬁrst part of the proof, we assume that the two triangles are
not coplanar. Say ΔABC determines a plane α and ΔA
B
C
determines α
.
If ΔABC and ΔA
B
C
are perspective from a point O since AA
and
BB
intersect, they determine a plane which contains AB and A
B
which
thus intersect in a point P
1
. Note that P
1
∈ α ∩ α
. Similarly, AA
and
CC
determine a plane so that AC ∩ A
C
= P
2
with P
2
∈ α ∩ α
. Finally,
BC ∩ B
C
= P
3
∈ α ∩ α
. It follows that P
1
, P
2
, P
3
lie on the line α ∩ α
.
Next suppose ΔABC and ΔA
B
C
are perspective from a line. As in
tersecting lines, AB and A
B
determine a plane π in which AA
and BB
intersect. Similarly, AA
and CC
intersect in a plane π
; and BB
and CC
intersect in a third plane π
. Since A, B, C are noncollinear, the planes
π, π
, π
intersect in a point, O. Now AA
= π ∩ π
, BB
= π ∩ π
, and
CC
= π
∩ π
. It follows that the three lines AA
, BB
, CC
are concurrent
at O.
60
A
A'
P1
B
B'
C
C'
O
P2
P3
Figure 30: The theorem of Desargues is true in Euclidean space as long as
all the intersections occur.
For the second part of the theorem, assume that ΔABC and ΔA
B
C
lie in a plane π. Suppose the two triangles are perspective from a point
O ∈ π. Let be a line through O, not lying in π. Choose P, P
distinct
points on , both diﬀerent from O. The lines PA and P
A
must intersect in
a point A
since they lie in the plane determined by intersecting lines AA
and PP
. Produce points B
, C
in a similar manner. Notice that the points
A
, B
, C
cannot be collinear: if they were, the planes determined by P
and each of the three sides of the triangle ΔABC would coincide, implying
that A, B, C were collinear. We thus have a new triangle ΔA
B
C
. By
construction, ΔABC and ΔA
B
C
are perspective from P, while ΔA
B
C
and ΔA
B
C
are perspective from P
. The line of perspectivity in both
cases is the intersection of π with the plane of ΔA
B
C”. This gives us that
A
B
∩ AB and A
B
∩ A
B
fall on this line of intersection. Since A
B”
meets this line just once, AB and A
B
intersect there as desired. The other
sides of the two triangles meet on this line as well, for the same reasons.
We leave the last part of the proof to the reader.
Exercise Figure 30 shows what is called a Desarguean conﬁguration, an
arrangement of ten points. Show that any one of the ten points can serve as
61
the center of perspectivity for a Desarguean conﬁguration with the same
ten points.
3 Aﬃne Planes
The most familiar algebraic structure is possibly the vector space, in partic
ular, R
n
. In this section, our goal is to understand how we can use vector
spaces to model Euclidean and projective spaces. The ﬁrst thing is to es
tablish what points, lines, and planes should be and then, that they satisfy
Hilbert’s axioms.
We take vectors to play the role of points. Playing the role of lines, we
have cosets of one dimensional spaces.
Deﬁnition 15. Let V be a vector space, W a subspace of V . Let u ∈ V .
The Wcoset determined by u is u + W = ¦u + w [ w ∈ W¦.
Denote the span of a set of vectors ¦v
1
, . . . , v
n
¦ by < v
1
, . . . , v
n
>.
The results that follow are from [5], starting with Lemma 3.2.1, p. 101.
Lemma 5. For any vector space V with subspaces W, W
, and vectors u, u
∈
V ,
1. u + W = u
+ W if and only if u −u
∈ W;
2. u + W ∩ u
+ W is empty or u + W = u
+ W;
3. if dimV = 2 and dimW = dimW
= 1, u +W ∩u
+W
= if and only
if W = W
.
Proof. 1. u+W = u
+W if and only if for every w
1
∈ W, there is w
2
∈ W
with u + w
1
= u
+ w
2
which is true if and only if u −u
∈ W.
2. If v ∈ u + W ∩ u
+ W, then there are w
1
, w
2
∈ W with v = u + w
1
=
u
+ w
2
, which is true only if u −u
∈ W, i.e., u + W = u
+ W.
3. If W =< w > and W
=< w
>, then W ,= W
means precisely that
¦w, w
¦ is a basis for V . In this case, suppose u = αw + βw
and u
=
α
w+β
w
. Take v = α
w+βw
and notice that u−v = (α−α
)w ∈ W,
and u
−v = (β
−β)w
∈ W
. It follows that v ∈ u + W ∩ u
+ W
.
62
Lemma 6. Two distinct vectors v, v
determine a unique 1dimensional
coset in V .
Proof. Let W =< v −v
>. Since v −v
∈ W, v
+(v −v
) = v ∈ v
+W so by
Lemma 5.1, v+W = v
+W. Suppose u+W
is another one dimensional coset
containing v and v
. The again by Lemma 5.1, v + W
= v
+ W
= u + W
,
which implies that v −v
∈ W
, thus that W
=< v −v
>= W.
Lemma 7. Two one dimensional cosets intersect in at most one point.
Proof. Suppose v, v
∈ u + W ∩ u
+ W. Then v = u + w
1
, some w
1
∈ W
and v
= u + w
2
, some w
2
∈ W so that v − v
∈ W. Similarly, v − v
∈ W
implying either v = v
or < v −v
>= W = W
.
Deﬁnition 16. A ﬁeld is a set F with at least two distinct elements, 0 and
1, together with two commutative and associative binary operations ⊕ and ¸
such that the following axioms are satisﬁed.
1. 0 is an identity for ⊕, i.e., 0 ⊕a = a for all a ∈ F.
2. For every a ∈ F, there is a
∈ F such that a ⊕a
= 0.
3. 1 is an identity for ¸, i.e., 1 ¸a = a for all a ∈ F.
4. Every element a ∈ F, except 0, there is ˆ a ∈ F such that a ¸ ˆ a = 1.
5. For all a, b, c ∈ F, a ¸(b ⊕c) = (a ¸b) ⊕(a ¸c).
Example 1. R is an example of an ordered ﬁeld because aside from satisfy
ing the axioms for a ﬁeld, it has an ordering that respects the ﬁeld operations
of addition and multiplication. In particular, a < b implies that a +c < b +c
and if 0 < a, then b < c implies that ab < ac.
Our favorite examples of ﬁelds besides R are Q, C and the integers modulo
p, as in the next example.
Example 2. Let Z
p
= ¦
¯
0,
¯
1, . . . , p −1¦ designate the set of residue classes
of integers mod p. That is,
¯
0 = ¦kp [ k ∈ Z¦,
¯
1 = ¦1 + kp [ k ∈ Z¦,
¯
2 = ¦2 +kp [ k ∈ Z¦, etc. Let ¯ a ⊕
¯
b := a + b and let ¯ a ¸
¯
b = ab. We leave it
as an exercise to verify that ⊕ and ¸ are welldeﬁned and that they satisfy
the axioms for a ﬁeld.
63
Now we are free to consider vector spaces with arbitrary ﬁelds of scalars.
Unless noted to the contrary, we assume V is a vector space deﬁned over
some ﬁeld F.
Consider that if u+W is a one dimensional coset in V and v ∈ u+W, then
v −u ∈ W implies that W =< v −u >. It follows that for any v
∈ u + W,
there is α ∈ F such that v
= u+α(v−u) = (1−α)u+αv. In other words, we
can coordinatize the points on the line = u +W by (α, β), where α, β ∈ F
satisfy α + β = 1. These α and β are then the aﬃne line coordinates for
v
with respect to v and u. On the other hand, for any α ∈ F, there is a
point (1 −α)u + αv ∈ u + W: the set of points on a line in V has the same
cardinality as F. In particular, since F must have at least two elements,
every line in V must have at least two distinct points.
We have proved the following.
Theorem 35. Let F be a ﬁeld with at least three elements. Let V be a
two dimensional vector space over F. The collection of vectors and one
dimensional cosets of V form the points and lines of the aﬃne plane Π(F),
which satisﬁes Hilbert’s axioms of incidence for a plane.
In the ﬁnite ﬁelds we have seen, there is k such that
k times
1 ⊕. . . ⊕1 = 0.
This also happens for certain inﬁnite ﬁelds, the study of which extends be
yond the scope of this course. When there is one such k, there must be a
minimal value of k for which it is true. That minimal k is called the charac
teristic of F. When there is no such k, the ﬁeld is said to have characteristic
0. R, C, and Q all have characteristic zero. Z
p
has characteristic p. It is not
diﬃcult to verify that the characteristic of a ﬁeld is either zero or a prime
number.
Assume now that the characteristic of F is not two. Then in F we have
1 ⊕1 = 2 ,= 0 so that 2 has an inverse with respect to ¸, multiplication. For
convenience, designate this inverse 1/2. Let u, v ∈ V = F
2
. Then we can
deﬁne the midpoint of u and v to be (1/2)(u + v).
We have u, v ∈ means = u+ < v − u > so that αu + βv ∈ for all
α, β ∈ F with α⊕β = 1. Does 1/2⊕1/2 = 1 in any ﬁeld F with characteristic
diﬀerent from 2? To see that the answer is yes, we invoke the distributive
law to get
(1/2 ⊕1/2) = 1/2 ¸(1 ⊕1) = 1/2 ¸2 = 1.
64
Next, we want to verify that in Π(R), (1/2)(u+v) is the same distance from
u as from v. Writing u = (u
1
, u
2
), v = (v
1
, v
2
) we have (1/2)(u + v) =
((u
1
+ v
1
)/2, (u
2
+ v
2
)/2). Using the distance formula to ﬁnd the distance
from u to (1/2)(u + v), we get
d(u, (1/2)(u + v)) =
(1/2(u
1
+ v
1
) −u
1
)
2
+ (1/2(u
2
+ v
2
) −u
2
)
2
=
(1/2(v
1
−u
1
))
2
+ (1/2(v
2
−u
2
))
2
= 1/2 d(u, v).
By symmetry, the distance is the same from u to (1/2)(u + v).
Deﬁnition 17. A triangle in Π(F) is a set of 3 noncollinear points and the
lines they determine.
Lemma 8. ¦u, v, w¦ ⊂ Π(F) is a triangle if and only if ¦v −u, w −u¦ is a
basis for F
2
.
Proof. We leave the proof as an exercise.
Lemma8 shows that if we ﬁx any triangle in Π(F), ¦u, v, w¦, and an
ordering of its vertices, then for each z ∈ Π(F), there are unique λ, μ ∈ F
with
z −u = λ(v −u) + μ(w −u).
This gives us z = (1 − (λ + μ))u + λv + μw. It follows that every point of
Π(F) has a unique triple of scalars (α, β, γ) where z = αu +βv +γw, where
α+β +γ = 1. These are the aﬃne coordinates of z with respect to ¦u, v, w¦.
Exercise
Check that when Π(F) = R
2
and ¦u, v, w¦ = ¦(0, 0), (1, 0), (0, 1)¦, the aﬃne
coordinates of a point (λ, μ) are (1 −(λ + μ), λ, μ).
We reconsider Theorem 21 in Π(F).
Exercise
Let ¦u, v, w¦ be a triangle in Π(F). A median of the triangle is a line
determined by one vertex and the midpoint of the line determining the
opposite side. Verify that the medians of the triangle are
m
1
= u+ < u −(1/2)v −(1/2)w >
65
m
2
= v+ < v −(1/2)w −(1/2)u >
m
3
= w+ < w −(1/2)u −(1/2)v > .
Now suppose the characteristic of F is 3. To make it easier to think about,
say F = Z
3
. In F we have 2 ⊕ 1 = 0 so −1 = 2 and 1 = −2. Referring to
the exercise above, let ¦u, v, w¦ be a triangle in Π(F) with medians m
1
, m
2
,
m
3
. Letting z = u +v +w, we can rewrite m
1
= u+ < z >, m
2
= v+ < z >,
m
3
= w+ < z >. It follows that the medians are either identical or parallel.
We claim they are parallel.
Note that z = u + v + w = −2u + v + w = (v − u) + (w − u). If
m
1
= m
2
, there is α ∈ F with u − v = α(v − u) + α(w − u). Note that
u −v = 2(v −u) +0(w −u). Since ¦v −u, w−w¦ is linearly independent, α
is unique, implying that α = 2 and α = 0, impossible as characteristic of F
is 3. It follows that the medians are distinct and parallel.
Theorem 36. Let F be a ﬁeld with characteristic not 2 or 3. Then the
medians of a triangle in Π(F) are concurrent.
Proof. Let ¦u, v, w¦ be a triangle in Π(F). Let z ∈ m
1
and say (α, β, γ) are
the unique aﬃne coordinates determined by our triangle with vertices in the
order given. We also have aﬃne line coordinates for z determined by m
1
:
z = u+δu−(1/2)δv −(1/2)δw. Note that z = (1+δ)u−(1/2)δv −(1/2)δw.
Since 1+δ−(1/2)δ−(1/2)δ = 1, it follows by uniqueness of aﬃne coordinates
that α = 1 + δ, β = (−1/2)δ, γ = (−1/2)δ. Similarly, z belongs to m
2
if
and only if there is ξ ∈ F with α = (−1/2)ξ, β = 1 + ξ, γ = (−1/2)ξ. We
conclude that m
1
and m
2
intersect if and only if the following equations can
be solved simultaneously.
(−1/2)ξ = 1 + δ
1 + ξ = (−1/2)δ
(−1/2)ξ = (−1/2)δ
By the last equation, ξ = δ. From the ﬁrst equation ξ = −2/3 = δ. The
aﬃne coordinates of the unique point of intersection of m
1
and m
2
is then
(1/3, 1/3, 1/3). We leave the rest of the proof to the reader.
66
Exercises
1. Complete the proof of the last theorem by verifying that the point
with aﬃne coordinates (1/3, 1/3, 1/3) lies on m
3
.
2. Consider the triangle ¦(0, 0), (1, 0), (0, 1)¦ ⊂ Π(Z
5
). What is the point
of concurrency of its medians?
3. Let ¦u, v, w¦ be a triangle in Π(F). Let x, y, z be points in Π(F)
with aﬃne coordinates given respectively by (α
1
, β
1
, γ
1
), (α
2
, β
2
, γ
2
),
(α
3
, β
3
, γ
3
). Show that x, y, z are collinear if and only if
det
⎛
⎝
α
1
β
1
γ
1
α
2
β
2
γ
2
α
3
β
3
γ
3
⎞
⎠
= 0.
3.1 The Aﬃne Group
Deﬁnition 18. Let G be a set of mappings on a set S. Suppose e ∈ G is
the identity mapping, that is, e(s) = s for all s ∈ S. If G is closed under
function composition, and if every element in G has an inverse in G then G
is a group.
The most fundamental example of a group is the set of permutations on
any set. The best way to understand this is to write out all the permutations
on a set of three objects ¦a, b, c¦ and see what happens if you start composing
them. In some sense, all groups are groups of permutations. This is the
content of Cayley’s Theorem, one of the big theorems you learn in a beginning
course in abstract algebra.
Groups are a critical element in the modern study of geometry.
Example 3. The set of invertible linear transformations on any vector space,
V , is a group designated GL(V ). This is easy to check and the reader is
urged to do so. When dim
F
V = n, GL(V ) can be identiﬁed with the set of
nonsingular n n matrices with entries in F. In this case we often write
GL(V ) = GL
n
(F).
Example 4. A vector space V is a group which we also designate V . In
what sense is a vector space a collection of mappings on a set? For any
v ∈ V , deﬁne t
v
: V → V by t
v
(u) = v + u, for all u ∈ V . We then have
67
t
v
◦t
w
(u) = t
v
(t
w
(u)) = t
v
(w+u) = v+w+u = t
v+w
(u) so that t
v
◦t
w
= t
v+w
.
The identity mapping is t
θ
, where θ is the zero vector. The inverse of t
v
is t
−v
.
Notice that unlike GL(V ), V is a commutative group, that is, while function
composition is generally not commutative, t
v
◦ t
u
= t
v+u
= t
u+v
= t
u
◦ t
v
.
Any group in which the operation is commutative like this is called an abelian
group.
Deﬁnition 19. Let V = F
2
. Fix T ∈ GL(V ) and v ∈ V . Deﬁne A(T, v) :
V → V by A(T, v)(u) = T(u) +v, for all u ∈ V . The aﬃne group Aﬀ(V ) is
the set ¦A(T, v)[ T ∈ GL(V ), v ∈ V ¦.
As a set, Aﬀ(V ) is the Cartesian product of GL(V ) and V , in other
words, an element of Aﬀ(V ) can be identiﬁed with an ordered pair (T, v), T ∈
GL(V ), v ∈ V . There are actually many diﬀerent ways to deﬁne “products”
of groups. The aﬃne group is an example of what is called a semidirect
product of GL(V ) and V .
Lemma 9. The aﬃne group maps lines to lines in Π(F).
Proof. Let u + αw ∈ = u+ < w >. For any T ∈ GL(V ), v ∈ V , we have
A(T, v)(u+αw) = T(u+αw)+v = T(u)+αT(w)+v ∈ v+T(u)+ < T(w) >.
Since T is an invertible linear transformation, T(w) cannot be zero if w ,= θ,
thus the image of the line under A(T, v) is a line v + T().
It is clear that if x ∈ ∩
, where and
are lines, then for any aﬃne
transformation A(T, v), A(T, v)(x) ∈ A(T, v)() ∩ A(T, v)(
).
Lemma 10. The aﬃne group maps triangles to triangles in Π(F).
Proof. By Lemma 8, ¦x, y, z¦ is a triangle in Π(F) if and only if ¦y−x, z−x¦
is a basis for V . Let A(T, v) ∈ Aﬀ(V ). We have A(T, v)¦x, y, z¦ = ¦T(x) +
v, T(y) + v, T(z) + v¦, and since T ∈ GL(V ),
¦T(y −x), T(z −x)¦ = ¦T(y) −T(x), T(z) −T(x)¦ = ¦T(y) + v −(T(x) + v), T(z) + v −(T(x) +
is also a basis for V . Lemma 8 implies that A(T, v)¦x, y, z¦ is a triangle.
Theorem 37. Let and
be lines in Π(F). There is an aﬃne transforma
tion A(T, v) such that A(T, v)() =
. If ¦x, y, z¦ and ¦x
, y
, z
¦ are trian
gles, there is a unique A(T, v) ∈ Aﬀ(V ) with A(T, v)(¦x, y, z¦) = ¦x
, y
, z
¦.
68
Proof. Let = x+ < y > and
= x
+ < y
> be distinct lines in Π(F).
Choose T ∈ GL(V ) so T(y) = y
. Let v = x
− T(x). Then A(T, v)(x+ <
y >) = T(x) + v+ < T(y) >= x
+ < y
>.
Given triangles ¦x, y, z¦ and ¦x
, y
, z
¦, there is a unique element T ∈
GL(V ) that maps ¦y −x, z −x¦ to ¦y
−x
, z
−x
¦. Consider that
A(T, −T(x) + x
)¦x, y, z¦ = ¦x
, T(y) −T(x) + x
, T(z) −T(x) + x
¦ = ¦x
, y
−x
+ x
, z
−x
+ x
If A(T, v)(x) = T(x) + v = x
, then v is uniquely determined. The result
follows.
Going back to our deﬁnition of “group,” we see it is easy to verify that
if G is a group of mappings on a set S, and if S
⊂ S, then the set H ⊂ G
of elements in G that map S
→ S
forms a group inside G. Thus, H is a
subgroup of G. Elements of H are said to leave S
invariant.
Let H be the subgroup of Aﬀ(V ) that leaves a given triangle invariant,
that is, that maps the triangle to itself with a permutation of its vertices.
The collection of H cosets is ¦g ◦ h[g ∈ Aﬀ(V ), h ∈ H¦. These can actually
serve as triangles. This is typical of the Erlangen Program of Felix Klein, a
hugely inﬂuential idea described in a lecture of Klein’s dating to 1872. At
that point in history, geometry was in a state of disarray. The notion that
groups could and should serve as a unifying force in geometry turned out to
be profound. It remains at the heart of a great deal of mathematics being
discovered today.
Exercises
Deﬁne a complete quadrilateral in Π(F) to be a set of four points, no three
of which are collinear, and the lines they determine. The four points are the
vertices of the quadrilateral. A quadrilateral ¦u, v, w, x¦ is a parallelogram
provided its vertices can be relabeled if necessary so that
u+ < v −u > x+ < w −x > and u+ < x −u > v+ < w −v >. The lines
u+ < w −u > and v+ < x −v > are the diagonals.
1. Prove that a quadrilateral ¦u, v, w, x¦ is a parallelogram if and only if
one of its vertices is the sum of the other two minus the third.
2. Let ¦u, v, w, x¦ be a parallelogram in Π(F), where the characteristic of
F is not two. Show that the diagonals of the parallelogram bisect one
another.
69
3. Prove that in an aﬃne plane over a ﬁeld of characteristic not two,
every parallelogram is the image under some aﬃne transformation of
¦(0, 0), (2, 0), (2, 2), (0, 2)¦.
4. Consider Π(C), the aﬃne plane over C. Let ϕ(z
1
, z
2
) = ( ¯ z
1
, ¯ z
2
), where
¯ z is the complex conjugate. (See below for more details, if this is
unfamiliar.) Prove that ϕ sends lines to lines and triangles to triangles
but that it is not an aﬃne transformation.
4 The Quaternions
This section represents something of a digression. Our objective is to describe
how the quaternions deﬁne rotations in R
3
. By the end of the section, we
should see that it is not as much of a digression as it ﬁrst appears. The
quaternions as rotations turn out to be an instance of a group acting on the
objects in a geometry. In the next section, we will take up the same topic in
the real projective plane.
Hamilton discovered the quaternions after an arduous search inspired by
the beautiful connection he discovered between the complex numbers and the
geometry of R
2
. Recall that the complex numbers, C = ¦a+bi [ a, b ∈ R¦ can
be identiﬁed with the points in R
2
= ¦(a, b) [ a, b ∈ R¦. This is signiﬁcant
because it allows C and R
2
to borrow structure from one another. To see this,
note ﬁrst that we can multiply elements of C simply using the distributive
law and the deﬁnition of i via i
2
= −1. Multiplication is commutative (easy
to check) and associative (tedious to check.) Recall further that the complex
conjugate of z = a + bi ∈ C is ¯ z = a − bi. We then have z¯ z = a
2
+ b
2
, a
nonnegative real number. This gives us the modulus or norm or absolute value
of z ∈ C, z =
√
a
2
+ b
2
. Notice that z is the distance from the associated
point in R
2
to the origin. We deﬁne division in C by z/w = z ¯ w/w
2
. This
makes sense whenever w ,= 0 and now we have that C is a ﬁeld.
Consider the collection of unit complex numbers ¦z ∈ C [ z = 1¦. Note
that this is precisely the unit circle in R
2
, usually designated S
1
. S
1
is the
one sphere, that is, the sphere which is one dimensional. S
1
is not ﬂat, it
is not a vector space, so you don’t have a precise idea of what “dimension”
means for such an object. A safe way to think about it is as follows: the
dimension of the object is the dimension of its tangent space. The tangent
space to a circle is a line, which is one dimensional, therefore the dimension
70
of S
1
is one.
If z ∈ S
1
, then z = cos θ +i sin θ for some θ. In fact, it is an easy exercise
in trigonometry to show that if w ∈ C, w = w(cos ϕ + i sin ϕ) . With z
and w in hand, we have
z.w = w(cos θ cos ϕ −sin θ sin ϕ) + i(sin θ cos ϕ + cos θ sin ϕ) =
w(cos(θ + ϕ) + i sin(θ + ϕ)).
We see then that multiplication by a unit complex number eﬀects a rotation
on R
2
: it does not change the distance from a point to the origin but it does
change its angle.
Hamilton wanted to extend this story to R
3
and came up empty handed
until he realized, at a famous moment when he was crossing a bridge in
Dublin, that he needed to view R
3
as sitting inside a four dimensional analog
to C, the quaternions H.
Deﬁne H = ¦q = a + bi + cj + dk [ a, b, c, d ∈ R¦ where i, j, k enjoy no
relations among themselves except for the following and their consequences:
i
2
= j
2
= k
2
= −1
ij = k, ji = −k, jk = i, kj = −i, ki = j, ik = −j.
Using these relations and the distributive law, we can deﬁne multiplication
of quaternions. Obviously multiplication is not commutative so H is not a
ﬁeld. It does, however, enjoy all the other properties of a ﬁeld so is a skew
ﬁeld.
As in C, we have conjugation in H given by ¯ q = a − bi − cj − dk. Then
q¯ q = a
2
+b
2
+c
2
+d
2
and the modulus of q is q =
√
q¯ q. Again, this allows
us to deﬁne division: q/w = q ¯ w/w
2
, for q, w ∈ H.
H can be identiﬁed with R
4
. Making the identiﬁcation, we write q =
a + bi + cj + dk = (a, b, c, d). The pure quaternions are those of the form
(0, b, c, d). Notice that we can designate any quaternion by (a, v), a ∈ R,
v ∈ R
3
. This is nice because it lets us write a rather involved looking
formulation that actually helps us prove some results.
If q = (a, v), and w = (b, u) are in H, then using the distributive law you
should have no trouble verifying that
qw = (ab −v u, au + bv +v u).
71
Here v u is the usual dot product in R
3
, that is, for v = (v
1
, v
2
, v
3
) and
u = (u
1
, u
2
, u
3
), v u = v
1
u
1
+v
2
u
2
+v
3
u
3
. v u is likewise the usual cross
product in R
3
, where i, j, k now play the roles of the standard unit vectors
in R
3
.
v u = det
⎡
⎣
i j k
v
1
v
2
v
3
u
1
u
2
u
3
⎤
⎦
We can identify R
3
with the pure quaternions and then the unit pure
quaternions become S
2
, the 2sphere, that is, the surface of a ball. The
unit quaternions can be identiﬁed with S
3
, the 3sphere, an object that has
inspired extensive research for over one hundred years.
Where are the rotations? Let R
θ,v
be the rotation of R
3
about the axis
determined by the unit vector v, through the angle θ. Notice that there are
two choices for v. If it makes things any easier we can always insist that if
v points “up”, then the rotation is counterclockwise.
The mapping u → quq
−1
is called conjugation by q. Conjugation comes
up frequently in group theory and in geometry.
Theorem 38. R
θ,v
is eﬀected by conjugating any element of R
3
, viewed as
a pure quaternion, by (cos(θ/2), sin(θ/2).v).
To prove/understand the theorem, the ﬁrst thing we do is verify that
conjugation maps pure quaternions to pure quaternions. For u = (0, u) and
q = (a, v),
quq
−1
= (a, v)(0, u)(a, −v) =
(−v u, au +v u)(a, −v) = (−av u + av u, . . . ) = (0, . . . )
where it doesn’t matter what . . . is, since we just wanted to show that u is
mapped to another pure quaternion.
An obvious check to whether or not q = (cos(θ/2), sin(θ/2).v) really is
a rotation of R
3
with axis < v > is to see what qvq
−1
gives us. Bearing in
mind that v is a unit vector, we have v v = 1. Also, v v = 0, the zero
vector. Then
qvq
−1
= (cos(θ/2), sin(θ/2).v)(0, v)(cos(θ/2), −sin(θ/2).v) =
72
(−sin(θ/2), cos(θ/2).v)(cos(θ/2), −sin(θ/2).v) =
(−sin(θ/2) cos(θ/2) + sin(θ/2) cos(θ/2), cos
2
(θ/2).v + sin
2
(θ/2)v) = (0, v).
This proves that v is ﬁxed under conjugation by q. This is promising, but
not a proof.
To ﬁnish the proof, we think about a matrix representation of R
θ,v
. Ma
trix representations depend on choice of ordered basis. A good basis to use
here would be ¦v, u, w¦ where ¦v, u, w¦ form a righthanded coordinate sys
tem of orthonormal vectors. This means that all three basis elements are unit
vectors, that they are orthogonal, in other words, have zero dot product, and
that v u = w, u w = v, and w v = u. In short, we can think of v,
u, w as arranged like the unit vectors in the x, y, z directions in R
3
. This
allows us to visualize R
θ,v
as a counterclockwise rotation in the yz plane.
The matrix representation is then
⎡
⎣
1 0 0
0 cos θ −sin θ
0 sin θ cos θ
⎤
⎦
To see this, think of where the unit vectors in the y and z direction are
mapped under the rotation. In particular, we have
v → v, u → cos θ.u + sin θ.w, w → −sin θ.u + cos θ.w.
Now we check:
quq
−1
= (cos(θ/2), sin(θ/2).v)(0, u)(cos(θ/2), −sin(θ/2).v) =
(0, cos(θ/2).u + sin(θ/2).w)(cos(θ/2), −sin(θ/2).v) =
(0, cos
2
(θ/2)u + cos(θ/2) sin(θ/2)w−sin(θ/2) cos(θ/2).u v −sin
2
(θ/2)wv) =
0, (cos
2
(θ/2) −sin
2
(θ/2).u + 2 sin(θ/2) cos(θ/2).w) = (0, cos θ.u + sin θ.w).
We leave it as an exercise to show that conjugation by q maps w →
−sin θ.u + cos θ.w. That completes the proof of the theorem.
73
Exercises
1. Show that conjugation by q ∈ H, q ,= 0, deﬁnes a linear transformation
on R
3
.
2. In the proof of Theorem 38, why did we not have to show that every
vector in R
3
underwent a rotation?
3. What conditions would we have to impose on q to get that conjugation
by q on pure quaternions preserves norm? In other words, when does
qwq
−1
have the same norm as w, for q ∈ H, w pure quaternion?
4. Find the explicit expression for q to eﬀect the rotation in R
3
through
30
◦
, about an axis that is normal to the plane x +y +z = 0. Take the
rotation that is counterclockwise when viewed along the axis from the
ﬁrst octant. Use q to determine where the point (−1, 1, 2) is mapped
to under the rotation. Give exact coordinates.
5 The Real Projective Plane
We consider here one model for the real projective plane, RP
2
.
Theorem 39. Let F be any ﬁeld and consider the vector space V = F
3
. By
deﬁning a projective point as a 1dimensional subspace of V , a projective line
as a 2dimensional subspace of V , and incidence as nonempty intersection,
we get a set, Π(V ), that satisﬁes the axioms for a projective plane.
Proof. We have four axioms to prove: the two that show nontriviality and
the two incidence axioms. Since F contains at least two elements, 0 and 1,
we have the following distinct points in Π(V ): < (1, 0, 0) >, < (0, 1, 0) >,
< (0, 0, 1) >, and < (1, 1, 1) >. We claim that no three of these are collinear,
that is, incident to the same line.
The line determined by < (1, 0, 0) > and < (0, 1, 0) > is the span of
¦(1, 0, 0), (0, 1, 0)¦, that is, the set of all vectors in V of the form (α, β, 0),
α, β not both zero in F. Since this set contains neither (0, 0, 1) nor (1, 1, 1),
neither of the latter points is collinear with < (1, 0, 0) > and < (0, 1, 0) >.
A similar argument applies to the other sets of three vectors. This veriﬁes
that Π(F) has four points, no three of which are collinear.
74
Next we consider a line in Π(F) and verify that it has three points.
Every line in Π(F) is determined by a linearly independent set of vectors
¦u, v¦ ⊂ F
3
. Note that the span of this set contains lines < u >, < v > and
< u +v >. It is easy to verify that these are distinct, thus, they correspond
to three distinct points on the line determined by u and v.
Next, let be determined by the span of ¦u, v¦ and
be determined by
the span of ¦w, x¦. Since planes through the origin in F
3
must intersect in
a unique line through the origin, and
intersect in a unique projective
point. Similarly, if < u > and < v > determine distinct points in Π(F), then
¦u, v¦ is a linearly independent set, thus spanning a 2dimensional subspace
of F
3
, which determines the unique line in Π(F) determined by < u > and
< v >.
There are other models for RP
2
. One that can be useful for visualization
is the unit sphere, S
2
, with antipodal points identiﬁed. In other words, a
point of RP
2
is realized as the pair of endpoints of a diagonal of the unit
sphere in R
3
. How is a line realized with this model? To understand this
part, notice that for a pair of nonantipodal points on the sphere, there are
many circles on the surface of the sphere that contain those points but there
is only one great circle that contains those points. A way to think of this is
to realize that you get a circle on the sphere precisely by intersecting S
2
with
a plane. When you intersect S
2
with a plane through the origin, you get a
great circle. If you pick two nonantipodal points on S
2
and the origin, this is
a set of three noncollinear points, thus uniquely determines a plane through
the origin. This plane, in turn, determines a unique great circle on S
2
, which
must then contain the antipodes of both original points. This great circle
identiﬁes the projective line determined by the two projective points.
This brief discussion shows right away how the S
2
model and the subspace
model interact. In particular, a one dimensional subspace of R
3
intersects
S
2
in a unique pair of antipodal points. This is a bijective correspondence
between points in the two models.
5.1 Homogeneous Coordinates
Designate a projective point now as [u], that is, let [x, y, z] be the projective
point corresponding to the nonzero vector u = (x, y, z). If z ,= 0, then
[u] = [x, y, z] = [x/z, y/z, 1]. In particular, the coordinates of [u] are deﬁned
only up to nonzero multiples, so are called homogeneous coordinates.
75
Homogeneous coordinates are especially convenient for localizing, that
is, for thinking of points that are somehow “near” one another in RP
2
as
belonging to an aﬃne space.
Deﬁnition 20. The line at inﬁnity,
∞
⊂ Π(F), is the projective line cor
responding to the space ¦(a, b, 0)[ a, b ∈ F¦.
We leave the veriﬁcation of the next theorem to the reader.
Theorem 40. The mapping ϕ : Π(F) → Π(F) given by ϕ(a, b) = [a, b, 1] is
a bijection onto the set Π(F) ¸
∞
.
This gives us a visual aid for understanding Π(F) as a subset of Π(F).
The plane z = 1 intersects all lines through the origin in F
3
except those
lying in the xyplane. In other words, < u >=< (a, b, c) > is a line in F
3
not
lying in the xyplane if and only if c ,= 0. Its intersection with the plane at
z = 1 is the image of the aﬃne point (a/c, b/c) under ϕ. Now we can think
of projective points [a, b, 1] as being near each other in the sense that they
comprise an aﬃne space.
Our choice of projective line to designate
∞
is arbitrary: we can excise
any line from Π(F) to get a copy of Π(F), that is a localization.
The ﬁrst six exercises that follow are taken from [7].
Exercises
1. Describe the line in Π(R), determined by the following pairs of points.
(a) [1, 1, 2] and [2, 3 −1]
(b) [0, 2, 1] and [−1, 1, 1]
(c) [2, 0, 1] and [−1, 1, 1].
(d) [1, 2, 0] and [2, 3, 4]
(e) [1, 2, 0] and [2, 3, 0]
2. Using the mapping described in the last theorem, ﬁnd the preimage in
Π(R), if there is one, for each of the following projective points.
(a) [2, 3, 1]
(b) [−2, −3, 1]
76
(c) [−2, −3, −1]
(d) [2, 4, 6]
(e) [2, −3, −1]
(f) [−1, 2, 0]
(g) [1/2, 1/3, 6]
(h) [1, 0, 0]
3. The following pairs of equations describe lines in RP
2
. Find the pro
jective point where each pair intersects.
(a) x + y + z = 0, and −x + 2y + z = 0
(b) 2x −y + 3z = 0 and −x + z = 0
(c) x = 0 and y = 0
(d) x = 0 and z = 0
(e) y = 0 and z = 0
(f) −x + 5y −z = 0 and 2x + 7y −3z = 0
4. Find the image of each of the following lines in Π(R) under the mapping
ϕ given in the theorem above.
(a) x = 7
(b) y = 8
(c) x + y = 4
(d) −x + 3y + 18 = 0
(e) 2x −7y = 21
(f) −y + 3x + 5 = 0
5. Determine whether each of the following sets three projective lines is
concurrent and ﬁnd the points of concurrency.
(a) x + y + z = 0, 2x −3y + z = 0, −x + 7y −6z = 0
(b) x + 2y −z = 0, 4x −y = 0, z = 0
(c) 2x −y + 3z = 0, −4x + 2y + z = 0, x −(1/2)y + z = 0
77
6. Which of the following sets of projective points is collinear? For those
that are, describe the line.
(a) [−1, −1, 1], [3, 3, 3], [1/2, 1/2, 4]
(b) [−2, 3, 0], [1, −2, 1], [2, 0, 0]
(c) [1, 1, −1], [2, 0, −2], [0, 1, 2]
7. Prove that ¦[u], [v], [w]¦ ⊂ Π(F) is a triangle if and only if ¦u, v, w¦ is
a basis of F
3
.
5.2 The Projective Group
Recall that GL
3
(F) is the group of nonsingular linear transformations from
F
3
to itself. GL
3
(F) can be identiﬁed with 3 3 matrices over F with
nonzero determinant. At ﬁrst blush, GL
3
(F) seems a natural choice for a
group of bijections on Π(F) but we see trouble right away when we notice,
for instance, that both I
3
and −I
3
ﬁx all projective points. This means that
on Π(F), I
3
and −I
3
deﬁne the same mapping. Indeed, for any α ∈ F, αI
3
is
the identity mapping on Π(F). This suggests that we consider the following
as the natural group of bijections on Π(F).
Deﬁnition 21. The projective group, PGL
3
(F), is the collection of map
pings
¯
T : Π(F) → Π(F) where
¯
T([u]) := [T(u)] for T ∈ GL
3
(F).
We put the cart before the horse with this deﬁnition: it really only makes
any sense after we prove the following theorem.
Theorem 41. If T ∈ GL
3
(F), then
¯
T deﬁned by
¯
T[u] = [T(u)] is a well
deﬁned bijective mapping on Π(F). Moreover,
¯
I
3
is the identity mapping on
Π(F),
¯
T ◦
¯
S = T ◦ S, , and
¯
T
−1
= T
−1
.
Proof. We ﬁrst must show that if [u] = [v], then
¯
T[u] =
¯
T[v] but that is
immediate because [u] = [v] if and only if there is nonzero α ∈ F with
v = αu, in which case
¯
T[u] = [T(u)] = [αT(u)] = [T(αu)] = [T(v)] =
¯
T[v].
The statement about
¯
I
3
is clear. Let T and S belong to GL
3
(F). For any
[u] ∈ Π(F), we have
¯
T ◦
¯
S[u] =
¯
T[S(u)] = [T(S(u))] = [T ◦ S(u)] = T ◦ S[u].
78
We leave the proof of the last statement to the reader. The fact that
¯
T is
bijective follows.
Theorem 42. The projective groups maps lines to lines and triangles to
triangles in Π(F).
Proof. A line ∈ Π(F) corresponds to a two dimensional subspace < u, v >⊂
F
3
and a mapping
¯
T ∈ PGL
3
(F) corresponds to the collection of mappings
αT ∈ GL
3
(F), α nonzero in F, T bijective. T(¦u, v¦) is a linearly inde
pendent set of two vectors hence corresponds to a two dimensional subspace
< T(u), T(v) >=< αT(u), αT(v) >, that is, to a projective line
∈ Π(F).
The argument for triangles is even easier, in light of Exercise 7 at the end of
the last subsection.
Next we consider the relationship between the aﬃne and the projective
groups.
Recall that an element of Aﬀ(V ) is A(T, v), where T ∈ GL(V ), v ∈ V
and A(T, v)(u) = T(u) +v. Identifying Π(F) with the plane z = 1 in F
3
lets
us think of the elements of Aﬀ(V ) as matrices with the form
A(T, v) =
⎛
⎝
a b x
c d y
0 0 1
⎞
⎠
where T =
a b
c d
∈ GL
2
(F) and v = (x, y) ∈ F
2
. Using block matrix
form, we can write
A(T, v) =
T v
00 1
.
We then have
A(T, v) ◦ A(T
, v
) =
T v
00 1
T
v
00 1
=
TT
T(v
) + v
00 1
.
Turning now to PGL
3
(F), we suppose
¯
P is a projective transformation
that maps Π(F) to itself. As usual, we identify Π(F) with the plane z = 1
in F
3
. It is not diﬃcult to verify that
¯
P must be all nonzero scalar multiples
of a matrix of the form
P =
⎛
⎝
T a
b
0 0 1
⎞
⎠
79
where det P = det T ,= 0. In other words, we can represent
¯
P uniquely with
a matrix of this form, that is, by an aﬃne transformation.
Theorem 43. Aﬀ(F
2
) can be identiﬁed as a subgroup of PGL
3
(F) that
maps
∞
to itself and that ﬁxes Π(F) ⊂ Π(F).
Exercises
1. Finish the proof of Theorem 41 by showing that
¯
T
−1
= T
−1
.
2. Let T linear be deﬁned on F
3
by T(1, 0, 0) = (0, 1, 0), T(0, 1, 0) =
(0, 0, 1) and T(0, 0, 1) = (1, 0, 1).
(a) Verify that T ∈ GL
3
(F).
(b) Find
¯
T[1, 2, −1].
(c) Find
¯
T(
∞
).
(d) Does
¯
T : Π(F) → Π(F)? If so, prove it. If not, describe
¯
T(Π(F)).
3. Complete the proof of Theorem 42 by showing that PGL
3
(F) maps
triangles to triangles. Show that if we consider a triangle an ordered
triple of noncollinear points, then there is exactly one element in the
projective group that sends a given triangle to another.
4. Verify that
¯
P ∈ PGL
3
(F) maps Π(F) to itself if and only if
¯
P cor
responds to the set of all nonzero scalar multiples of a matrix of the
form
P =
⎛
⎝
T a
b
0 0 1
⎞
⎠
for some T ∈ GL
2
(F).
5. Verify that a mapping of the form given in the last problem ﬁxes
∞
.
80
5.3 Curves
Curves are a source of much comment in calculus. Typically, in that context,
you study pieces of curves, in particular, the pieces that can be described
as graphs of functions, y = f(x), for example, y =
√
x. More generally, we
deﬁne a real algebraic plane curve, that is, a curve in R
2
, as the zero set of
a polynomial in x and y. Instead of thinking of functions, we are thinking
of polynomial equations. The equation associated to the function y =
√
x
would be y
2
= x or x −y
2
= 0.
A homogeneous polynomial of degree n is a sum of monomials of degree
n. x − y
2
is an example of an inhomogeneous polynomial of degree 2. Note
that the degree of the polynomial is that of the highest degree monomial.
In Π(F), an algebraic curve is the set of zeroes of a homogeneous poly
nomial. We can understand curves in F
2
by “lifting” to Π(F) through the
process of homogenization. To homogenize a polynomial in x and y of degree
n, multiply each monomial by z
d−n
, where d is the degree of the monomial.
We homogenize a polynomial equation in x and y by ﬁrst writing it in the
form p(x, y) = 0, then homogenizing p(x, y).
Example 5. If we homogenize x − y
2
, we get xz − y
2
. To homogenize y =
x
2
− 2, rewrite to get x
2
− y − 2 = 0 then homogenize the left hand side to
get x
2
−yz −2z
2
= 0.
Exercises
1. Consider the equation 2x + 3y = 1.
(a) What does this equation describe in Π(R)?
(b) Homogenize the equation.
(c) What does the homogenized equation describe in RP
2
?
2. Consider the set of points in R
2
given by (t, t
2
/2).
(a) What does this set of points describe?
(b) Write an equation that gives a relation satisﬁed by the x and y
coordinates of the points on this curve.
(c) Homogenize the equation.
81
(d) Describe the set of points in Π(R) that satisfy the homogeneous
equation using parameters s and t. (Hint: Think of embedding
the points (t, t
2
/2) into R
3
by putting them in the plane z = 1.
Now homogenize the resulting parametric description.)
5.4 Conics
Recall that in R
2
, we can deﬁne a conic section to be the set of zeroes of any
degree two polynomial.
ax
2
+ by
2
+ cxy + dx + ey + f.
When the constant term, f, is not zero, and at least one of a, b, c is not zero,
the conic is nondegenerate, that is, it is a bona ﬁde parabola, hyperbola or
ellipse, as opposed to one or two lines or a single point. If we homogenize
our deﬁning polynomial, we get
ax
2
+ by
2
+ cxy + dxz + eyz + fz
2
It may be hard to see what this deﬁnes. Note that we can get rid of cross
terms by rotations, though. To help you believe that, we note that if we send
x → x + y and y → x − y then xy → x
2
− y
2
. Now look at the xyplane,
and compare the x axis to the line x + y, and the y axis to the line x − y:
we rotate through an angle of 45
◦
to eﬀect the transformation. This lets us
see that we lose no generality if we consider that the deﬁning polynomial for
a conic in RP
2
is given by ax
2
+by
2
+ fz
2
. Next note that scaling in the x,
y, z directions separately is deﬁned by a diagonal matrix which determines
a projective transformation. To understand how to think of a conic in RP
2
,
all we care about are the zeroes of the polynomial. It follows that the only
thing that matters here is the ratio of signs of the coeﬃcients of x
2
, y
2
, and
z
2
. This means that we can think of a conic section in the projective plane as
associated to the surface given by x
2
+y
2
−z
2
= 0, that is, a cone. Diﬀerent
conic sections then arise via a transformation of the cone that amounts to
a rotation about the origin in R
3
or scaling in various directions, neither
of which change the cone into a diﬀerent type of surface. This makes the
following theorem transparent.
Theorem 44. Under projective transformations, all (nondegenrate) conic
sections are the same. In other words, given any two conic sectionsin RP
2
,
there is a projective transformation that sends one to the other.
82
5.5 Degree
Our goal here is to show that the degree of a curve in
¯
Π(F) is actually a
geometric feature of the curve, not just an attribute of one polynomial that
may deﬁne the curve. We need a few results about homogeneous polynomials
to get there. This material draws from Walker’s classic book on algebraic
curves [10].
Theorem 45. The factors of a homogeneous polynomial are homogeneous.
Proof. Suppose F = fg, where F is homogeneous and f is not homogeneous.
Write f and g using homogeneous components as follows:
f = F
i
+ . . . + F
i+j
g = G
k
+ . . . + G
k+
where F
i
, F
i+j
are nonzero,homogeneous with degrees respectively i and i+j,
and G
k
, G
k+
are nonzero homogeneous of degrees k and k + . Attempting
to write fg using homogeneous components we get
fg = F
i
G
k
+ (F
i
G
k+1
+ F
i+1
G
k
) + . . . + F
i+j
G
k+
where F
i
G
k
and F
i+j
G
k+
are both nonzero with degF
i
G
k
= i+k < degF
i+j
G
k+
,
contradicting F homogeneous.
Let F
n
designate a homogeneous polynomial of degree n. F
n
can have
any number of variables. Say it has k variables, that is, F
n
= F
n
(x
1
, . . . , x
k
).
Let f be the nonhomogeneous polynomial in k − 1 variables that we get by
letting x
i
= 1, for some i ∈ ¦1, . . . , k¦. We say F
n
and f are associates.
The next three results, variations on a theme, follow immediately from
the theorem.
Corollary 5. The factors of F
n
are associates of the factors of f where F
n
and f are associates.
Corollary 6. F is irreducible if and only if its associates are irreducible.
A ﬁeld is algebraically closed provided all polynomials in a single variable
with positive degree factor into linear factors. Our favorite example of an
algebraically closed ﬁeld is C. Note that R is not algebraically closed as, for
example, x
2
+ 1 does not factor in linear factors if we restrict to coeﬃcients
in R.
83
Corollary 7. Let D be algebraically closed and F
n
homogeneous degree n in
D[x
0
, x
1
]. There are a ,= 0, a
i
, b
i
in D with F = aΠ
n
i=1
(a
i
x
1
−b
i
x
0
). In other
words, F factors into linear factors.
Now consider an algebraic curve in
¯
Π(F). This is the set of projective
points that satisfy F
n
(x, y, z) = 0, where F
n
(x, y, z) is homogeneous of degree
n. Let be any line in
¯
Π(F). Think of as determined by a two dimensional
subspace spanned by vectors u and v in F
3
. We can think of as the set of all
nonzero points su+tv, where s, t ∈ F. In other words, [s, t] are homogeneous
coordinates for a point on .
Consider what it means for a point to lie on both C and . Because
it lies on , it is su + tv, for some (s, t) ∈ F
2
¸ (0, 0). Because it lies on
C, F
n
(su + tv) = 0. A point of intersection of C and then corresponds
precisely to a solution [s, t] to F
n
(su+tv) = 0. We are now thinking of s and
t as variables so that F
n
(su +tv) is homogeneous degree n in two variables.
Our corollary above allows us to say that, counting multiplicity, there are n
solutions to F
n
(su + tv) = 0 in
¯
Π(F), as long as F is algebraically closed.
This proves the following theorem.
Theorem 46. Assume F is algebraically closed. The degree of a curve in
¯
Π(F) is the number of intersections it has with a line in
¯
Π(F)
The deﬁnition of degree as a geometric property of a curve is the ﬁrst sub
plot in the story of intersections and singular points of curves. An important
tool in the telling is resultants. We end these notes with a brief discussion of
this idea.
Deﬁnition 22. Let f(x) = a
0
+ a
1
x + a
2
x
2
+ . . . + a
n
x
n
, g(x) = b
0
+ b
1
x +
b
2
x
2
+ . . . + b
m
x
m
, a
i
, b
i
∈ F, a
n
, b
m
,= 0. The resultant, R(f, g) is the
determinant
a
0
a
1
. . . a
n
0 a
0
a
1
. . . a
n−1
a
n
.
.
.
b
0
b
1
. . . b
m−1
b
m
.
.
.
0 b
0
. . . b
m
where there are m rows of a
i
s and n rows of b
i
s.
84
Note that the resultant is the determinant of an m+ n m + n matrix.
Example 6. Let f(x) = x
2
−2x + 1 and let g(x) = x −1. Then
R(f, g) =
1 −2 1
−1 1 0
0 −1 1
= 0.
The resultant of a polynomial and its ﬁrst derivative is called the discrim
inant of the polynomial. We are used to seeing the discriminant deﬁned for
a quadratic polynomial f(x) = ax
2
+ bx + c as b
2
−4ac. If you think about
what the discriminant gives us, you start to get a sense of what resultants are
good for more generally. In particular, when the discriminant of a quadratic
polynomial is zero, the polynomial has a repeated root, that is, a root in
common with its derivative. A quick calculation reveals that the deﬁnition
we have here deﬁnes the discriminant of f(x) = ax
2
+bx+c as −a(b
2
−4ac).
Obviously, this is zero if and only if b
2
− 4ac = 0. As is typically the case
with determinants, we only care whether the resultant is zero or not zero so
for what we need, the two deﬁnitions are close enough.
Theorem 47. R(f, g) = 0 if and only if f and g have a common noncon
stant factor.
Proof. We sketch the proof.
Start by noting that f and g have a nonconstant common factor if and
only if there are ϕ(x) and ψ(x) with degree ϕ(x) < degree g(x), degree ψ(x) <
degree f(x) and fϕ = gψ. Now think of ϕ(x) = α
0
+ α
1
x + . . . + α
k
x
k
,
ψ(x) = β
0
+ β
1
x + . . . + β
x
, where α
i
, β
i
are unknowns. After doing the
multiplications to write fϕ = gψ, we see that we have equations giving us
a
0
α
0
= b
0
β
0
, a
0
α
1
+ a
1
α
0
= b
0
β
1
+ b
1
β
0
, etc. Essentially, the resultant is the
determinant of the (transpose of the) coeﬃcient matrix for this large homo
geneous system of equations. The system has a nontrivial solution— α
i
s, β
i
s
not all zero— precisely when the determinant of this matrix, the resultant,
is zero.
In the projective plane, an algebraic curve is deﬁned by a homogeneous
polynomial in three variables. Two such polynomials with a nonconstant
common factor deﬁne curves with a common component. How can we use
resultants to pick up on these common components?
85
Let f(x
1
, . . . , x
k
) be a polynomial with coeﬃcients in a ﬁeld F. We can
think of f as belonging to D[x
k
], where D = F[x
1
, . . . , x
k−1
]. D is an example
of an algebra, that is, a vector space that is closed under multiplication.
Example 7. Let f(x, y, z) = 2xy −x
2
+z
2
. We can think of f as belonging
to D[x], D = R[y, z]. Then we have f = z
2
− (2y)x −x
2
, so the “constant”
term is z
2
, the coeﬃcient of x is −2y and the coeﬃcient of x
2
is −1.
This idea allows us to deﬁne resultants for polynomials in several vari
ables. A resultant for f(x
1
, . . . , x
k
) and g(x
1
, . . . , x
k
) is then a polynomial
in k −1 variables.
Theorem 48. If F
n
, G
m
are homogeneous of degrees n, m in F[x
0
, x
1
], then
R(F
n
, G
m
) = 0 if and only if F
n
and G
m
have a common nonconstant factor.
Exercises
1. Find the intesection points of f(x, y, z) = x
2
+ y
2
− z
2
and the line
x + y + z = 0.
2. Find the resultant of the two polynomials from the last exercise. (Note
that there are three possibly diﬀerent resultants. Find all of them.)
3. Using the deﬁnition of discriminant given in the text, ﬁnd the discrim
inant of x
3
+ px + q.
4. Consider the determinant
a
0
a
1
a
2
. . . a
n−1
a
n
−λ 1 0 . . . 0 0
0 −λ 1 . . . 0 0
0 0 0 . . . −λ 1
.
Argue that this must be equal to a
0
+ a
1
λ + a
2
λ
2
+ . . . + a
n
λ
n
.
5. How can you tell, without doing the calculation, what the resultant of
f(x, y) = x
2
−2xy +y
2
and g(x, y) = x−y is? Now do the calculation.
86
References
[1] Coxeter, H.S.M, “Introduction to Geometry,” Second Edition, John Wi
ley and Sons, Inc., New York, 1989.
[2] Euclid, “The Thirteen Books of The Elements,” Vol. 1 (Books I and II),
Translated with introduction and commentary by Sir Thomas L. Heath,
Second Edition, Dover Publications, Inc., New York, 1956.
[3] Fishback, W. T., “Projective and Euclidean Geometry,” Second Edition,
John Wiley & Sons, Inc., New York, 1969.
[4] Hilbert, David, “Foundations of Geometry,” Second English Edition,
Translated by Leo Unger, Revised and Enlarged by Dr. Paul Bernays,
Open Court, LaSalle, Illinois, 1999.
[5] Loustau, John and Meighan Dillon, “Linear Geometry with Computer
Graphics,” Marcel Dekker, Inc., New York, 1991.
[6] “MacTutor History of Mathematics,” http://wwwhistory.mcs.st
andrews.ac.uk/history/index.html, University of St. Andrews, Scotland
[7] Malkevich, Joseph, “The Real Projective Plane,” web notes,
http://www.york.cuny.edu/ malk/mycourses/math244/realprojectiveplane.html
[8] Reid, Constance, “Hilbert,” SpringerVerlag New York, 1996.
[9] Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/
[10] Walker, Robert J., “Algebraic Curves,” Springer Verlag, New York,
1978.
87
today is projective geometry. Although this is not a course in projective geometry, we will study certain projective planes here. In that sense, this course forms a bridge from high school geometry to more advanced studies. The high school geometry course you probably had was primarily a course in synthetic Euclidean geometry. Synthetic geometry starts from ﬁrst principles: deﬁnitions, and axioms or postulates, which are truths held to be selfevident. The heart of the subject is logical synthesis applied to the axioms in a rigorous way to uncover the facts (theorems) which relate the objects of the geometry. This remains an attractive subject for students of any age with any background because all assumptions are, ideally, clearly stated at the beginning of the program. The objects are familiar and the relations among them rich. No previous knowledge is necessary. Planar Euclidean geometry is the geometry of straight edge and compass. The straight edge is unmarked, so cannot measure length, and the compass does not stay open once you lift it from the page. The objects and all relations among them must be produced using only a straight edge and compass applied to a ﬂat surface. One approach to studying Euclidean geometry is constructive, and usually a portion of the high school course deals with actual constructions. You can think of the synthetic approach as supplying detailed instructions for allowed constructions. The book that brought synthetic geometry to us in the West was The Elements of Euclid, a masterwork dating from around the year 300 B.C. Not only has The Elements been studied more or less continuously since it was written, it has remained the most important source for the subject since that time. High school students usually study textbooks based on The Elements, not The Elements itself. These textbooks are, more or less, the products of interpretations, corrections, and “improvements” to The Elements— many of questionable merit — that have been incorporated into the subject since the time of Euclid. One view of these sorts of textbooks is that they attempt to provide the elusive “royal road to geometry,”1 ﬁt for schoolchildren.[2], p. vi. Our course will start with The Elements itself, in a translation that dates to the beginning of the last century. Using this as a departure, we explore
There is a legend that Ptolemy, the king of Egypt, asked Euclid, his tutor, if there was not a shorter way to learn geometry. Must one go through The Elements? Euclid’s response was that there was “no royal road to geometry.” Similar stories are attributed to other mathematicians in response to complaints from other kings trying to learn mathematics. See [2], p. 1.
1
2
some developments in Euclidean geometry since the glory days of Greece. Since a great deal has happened over the last few centuries, what you see here represents a very small sampling of topics taken from what is available for study under the heading ‘geometry.’ Well before the twentieth century, mathematicians recognized that The Elements had gaps and inaccuracies. There were many signiﬁcant contributions to rectifying The Elements but David Hilbert’s Grundlagen der Geometrie (Foundations of Geometry), dating from the early 1900s, became the deﬁnitive correction and essentially closed the matter once and for all. Hilbert was able to control some of the diﬃculties with The Elements that had been recognized but had remained unresolved since antiquity. One of the more signiﬁcant contributions of Hilbert was his recognition of the necessity of leaving certain terms undeﬁned in an axiomatic development. Aristotle, who predated Euclid, thought a great deal about deﬁning basic geometric objects and noted that “ ‘the deﬁnitions’ of point, line, and surface ... all deﬁne the prior by means of the posterior” [2], p. 155. In other words, for predeccesors of Euclid through the dawn of the 20th century, it was common but uneasy practice to work with deﬁnitions where, for instance, you might use lines to deﬁne points and points to deﬁne lines. Moreover, it was wellknown that this was a logical subversion. Hilbert recognized that the solution to the problem was to jettison some deﬁnitions altogether. One must start with a minimal set of basic terms that remain undeﬁned: their properties are then detailed as postulates. For example, we do not deﬁne ‘point’, ‘line’, or ‘plane’ but we understand the nature of these objects through the postulates, for example, two distinct points determine a unique line, and a line together with a point not on that line determine a unique plane. All other deﬁnitions, for example ‘angle’, are then carefully crafted using this basic set of undeﬁned terms. Our foray into synthetic geometry includes contributions from the Renaissance that amounted to the initiation of the study of the projective plane through the addition of ideal points to the Euclidean plane. Our study of Euclid and Hilbert culminates with the nine point circle. If Euclidean planar geometry derives from straightedge and compass constructions, aﬃne geometry derives from straightedge constructions alone. This sounds poorer than Euclidean geometry but we come into a wealth of ideas when we use linear algebra to develop a model for aﬃne geometry. This is an extension of the familiar ideas of coordinate geometry and leads naturally to the three dimensional vector space model for a projective plane, an important tool in use today. 3
Projective geometry dates to the 16th century, thus predates many of the fundamental ideas that we think of as typical of the modern era in mathematics, such as sets and functions. Nonetheless, projective geometry has a strong ﬂavor of modernity as it provides a setting in which curves and surfaces have few or no exceptional conﬁgurations. This gives projective geometry tremendous power which we exhibit with a brief study of plane algebraic curves. Algebraic curves can be described parametrically with polynomials. Plane curves, that is, those that lie in R2 or C2 , necessitate the introduction of calculations and tools that are manageable but rich enough to give the reader an appreciation of some of the ideas handled in modern algebraic geometry. Our goal in this ﬁnal section of the course is Bezout’s Theorem, which is a description of the intersection of two curves in a plane.
2
2.1
Synthetic Geometry
Background on Euclid and The Elements
There are no contemporary accounts of Euclid’s life but there are some details about Euclid that scholars have been able to cobble together indirectly and about which there is little controversy. It is generally held that Euclid founded a school of mathematics in Alexandria, Egypt.2 In particular, Archimedes, 287212 B.C., sometimes described as the greatest mathematician who ever lived, studied in Alexandria at Euclid’s school and cited Euclid’s work in his own writing. It is clear then that Euclid predated Archimedes. On the other hand, The Elements has detailed references to the works of Eudoxus and Theaetetus. To have learned their work, Euclid would have to have gone to Plato’s Academy.[2], p. 2. The Academy was established outside Athens around 387 B.C.3 There is general agreement that Euclid was too young to have studied with Plato at the Academy, and too old to have taught Archimedes in Alexandria. This helps narrow down the dates for Euclid which are currently accepted as about 325265 B.C. [6] As a student at the Academy, Euclid would have been the product of
Alexandria is named for Alexander the Great, who, as a child, studied with Aristotle. This is the origin of the word academic. Academy is actually the name of the place where Plato set up his institution. The Academy remained in use until 526, another astonishingly longlived force in the intellectual world.
3 2
4
A plane surface is a surface which lies evenly with the straight lines on itself. There is room for doubt there. each of the equal angles is right. Aristotle. 2. as several books of The Elements have nothing whatever to do with the construction of the platonic solids. a student of Plato himself. When the lines containing the angle are straight. The Elements. even if we restrict attention to the West.a rich tradition of careful thought.[2]. though. 7. p. starts with a set of 23 deﬁnitions. 10. The extremities of a surface are lines. We quote from [2]. schooled in the writings and teachings of Plato. including geometry. 9. p. One commonly accepted view of The Elements is that it pulled together what was known in geometry at the time. 5 . and the straight line standing on the other is called a perpendicular to that on which is stands. The extremities of a line are points. that were accepted at that time. the angle is called rectilineal. and the other Greek geometers. A surface is that which has length and breadth only. suggested in his own writings that his students had sources which codiﬁed the principles of mathematics and science. In other words. A straight line is a line which lies evenly with the points on itself. with the goal of proving that the ﬁve platonic solids are the only solutions to the problem of constructing a regular solid. 3. Euclid’s was not the ﬁrst geometry text. Aristotle. A point is that which has no part. 2. A line is a breadthless length. These sources presumably would have been available to Euclid as well. which is in thirteen books. When a straight line set up on a straight line makes the adjacent angles equal to one another. 4. 8. 153154. This does not diminish its greatness but it is important to maintain perspective. 1. A plane angle is the inclination to one another of two lines in a plane which meet one another and do not lie in a straight line. 5. 6.
An acute angle is an angle less than a right angle. and such a straight line also bisects the circle. a rhombus that which is equilateral but not rightangled. 18. And the point is called the center of the circle. Of trilateral ﬁgures. A ﬁgure is that which is contained by any boundary or boundaries. 21. 14. 16. 17. an oblong that which is not equilateral but rightangled. A boundary is that which is an extremity of anything. that which has two sides equal. Rectilineal ﬁgures are those which are contained by straight lines. an isosceles triangle. 19. trilateral ﬁgures being those contained by three. 12. Further. quadrilateral those contained by four. And the center of the semicircle is that of the circle. 6 . and a rhomboid that which has its opposite sides and angles equal to one another but is neither equilateral nor rightangled. a rightangled triangle is that which has a right angle. that which has three angles acute. and a scalene triangle that which has its three sides unequal. A diameter of the circle is any straight line drawn through the center and terminated in both directions by the circumference of the circle. A circle is a plane ﬁgure contained by one line such that all the straight lines falling upon it from one point among those lying within the ﬁgure are equal to one another. 13. of trilateral ﬁgures.11. An obtuse angle is an angle greater than a right angle. A semicircle is the ﬁgure contained by the diameter and the circumference cut oﬀ by it. 22. and multilateral those contained by more than four straight lines. Of quadrilateral ﬁgures. an equilateral triangle is that which has its three sides equal. a square is that which is both equilateral and rightangled. and an acuteangled triangle. and obtuseangled triangle that which has an obtuse angle. 20. 15. All other quadrilaterals are called trapezia.
In standard usage today. being in the same plane and being produced indeﬁnitely in both directions..mcs. Legendre’s reworking of The Elements. make sure that you know what someone means when he says “circle” or “triangle. for instance. When we talk about the angle at which curves intersect. he directs his energies towards lines and circles. 4 7 . we usually call a curve. and plane.4 and dates to 1794. “The angle of intersection of curves at a point would be deﬁned as the angle of intersection of their tangents at that point. http://wwwhistory.” Since we are raised from an early age on calculus. and everything else. not necessarily straight lines. we may feel awkward because we know that curves may or may not have deﬁned tangents at a particular point.standrews. Euclid’s deﬁnition of triangle refers to the space enclosed by what we call a triangle. Another note is that Euclid’s deﬁnition of a “rectilineal ﬁgure” at ﬁrst seems to refer to what we would call a “polygon” but closer examination reveals a distinction. “El´ments de g´om´trie . Mathematicians and philosophers succeeding Euclid left a wealth of commentary on the deﬁnitions. do not meet one another in either direction. circles. we would call a disk. Euclid’s predecessors and contemporaries included. What Euclid calls a line. we may get a little nervous.ac. Euclid’s deﬁnitions thus contain a classiﬁcation of curves into three types: lines.” Context According to The MacTutor History of Mathematics Archive.. Euclid is not concerned with this problem as. and think something like. The object Euclid called a triangle is today called a 2simplex.uk/history/index. once again.23. is due to Legendre. The second and fourth deﬁnitions indicate a distinction between the notions of line and straight line. a triangle is three noncollinear points and the line segments they determine. That may not be so strange but it is worthwhile to note that more detailed classiﬁcations of curves were of great interest to geometers through the centuries.” in Euclid’s terminology. In spite of this nicely detailed dictionary. especially those of point. Parallel straight lines are straight lines which. which present no such ambiguities.html. line. We are used to thinking of an angle as a “rectilineal angle. was the leading elementary text on the topic for around 100 e e e years” and actually formed the basis for most subsequent texts on Euclidean geometry in Europe and the United States. What Euclid calls a circle. The deﬁnition of a line as the shortest distance between two points. The deﬁnition of angle refers to lines.
. viz. [2]. If the interior angles on the same side are less than two right angles. p. that is. p.often makes it clear but in any case. 220 for other formulations of the parallel postulate and the associated history. [2] p. The concern was whether it should really be a theorem. if produced indeﬁnitely. Exercise What is the usual deﬁnition of a circle? Show that Euclid’s deﬁnition is equivalent to the more familiar deﬁnition. 154155 1. Euclid reveals that he actually assumes that there is a unique line determined by each pair of points. it is not a dumb question. A ﬁnite straight line can be produced continuously in a straight line. (The Parallel Postulate) Consider a straight line falling on two straight lines. Euclid’s ﬁve postulates follow the deﬁnitions. p. All right angles are equal to one another. 4. There is another hidden assumption there. 2. that the line determined by a segment is unique. 202219. meet on that side on which the angles are less than the two right angles.) The ﬁfth postulate was controversial from Euclid’s time through succeeding centuries. (See [2]. The second postulate says that a line segment (ﬁnite straight line) can be extended into a line. This is an important part of the history of mathematics partly because all the fussing over the 8 . A straight line can be drawn from any point to any point. as well. 5. Much later in The Elements. One may describe a circle with any center and any radius. 3. A brief history of attempts to prove it is given in [2]. The Parallel Postulate is not what we are used to seeing. 195. whether the ﬁfth postulate actually followed from the other four. then the two straight lines.[2]. 196. The deﬁnition of a circle is not what we are used to. Exercise What is the usual formulation for the existence of parallel lines in a plane? Show that this follows from Euclid’s fourth and ﬁfth postulates. p.
i. 2. 4. for example.ncsa. 5.) The development of noneuclidean geometries by Lobachevsky in Russia and Bolyai in Hungary proved that the parallel postulate is independent of the other four. In other words. 2.edu/Cyberia/Cosmos/CosmosShape.. If equals be subtracted from equals. if a ∼ b and b ∼ c. ∼ is symmetric. A relation ∼ on S is called an equivalence relation provided 1. Euclid states ﬁve “common notions” in addition to the deﬁnitions and postulates.. the geometry of the surface of a sphere. Let S be a set. although it was many years before mathematicians accepted and understood the signiﬁcance of noneuclidean geometries.e. if you keep the other four postulates but excise or change the parallel postulate. 3. These are often referred to as “axioms” in the literature. a ∼ b implies b ∼ a.e. Things which are equal to the same thing are also equal to one another. 155: 1. Consider. i.uiuc. i. 3. a ∼ a for all a in S. Noneuclidean geometry may seem like an interesting idea that is of purely academic interest but nothing could be farther from the truth. p. an idea that is relatively modern. Things which coincide with one another are equal to one another. 9 . (See p. the wholes are equal.e. Other noneuclidean geometries are routinely used in the study of space— as in the cosmos. (See http://archive. you still get a viable axiomatic system. ∼ is reﬂexive.Parallel Postulate led to the development of noneuclidean geometries during the 19th century. the remainders are equal. we should pause to discuss equivalence relations. The common notions look a bit strange because of where we stand in the history of the subject. settling a controversy that persisted from Euclid’s time.html for example. The whole is greater than the part. To discuss them further. That was an enormous breakthrough. You realize how important this might be to humans when you think about traversing large distances across the surface of the earth.) We quote from [2].. If equals be added to equals. then a ∼ c. ∼ is transitive. 221 in [2].
in 1903 [4]. Diﬀerent angles that have the same measure are not equal. the name for a group of mathematicians (with changing membership) based in Paris from the 1930s through the present. This may seem like splitting hairs but in some sense it underlies aspects of The Elements that bothered Euclid himself. Hilbert’s work was certainly an important impetus in this direction. then for a jump in the evolution of mathematics. In modern English usage. if not for a genuine revolution. in 1891. The importance of axiomatization and abstraction in mathematics was explored more deeply than ever before. and independent basic notions. The late nineteenth and early twentieth centuries saw major developments in Euclidean geometry. who was the most important commentator of his time and possibly for the next hundred years. Homeomorphism is a topological equivalence relation.[9] Bourbaki. This movement was manifested most obviously in The Principia Mathematica of Russell and Whitehead. postulates. Regardless of how you respond to the role of abstraction in mathematics. most notably with the works of Pasch. Congruence is a geometric equivalence relation. The evaluation of Euclid’s postulate system was to whether the deﬁned terms. By the time later editions of Hilbert’s Grundlagen der Geometrie were published. the word equal is used to mean congruent. consistent. We are so used to thinking about congruence that it seems strange to have to say much about it but remember that equivalence relations per se were not familiar to Euclid and remember that he was working the subject from the most basic principles. Veronese.Equality is the archetype for equivalence relations. but congruent. Isomorphism is an algebraic equivalence relation. and common notions form a set of complete. The ﬁrst common notion says that equality is transitive. we say angles α and β are equal only if α and β are diﬀerent names for the same angle. Fermat’s Last Theorem) and in bringing certain subjects— 10 . initiated an ambitious project of putting all of mathematics on a ﬁrm axiomatic foundation. a level of ‘purity’ that is fundamentally misguided— it was and remains a profound inﬂuence. who sought to reduce all of mathematics to formal logic and set theory. and David Hilbert. Whether or not one prefers this approach to mathematics— some argue that examples rather than axiomatization should drive the subject and that a focus on abstraction leads one astray of applications. the stage was set. in 1882. Beware that in the Heath edition of Euclid. it has been hugely eﬀective both in dispatching certain problems that remained intractable through the ages (for example. We have mentioned Legendre.
(Being) what it was required to do. Again. p. BC is equal to BA. CB is equal to AB. Proposition 1. But CA was also proved equal to AB. its proof. Let A be the given point. again. and ﬁgure from [2]. since the point A is the center of the circle CDB. Therefore the triangle ABC is equilateral. BC are equal to one another. Early commentators noticed that there is an unarticulated assumption used in this proof. 244. (See [6] for more information about Bourbaki and their project. and it has been constructed on the given ﬁnite straight line AB. Thus it is required to place at the point A (as an extremity) a straight line equal to the given straight line BC. B let the straight lines CA. And things which are equal to the same thing are also equal to one another. 11 . Texts that resulted from their work are still very much in use. to the points A. therefore. and from the point C. p. algebraic geometry— to a better place for further development. AC is equal to AB. each of the straight lines CA. Extensive comments elaborating on this point are in [2]. with center B and distance BA let the circle ACE be described.) We consider some of the results (propositions) Euclid proves in The Elements and some possible problems with the proofs. Let AB be the given ﬁnite straight line. 241. Proposition 2. Thus it is required to construct an equilateral triangle on the straight line AB. We continue on to Proposition 2. CB be joined. in which the circles cut one another. we quote from [2]. Now. 2423. p. again quoting statement. namely that the circles must intersect. and the accompanying ﬁgure. To place at a given point (as an extremity) a straight line equal to a given straight line. Proposition 1. CA is also equal to CB. and BC the given straight line. since the point B is the center of the circle CAE. With center A and distance AB let the circle BCD be described. therefore. Therefore the three straight lines CA. On a given ﬁnite straight line to construct an equilateral triangle. proof. AB.notably. The correction involves an assumption about the continuity of a line or curve.
Then. The points E and F are respectively points on the lines determined respectively by the segments DA and DB.C D A B E Figure 1: Proposition 1 From the point A to the point B let the straight line AB be joined. And in these DA is equal to DB. we can construct a line segment AL which is congruent to BC. with center D and distance DG let the circle GKL be described. therefore the remainder AL is equal to the remainder BG. BF be produced in a straight line with DA. and on it let the equilateral triangle DAB be constructed. Again. given a point A and a line segment BC. But BC was also proved equal to BG. The statement of the proposition is that. since the point B is the center of the circle CGH. with center B and distance BC let the circle CGH be described. BC is equal to BG. DL is equal to DG. therefore AL is also equal to BC. therefore each of the straight lines AL. Therefore at the given point A the straight line AL is placed equal to the given straight line BC. Euclid is invoking 12 . Let the straight lines AE. And things which are equal to the same thing are also equal to one another. BC is equal to BG. Euclid’s ﬁrst step is to employ Proposition 1 to construct an equilateral triangle DAB. and again. DB. since the point D is the center of the circle GKL. (Being) what it was required to do.
By construction. using a straight edge and compass. but note that while H is arbitrary. as in the picture. if A is on the line determined by BC. G and H are points on this circle. radius BC. you would probably mark the endpoints of the given segment with the compass. to get AL ∼ BC. Euclid constructs two circles. we ought to be able to construct a segment extending oﬀ the given segment with the same length. The next step invokes Common Notion 3. do you think Euclid missed something or do you think he made a deliberate choice? If you think about constructing a segment of a given length. radius DG. move the compass to a new location. that a line segment is part of a line. and contains points K and where is the intersection point of the line AD and the circle. G must be the point where this circle intersects the line BD. the circle centered at D we now call GKL. Similarly. = Try to answer the following questions. essentially. Does Euclid’s proof apply in that case? If the proof needs modiﬁcation. does the proof still apply? Does it change depending on whether A is between B and C? What if A is an endpoint of the segment? In this case. which allows us to DA = subtract line segments. The second circle has center D. Next. = = ∼ DB. DL ∼ DG.K H D A C B G F L E Figure 2: Proposition 2 the second postulate here. As radii of the same circle. the ﬁrst with center B. which says. The circle centered at B we now call CGH. BC ∼ BG. mark 13 . in our language. Does a correct proof of Proposition 2 depend on the relative positions of the point A and the segment BC? For instance.
Thus it is required to cut oﬀ from AB the greater a straight line equal to C the less. Now. 246. But C is also equal to AD. We quote from [2]. since the point A is the center of the circle DEF . from AB the greater AE has been cut oﬀ equal to C the less. This violates the stricture that the compass used for Euclidean geometry does not stay open when we lift it oﬀ the page. We quote 14 . Therefore. Again. to cut oﬀ from the greater a straight line equal to the smaller. At the point A let AD be placed equal to the straight line C. Proposition 3. so that AE is also equal to C. look for implicit assumptions about the relative positions of objects and ask if they make a material diﬀerence to the proof. C be the two given unequal straight lines and let AB be the greater of them. The next two propositions are the ﬁrst substantial theorems about triangles. p. C is equal to AD. you can transfer information about length from one location to another.C D E A B F Figure 3: Proposition 3 the endpoints of the new segment. Let AB. Proposition 3 is a corollary that says we can cut a given line segment to the match the length of a smaller segment. Proposition 4 is the familiar sideangleside criterion for triangle congruence. Given two unequal straight lines. Proposition 2 proves that even with a ﬂoppy compass. given the two striaght lines AB. and with center A and distance AD the circle DEF be described. Each has spawned much comment through the centuries. and use the straight edge to connect the dots. C. Therefore each of the straight lines AE. AE is equal to AD. (Being) what it was required to do.
Therefore the base BC will coincide with EF ] and will be equal to it. p. if the triangle ABC be applied to the triangle DEF . If two triangles have the two sides equal to two sides respectively. Again. In fact. the angle ABC to the angle DEF . then the point B will also coincide with E. 225.from [2]. hence the point C will also coincide with the point F . p. that is. This time. hence the base BC will coincide with the base EF . AC equal to the two sides DE. And the remaining angles will also coincide with the remaining angles and will be equal to them. aver that Euclid himself was uncomfortable about employing this argument [2]. I say that the base BC is also equal to the base EF . p. (Being) what it was required to prove. DEF be two triangles having the two sides AB. You should supply one as an aid to follow the argument. [For if. and if the point A be placed on the point D and the straight line AB on DE. namely those which the equal sides subtend. the base BC does not coincide with the base EF . two straight lines will enclose a space: which is impossible. and the angle BAC equal to the angle EDF . the business of picking up one triangle and dropping it down onto another for comparison. because AC is again equal to DF . because AB is equal to DE. and the angle ACB to the angle DF E. al. namely AB to DE and AC to DF . 249. Heath argues that Common Notion 4 and the bracketed material in the proof were added by later commentators to justify this step in Euclid’s proof. and have the angles contained by the equal straight lines equal. [2]. the angle ABC to the angle DEF . we omit Euclid’s ﬁgure. DF respectively. Let ABC. Proposition 4. But B also coincided with E. the triangle will be equal to the triangle. and the remaining angles will be equal to the remaining angles respectively. they will also have the base equal to the base.” that is. 2478. because the angle BAC is equal to the angle EDF . For. when B coincides with E and C with F . 15 . and remaining angles will be equal to the remaining angles respectively. The controversy about the proof is to do with the socalled “method of superposition. Therefore etc. namely those which the equal sides subtend. AB coinciding with DE. Heath et. the triangle ABC will be equal to the triangle DEF . and the angle ACB to the angle DF E. the straight line AC will also coincied with DF .
If superposition were allowed. But F Cwas also proved equal to GB. Then. and. while the base BC is common to them. which we take up in the next section. and the remaining angles will be equal to the remaining angles respectively. and the angle CBD to the angle BCE. 6. that is. And. 251. I say that the angle ABC is equal to the angle ACB. 249250 for more details regarding this and other diﬃculties with the proof of Proposition 4. pp.Why is superposition problematic? The argument. from AE the greater let AG be cut oﬀ equal to AF the less. but people still enjoy puzzling over it. we could use it there and dispatch the proof in one line. the asses’ bridge. F C are equal to the two sides CG. known famously as pons asinorum. CE be produced further in a straight line with AB. and the triangle AF C is equal to the triangle AGB. the angle ACF to the angle ABG. GB be joined. Using it in this proof and not the others belies a conﬂict in the underlying ideas. the angle F AG. namely those which the equal sides subtend. AB respectively. Let ABC be an isosceles triangle having the side AB equal to the side AC.) We will grant Proposition 4 and proceed to consider Proposition 5. In isosceles triangles the angles at the base are equal to one another. p. therefore the triangle BF C is also 16 . or (2) that it refers to the picture Euclid supplied for his proof. 415 We quote the statement. Hilbert resolves the conﬂict with a dramatic ﬂourish. Consider the pains Euclid took to prove Proposition 2. AC are equal to the two sides GA. The actual origin and meaning of the name is lost.[1]. AC. (See [2]. the two sides F A. and the angle BF C is equal to the angle CGB. and in these AB is equal to AC. one could use it to prove many of the propositions in The Elements with little trouble. if the equal straight lines be produced further. and let the straight lines BD. proof. Proposition 5. the remainder BF is equal to the remainder CG. p. the angles under the base will be equal to one another. Let a point F be taken at random on BD. which goes back to the 16th century. since the whole AF is equal to the whole AG. and the angle AF C to the angle AGB. [2]. since AF is equal to AG and AB to AC. p. GB respectively. therefore the two sides BF . and they contain a common angle. and let the straight lines F C. Therefore the base F C is equal to the base GB. The most common stories behind the name are (1) that it refers to those who cannot follow the proof . and ﬁgure from [2]. is that if superposition is indeed a legitimate method of proof.
it can be argued that this is the board from which a great many of the fundamental theorems of Euclidean geometry spring. quite a diﬀerent approach to the subject altogether. This is. pulling together known results but from early on in the program.3.D. and they are at the base of the triangle ABC. First. Finally. Section 1. therefore the angle F BC is equal to the angle GCB. references in Aristotle indicate that geometers who preceded Euclid knew pons asinorum but had a diﬀerent proof. and the angle BCF to the angle CBG. This helps us understand the role of The Elements in history. in fact. 17 . and indeed. Coxeter’s approach to teaching geometry in [1] (cf. and in these the angle CBG is equal to the angle BCF . since the whole angle ABG was proved equal to the angle ACF . pp. the remaining angle ABC is equal to the remaining angle ACB. and the remaining angles will be equal to the remaining angles respectively. 252254. namely those which the equal sides subtend. Accordingly.A B C F D G E Figure 4: Pons Asinorum equal to the triangle CGB. to a large extent.E.) Second. Q. he forges his own way through the thicket. Pons asinorum is signiﬁcant for several reasons. See [2]. Therefore etc. Euclid was. But the angle F BC was also proved equal to the angle GCB. and they are under the base.
Our goal for this part of the course is the nine point circle but for now we turn away from Euclid to 18 . Let ABC be an isosceles triangle. And the angle BAC is equal to the angle CAB for it is the same. namely BC to BC. p. Book II is concerned with parallelograms and Book III takes up circles. Let us conceive this one triangle as two triangles. and then laying it down again upon itself. for these are the angles subtended by the equal sides AB. (Compare also [1]. the two sides AB. “Euclid: Surely that has too much of the Irish Bull about it.. Therefore all the corresponding parts (in the triangles) are equal. and AC to AB. Pappus’s proof is often presented without attribution and with modiﬁcation.3. a prominent commentator who came around 700 years after Euclid. 254.” while faintly suspect. we invoke a rebuke to commentators who proposed to improve Euclid’s proof by means such as these.Proclus. that the triangle is picked up from the page and reﬂected across a median or angle or base bisector (none of which have been deﬁned) and dropped down again upon itself. Before leaving this topic.) Here is Pappus’s proof. as quoted from [2].. and the angle ACB to the angle ABC. which is quite possibly what you saw in high school..e. L. Section 1. namely. Dodgson (a. relates the proof that Pappus described. AC are equal to the two side AC. this approach is at best unsettling and Pappus’s original approach.a.or upon a trace it has left on the page after it has been picked up. AB. Note that a square here is a geometric ﬁgure. as cited in [1]. then it is a right triangle. 6: “Minos: It is proposed to prove [pons asinorum] by taking up the isosceles Triangle. to “conceive this one triangle as two triangles. Since AB is equal to AC.. and let us argue in this way.k. the triangle ABC to the triangle ABC (i. Note that this does not cover the part of the theorem describing the angles under the base of the triangle. Therefore in isosceles triangles the angles at the base are equal. AC. Lewis Carroll) from Euclid and His Modern Rivals. In these books of The Elements. seems elegant in comparison. Euclid does not work with numbers. turning it over. Proposition 47 is the Pythagorean Theorem. the angle ABC to the angle ACB. to deserve a place in a stricly philosophical treatise?” Book I of Euclid’s Elements ends with Proposition 48. and reminds one a little too vividly of the man who walked down his own throat. ACB). and AB equal to AC. p.. This is a quote of C. In light of previous remarks. the theorem that says that if the squares of two sides of a triangle add up to the square of the third side.
His goal was a description of the foundation of Euclidean geometry that would yield all the familiar theorems while meeting modern standards of rigor. b. 2. the assumptions about them. B. c. Hilbert. Hilbert was the one responsible for acknowledging the necessity of starting with terms that were and would remain. the notion to be deﬁned.his most successful modern rival. [4]. denoted A. etc. II. By Hilbert’s time. Hilbert singled out the following undeﬁned terms referring to objects: points. This may seem like a triviality but it was an important idea in the early development of logic and set theory and. All human knowledge thus begins with intuitions. The undeﬁned terms referring to relations among points. 2. by implication.” Second Part. 168 (cf.. p. undeﬁned. whom Hilbert quotes before the introduction of Foundations of Geometry. “Elements of Transcendentalism. it had long been recognized that is was impossible to explain something like a straight line “by any regular deﬁnition which does not introduce words already containing in themselves.. planes. etc. and spaces. C. Points and lines are the elements of plane geometry.) Predecessors of Hilbert started with material approximations to terms like point and line. denoted α. p.2 Hilbert’s Axioms Hilbert’s goal in studying The Elements was to sort out the objects Euclid studied. The points are called the elements of line geometry. denoted a. Hilbert’s convention. and congruent. Euclid likely set the standard for rigor but standards change and Hilbert wanted to bring Euclid’s work up to a modern level of acceptability. lines. lines. which we follow.” [2]. is that when we refer to points and lines with language such as. in some sense. β. Critique of Pure Reason. and the relationships among them. γ. Aristotle’s remark about deﬁning the prior with the posterior. between. Points. frankly. 19 . planes. and space. In his time. etc. goes back to Kant. Several of Euclid’s deﬁnitions had to be jettisoned. lines and planes are the elements of space geometry. “A and B are points” it is to be understood that A and B are distinct. Kant. are lies on. proceeds thence to concepts and ends with ideas. so that we can proceed to our goal on a more secure foundation.
Hilbert knew set theory but did not use its language and conventions in the Grundlagen. 7. 4. I. lines. Every plane contains at least one point. 8. Every set of three noncollinear points is contained in a plane. There exist at least four points which do not lie in a plane. If two planes have a point in common. Hilbert did not use these words in his axioms but did later on in the text. If two points of a line a lie in a plane α. Points that do not lie on a single line are noncollinear. Axiom of Parallels. that points. Points that lie in a single plane are coplanar. 5. and Axioms of Continuity. 2. Axioms of Congruence. The 20 . We denote by AB the unique line determined by points A and B. and planes are all diﬀerent sorts of objects. These are not verbatim but are only slightly modiﬁed from the translation of Grundlagen [4] that we are using. then they have at least one more point in common. The plane determined by the three noncollinear points is unique. in other words. Points that lie on a single line are collinear. then all of the points of a lie in α. This axiom is a guarantee against triviality.Hilbert’s axioms are organized into ﬁve subsets: Axioms of Incidence. we are so used to its basic ideas and notation that it seems natural to use them here. Modern set theory dates from the late 19th century with the work of Georg Cantor. Points that are not in a single plane are noncoplanar. 3. The line determined by two points is unique. 6. Axioms of Order. There exist three points that do not lie on a single line. Axioms of Incidence 1. Two points A and B determine a line a that contains them. Every line contains at least two points. Although we will not deal with it in a deep way. The underlying idea is that lines and planes are sets.
[4] When the point B lies between points A and C. There is exactly one plane containing both a and b. The intersection of a and α is empty or a single point. or A ∈ α ∩ β. we write A ∗ B ∗ C. The intersection of two lines in a plane is a single point or empty. If the line a lies in the plane α. C are collinear and C ∗ B ∗ A. If A and C are two points there exists a point B ∈ AC with A ∗ C ∗ B. Hilbert attributes this set of axioms to Pasch. The intersection of two planes is empty or exactly one line and no other points. Let a be a line and A a point not on a. we write A ∈ a or A ∈ α. Let a and b be lines with a single point of intersection. Designate the line determined by points A and B by AB. we think of a as a subset of α and write a ⊂ α. prove the following theorems. A and B are the endpoints of AB. 3.elements of these sets are points. or the planes α and β. If A ∗ B ∗ C then A. There is a exactly one plane containing a and A. C is inside AB otherwise C is outside AB. A ∈ a ∩ α. Note that Hilbert does not use the notion of sensed or directed line segments. If the point A belongs to the lines a and b. in particular. segments and angles. write A ∈ a ∩ b. 21 . the line a and plane α. Axioms of Order Line segments and angles are not undeﬁned terms but the notion of betweenness is and it underlies a great many of the familiar ideas from Euclidean geometry. 5. B. 4. 2. 2. Let α be a plane and a a line that does not lie in α. a German mathematician of the late 19th centery. Hilbert deﬁnes the line segment AB or BA to be the point A and B together with the points between A and B. If A is a point lying on the line a or the plane α. II. 1. When A ∗ C ∗ B. 1. Exercises Using only Hilbert’s axioms as stated above.
Of any three points on a line. it is possible to order them A. lines. K so that A ∗ B ∗ C.3. 4. and bottles of beer. designated as sides of the line. . B. C be noncollinear points and let a be a line in the plane ABC that does not contain any of A. . or C but that passes through AB. The axioms to this point allow us to prove several things that seem intuitively obvious when we talk about space. . Exercise Use Axioms of Incidence and Betweenness to prove that for any points A and B there is a point C on AB with A ∗ C ∗ B. tables. . AC intersects a. Given any ﬁnite number of points on a line. The argument makes use of Axiom II. Theorem 3. B ∗ C ∗ D. and planes. Between any two points on a line there exist an inﬁnite number of points. 22 . Let A.4 and is not as easy as one might expect. For us. Hilbert in fact remarked famously that his axiom system should be as applicable to chairs. . Then a must also pass through AC or BC. As a line divides a plane into two sets. C. . The Axioms of Order are what allow us to view polygons as having insides and outsides. Theorem 1. They also give us the notion of a ray. we state some of the consequences of the Axioms of Order. I ∗ J ∗ K. B. [8] Omitting proofs. as to points. Theorem 2. and a plane divides 3space into two sets. it must exit through another. . For points A and B on the same side of a. but remember that neither Euclid nor Hilbert was discussing Rn . a point divides a line into two sets. . B. for example. there is no more than one that lies between the other two. the segment AB does not intersect a. The points of α that do not belong to a fall into two collections. For points A and C on opposite sides of the line. Let a be a line and α a plane containing a. the sides of a triangle are like doors to its interior and if a line enters through one door. space typically means Rn . The last axiom says that in a plane.
If we are to have triangles. Thus every point of a line partitions it into two rays. = = = This axiom allows the addition of segments. Recall that this was the second proposition in Book I of The Elements.. O. we can ﬁnd a point B on either side of A so that AB ∼ A B . k) or ∠(k. as did Euclid’s Common Notion 3. that segment congruence is reﬂexive. Let A. = The second axiom is a form of the ﬁrst of Euclid’s Common Notions. In that case. On the same or another line a let A B and B C be two segments which except for B also have no point in common. Let α be a plane and h. A are then said to lie on the line a on one and the same side of O and the points A. The ﬁrst axiom of congruence allows the possibility of constructing congruent segments. if AB ∼ A B and BC ∼ B C then AC ∼ A C . On the line a let AB and BC be segments which. The totality of points of a that lie on one and same side of O is called a ray emanating from O. Deﬁnition 2. Given a segment AB on a line a and a line a with a point A ∈ a . = = = Exercise Prove that every segment is congruent to itself.e. The pair of rays h. except for B. i. 3. Prove that segment congruence is also symmetric and transitive. k any two distinct rays emanating from O in α and lying on distinct lines. 23 . III. 1.Deﬁnition 1. h). we need angles and some axioms guiding us in how to think about the notion of congruence. Axioms of Congruence When discussing segments and angles we use the words congruent and equal interchangeably. B be four points of the line a such that O lies between A and B but not between A and A . The points A. k is called an angle and is denoted ∠(h. If AB ∼ CD and EF ∼ CD then AB ∼ EF . 2. B are said to lie on the line a on diﬀerent sides of the point O. A . have no point in common.
Notice that Axiom III. we use a to distinguish a line a from a ray a contained in a. it says that “obtuse angles” are excluded by this deﬁnition. An angle partions the points of a plane into the interior of the angle and the exterior of the angle. 4. 4. Show that the interior of an angle is convex. 5 24 . Show that HK lies in the interior of ∠(h. Let H be a point of ray h and K be a point of the ray k where h and k form an angle ∠(h. As with segments. We continue with axioms of congruence.Straight angles and angles that exceed straight angles are excluded by this deﬁnition.5 also gives us ∠BCA ∼ B C A . then the segment AB is entirely contained in R. every angle is congruent to itself. 5. AC ∼ A C . k) and ∠(k. Exercises 1. This may be an error in the translation. Is the exterior convex as well? 3. Thus ∠(h. If ΔABC and ΔA B C are triangles with AB ∼ A B . Moreover. A ray that lies in the interior intersects HK. 2. then the congruence ∠ABC ∼ ∠A B C is also and∠BAC = = satisﬁed. h) denote the same angle. k). we are not treating angles as having sense or direction. Every angle in a given plane can be constructed on a given side of a given ray in a uniquely determined way. Give a precise deﬁnition of the interior and the exterior of an angle.5 When necessary. If R is any region of a plane with the property that whenever points A and B are in R. k). then R is convex. = In [4]. = = ∼ ∠B A C . Show that neither set is empty. A ray emanating from O lies either entirely inside or entirely outside the angle.
there is a proposition that says a right angle is congruent to its supplement.4). = = = 25 . ∠ABC ∼ ∠A B C .5 implies ∠BAC ∼ B A C and ∠ACB ∼ A B C . ∠A ∼ ∠A . Axiom III. = = = Theorem 5 (SAS or ﬁrst congruence theorem for triangles). using uniqueness of angle construction and Axiom III. Two angles with a common vertex and a common side are supplementary provided their other sides form a line. = Choose C on A C so that either A ∗C ∗C or A ∗C ∗C and A C ∼ AC. Given isosceles ΔABC with AB ∼ BC.4. = We have shown that C is uniquely determined. Axiom III. we can write AB ∼ BC. It = = remains to show AC ∼ A C . Proof.Exercise Prove the uniqueness of segment construction. then ΔABC ∼ ΔA B C . Theorem 4. Axiom III. unless C = C . we revisit pons asinorum. If ∠A ∼ = ∠A . Deﬁnition 3.5 does not quite give us the sideangleside (SAS) criterion for triangle congruence but it is very close. If ΔABC and ΔA B C satisfy AB ∼ A B . This violates uniqueness of angle = construction as given in Axiom III. By III. Before settling the congruence theorems for triangles. BC ∼ AB. The base angles of an isosceles triangle are congruent. Note then that we have SAS for triangles ΔABC and ΔA B C since AC ∼ A C . We also have ∠ABC ∼ ∠A B C”. we see Hilbert taking one of Euclid’s propositions as an axiom. Start with the assumption that a segment congruent to AB is constructed two ways on a ray emanating from A to B and to B . = Proof. AB ∼ A B and ∠B ∼ ∠B .5. Recall that Euclid deﬁned a right angle in terms of lines impinging on one another. A bit later in Book I. = = ∠B ∼ ∠B (III. then. then = = = ΔABC ∼ ΔA B C .5. Once again. Theorem 6 (ASA or second congruence theorem for triangles). A right angle is an angle congruent to one of its supplements. ∠A ∼ ∠C. and AB ∼ A B . BC ∼ B C . The next theorem is commonly called anglesideangle (ASA).5 implies that ∠ABC ∼ = = = = A B C .
Each angle marked with double arcs is supplementary to the one marked with a single arc. we can as= sume AB ∼ A B and BC ∼ B C . Now notice that we can treat any pair of vertical angles as supplements to a given angle which is congruent to itself. as desired. A ∗ B ∗ D . Proof. ∠ABC ∼ ∠A B C so that. thus to prove the sidesideside (SSS) criterion for triangle congruence. By SAS. = Again by SAS. ΔABC ∼ A B C . This gives us ∠ADC ∼ ∠A D C . ∠A ∼ ∠A and AB ∼ A B . BD ∼ B D . Vertical angles are congruent. Choose C exactly as in the last proof and note that this gives us SAS for ΔABC and ΔA B C as AC ∼ A C . Proof.C C C A B A B Figure 5: SAS for triangles Proof. = We conclude that ∠DBC ∼ ∠D B C . in = = turn implying that CD ∼ C D . thus that = ΔABC ∼ ΔA B C by SAS. D with A ∗ B ∗ D. the supplements are thus congruent. We conclude that AC ∼ A C . The picture above applies here as well. Consider the angles in Figure 7. The next lemma allows us to add and subtract congruent angles. thus SAS for ΔCBD and C B D . = = = Now choose points D. ΔADC ∼ ΔA D C . again. The last theorem gives us = = = ∼ ΔA B C which in turn implies that ∠ABC ∼ ∠A B C . Without loss of generality. = Corollary 1.4. By the previous theorem. unless C = C . Supplements of congruent angles are congruent. By ΔABC = = assumption. = Theorem 7. uniqueness = of angle construction is violated. The latter is congruent to itself by Axiom III. Suppose ∠ABC ∼ ∠A B C . The reader 26 .
27 .C C A B D A B D Figure 6: Supplements of congruent angles are congruent Figure 7: Vertical angles are congruent.
L ∈ and choose K ∈ k . = = Following through. Choose B on k so that OB ∼ OB . Construct an angle congruent to ∠(h. k ) are supplements and congruent so already are right angles. ).e. diﬀerent sides) of and that h . k . OL ∼ O L . k. KH ∼ K H and ∠OKH ∼ Using supplements of ∠OHL = = = ∼ ΔO K H . KL ∩ h = H. a unique point. Choose H ∈ h to satisfy OH ∼ O H . Fix a point O and a ray h emanating from O. ) = Proof. Choose a point B oﬀ the line h and let k be the ray emanating from O passing through B. Theorem 8. We complete the proof for the case where h. k). by taking a ray k emanating from O on the side of h that does not contain B. If ∠(h. k are on the same side of and h . H ∈ K L as claimed. ∠O L K ∼ ∠O L H implying that L K = L H . By uniqueness of congruent angles on a given side of a ray. Consider two cases. implying that ∠(h. also by = = = 28 . Suppose that h. = = Since h is interior to ∠(k. ). be rays emanating from a point O. we can suppose h is interior to ∠(k. k ). k) and ∠(h. We now have a procedure that allows us to construct a right angle on a given line. we get ASA for ΔOKH = = ∠(h . We claim that H ∈ K L . We can construct a right angle. Lemma 1. k) ∼ ∠(h . Recall that we deﬁned a right angle as one that was congruent to its supplement. Let = A be the point where h intersects BB . ) ∼ ∠(k . L ∈ so that OK ∼ O K . k are on the same side (respectively. Case 1: A = O. Changing labels if necessary. Proof. it follows that KH ∼ K H . and OB ∼ OB . k ) and ∠(k.should supply a picture to help understand the hypotheses of the result as well as the proof. diﬀerent sides) of . Using congruence of supplements of congruent angles. = = ∼ ∠(h . = Since KL ∼ K L . HL ∼ H L . k are on the same side (respectively. = Note that we have SAS for ΔOKL ∼ ΔO K L and for ΔOHL ∼ ΔO H L . Let h. Let h . be rays emanating from a point O . then ∠(h.. i. k . ). we get ∠OLK ∼ ∠O L K and ∠OLH = ∠OLK ∼ = = ∠O L H . k) ∼ ∠O K H . Case 2: If A = O then we have SAS for ΔAOB and ΔAOB since AO ∼ AO. Here ∠(h. the reader can supply details of the other case. = = = ∼ ∠O H L . ). the same side of . ∠AOB ∼ ∠AOB by construction. Pick K ∈ k.
The implication is that ∠OAB ∼ ∠OAB . We must work around the fact that we do not have transitivity of angle congruence in the next proof. we get congruent angles as marked. As corresponding angles. The next result brings us closer to SSS as well as transitivity for angle congruence. ∠ABC1 ∼ Now SAS applies to ΔBC1 A = = ∠ABC2 . Using pons asinorum. Let C1 and C2 be on opposite sides of a and suppose AC1 ∼ AC2 and BC1 ∼ BC2 . The reader should complete the proof by addressing cases in which C1 C2 intersects AB. Let a be a line through points A and B.B k h A O k Figure 8: Construct a right angle construction. thus must be right angles. Then = = ∠ABC1 ∼ ∠ABC2 . We invoke Lemma 1 to subtract congruent angles so that ∠BC1 A ∼ ∠BC2 A. Notice that these = are supplements. Lemma 2. = ∼ ΔBC2 A. I am indebted to Wilson Stothers for pointing out errors in a previous version of this proof. at an endpoint or otherwise. 6 29 . = Proof. 6 Figure 9 shows one arrangement of points that satisﬁes the hypotheses of the lemma.
corresponding sides are congruent. This gives us ΔABC ∼ ΔA B1 C = ΔA B C . k ) ∼ ∠(h . = Take B2 on the other ray so that AB ∼ A B2 . If ∠(h. k) ∼ ∠(h . we can also say ΔA B C ∼ = ΔABC. It follows from there that = = ∼ C B1 and CB ∼ C B2 implying C B1 ∼ C B2 . By Lemma 1. k ) and ∠(h . as desired. take the point B1 so that AB ∼ A B1 . Lemma 1 also applies to A B C and A B2 C = so that ∠C A B ∼ ∠C A B2 . We have CA ∼ C A by = = assumption. we CB = = = have ∠C A B1 ∼ ∠C A B2 . The ﬁrst result in this direction corresponds to Axiom III. then corresponding angles are congruent as well so that the triangles themselves are congruent. B is on same side of A C as one of the rays making these angles. = = ∼ A B1 ∼ A B2 . ∠BAC ∼ ∠B1 A C and ∠BAC ∼ ∠B2 A C by construction. k ). k ) then ∠(h.C1 A B C2 Figure 9: This is one conﬁguration possible given the hypotheses of Lemma 2. If in triangles ΔABC. Theorem 10. Proof. also by construction. This gives us SAS for and AB = = ΔABC ∼ ΔA B1 C and ΔABC ∼ ΔA B2 C . The SSS criterion for triangle congruence implies that angle congruence is an equivalence relation. On that ray. Construct one angle congruent to ∠BAC at A on each side of the ray emanating from A through C . Theorem 9 (SSS or the third congruence theorem for triangles).2. Uniqueness of angle construction implies that = B = B1 . k) ∼ = = = ∠(h . = Note that since segment congruence is symmetric. 30 . ΔA B C .
Lemma 3.B B1 B A C A C B2 Figure 10: SSS for triangle congruence Proof. Let H ∈ h . To show is interior to ∠(h . k) emanating from vertex = O. k respectively so that OB ∼ O B ∼ O B . k). O . k) ∼ ∠(h . k ) implies by the theorem = = that ∠(h . ∠(h. k ) ∼ ∠(h. = = ∼ ∠(h . This gives us SAS for ΔOHK ∼ = = = 31 . k) ∼ ∠(h . Angle congruence is an equivalence relation. k) ∼ ∠( . choose a point A. with ∠(h. k ). = Our next goal is to establish means for comparing angles. Let emanate from O interior to ∠(h. ∠(h k ) emanating from O . k ) ∼ ∠(h . thus SSS. B on k . Proof. Choose a point B on k and B . ) = = Proof. k ). we need prove that if = H = O is a point on h and K = O is a point on k . ) ∼ ∠(h . SSS allows us to say ΔAOB ∼ ΔA O B . Since ∠(h . k ) is interior to ∠(h. The unique ray emanating from O on the k side of h that satisﬁes ∼ ∠(h . k). We need two lemmas. k ) as desired. Choose A and A on h . k ). ∠(h. k). Label the vertices of the angles O. This gives us SAS for ΔAOB ∼ ΔA O B and = = = ΔA O B ∼ ΔA O B . On h. implying ∠(h. Transitivity is now immediate by the theorem. Let be the unique ray emanating from O on the k side of h with ∠(h. O respectively. k) = Corollary 2. ) and ∠( . ). then H K intersects . Suppose ∠(h. K ∈ k be diﬀerent from O and choose H ∈ h and K ∈ k so that OH ∼ O H and OK ∼ O K . h respectively so that OA ∼ O A ∼ = = O A .
) respectively. Suppose is the ray emanating from O on the k side of h such that ∠(h. ). k ) and exterior to its supplement ∠(h . k ) ∼ ∠(h. k) < ∠(h . This gets us SAS for ΔOHL ∼ ΔO H L . ∠(h. ) be as in the theorem. It is clear then that for any two angles α and β. Let ∠(h. k ) < ∠(h . Suppose everything is as hypothesized. k). = Theorem 12. ) = = as well as ∠(h. Let k emanate from O on the k side of h so that ∠(h. Proof. there is a unique ray k emanating from O. Since is interior to ∠(h. Let ∠(h. We conclude that L ∈ H K . 32 . k ) emanating from O . that violates = uniqueness of angle construction. ) ∼ ∠(h . ) emanating from O . ) with ∠(h. and ∠(h . Suppose k emanates from O on = the side of h such that ∠(h. ) and ∠(h . k ) ∼ ∠(h . we can let L = HK ∩ . ). k) emanating from O. ) = which is a contradiction. ) provided the construction of an angle congruent to ∠(h. ) ∼ ∠(h . one and only one of the following statements is true: α < β. k) ∼ ∠(h . k). k) < ∠(h . By the previous lemma. emanating from O. interior to ∠(h. k ). be right angles with supplements ∠(h. Let ∠(h. ∠O H K = = ∠O H L . k) using h and a ray emanating from O on the side of h yields a ray interior to ∠(h . k ). with interior to ∠(h. By uniqueness. Proof. ) be given. ). k) ∼ = ∠(h . k). Changing labels. k ). Suppose ∠(h. Note that k is interior to ∠(h . ∠(h . Then is interior to ∠(h. α ∼ β. All right angles are congruent.ΔO H K . Pick L ∈ so that OL ∼ O L . We write ∠(h. Note now = = that ∠OHL = ∠OHK ∼ ∠O H K . k ). k) and ∠(h . by the ﬁrst triangle congruence and = ∠OHK = ∠OHL ∼ ∠O H L by the second. which proves the result. k) = if and only if k is exterior to ∠(h . we get the result in the other direction as well. k) but say that k is interior to ∠(h . ). k) and ∠(h . k ) ∼ ∠(h . We conclude that interior implies k exterior. Theorem 11. This gives us ∠(h. k ) > ∠(h . ) ∼ ∠(h . Since = ∠(h . α > β. and both k and k are on the side of h. ). A right angle is deﬁned as one congruent to its supplement.
The interior angles of the triangle are ∠ABC. it follows that D = B which is absurd. = ∼ ∠ACB. implying that D ∈ BC. The exterior angle of a triangle is greater than either of the interior angles that are not adjacent to it. we get a ray interior to ∠ACB that intersects AB at a point B . Proof. use the fact that ∠CAD is congruent to its vertical. the implication Suppose ∠CAD = = ∼ ∠CAB. An angle smaller than its supplement is acute. We conclude that ∠CAD ∼ = ∠ACB. 33 . By congruence of supplements. which. We conclude that ∠CAD > ∠ACB. Then ΔACD ∼ ΔACB by SAS. is impossible. Deﬁnition 5. Let ΔABC be a triangle. and ∠BAC. as shown above. Exterior angles play an important role in The Elements. ∠BCA. To show that ∠CAD > ∠ABC. Given ΔABC. ∠ACD would being that ∠ACD = be congruent to the supplement of ∠ACB. An angle greater than its supplement is obtuse. though. with exterior ∠CAD congruent to interior angle ∠ACB .k k k l h l h Figure 11: Right angles are congruent Deﬁnition 4. choose D ∈ AB with A ∗ B ∗ D∗ and AD ∼ CB. The exterior angles of the triangle are the supplements of the interior angles. Since D ∈ AB. Exercises Prove the following corollaries to Theorem 13. Suppose next that ∠CAD < ∠ACB. This gives us a triangle ΔAB C. If we construct an angle congruent to ∠CAD at C on the AB side of C. Theorem 13.
that extends on the side of AB that does not contain C. since they = = ∼ ∠EBD by construction. 34 . (Hint: Start by picking a point on the greater side that cuts oﬀ a segment with the length of a smaller side. (This is the angleangleside criterion.1. and AC ∼ BD by are vertical angles. Triangles ΔACO and ΔBCO have SAS by pons asinorum. ∠EAC = = construction. Every angle can be bisected. k. Pick A ∈ h and B ∈ k so that OA = OB. But ∠ABD. Let AB be given and let C be a point oﬀ AB. It follows that ∠AOC ∼ = ∠BOC. Axiom IV: Axiom of Parallels Deﬁnition 6. Every segment can be bisected.) 2. 3. Pick the point D on k that satisﬁes AC ∼ BD. hence. = Corollary 3. k) emanate from O. that OC bisects ∠(h. k). taking a ray. and ∠BCA ∼ B C A . Let AB have midpoint C. If AB ∼ A B . Two lines are parallel provided they lie in the same plane and do not intersect. the greater angle lies opposite the greater side. Proof. A triangle with two equal angles is isosceles. that CD bisects AB. that we deﬁne something that exists.) Theorem 14. then ∠BAC is exterior to ΔBAD. say. Construct an angle congruent to ∠CAB with vertex B. When we deﬁne something. CC has a nonempty intersection with AB. We claim that E cannot coincide with A or B. Note now that ΔACE ∼ ΔBDE by AAS: ∠AEC ∼ BED. it is important to establish that the deﬁnition is not vacuous. Proof. If it did coincide with A. We conclude that AE ∼ EB. = Since C and C lie on opposite sides of AB. then the = = = triangles are congruent. Note that ΔAOB is isosceles. or AAS. thus. Let ∠(h. that is. Call that intersection point E. In every triangle. ∠BAC ∼ B A C . which is interior to ΔBAD. was constructed to be congruent to ∠BAC.
C B A E D Figure 12: Segments can be bisected Consider a. The construction we described above thus yields the unique parallel to a through B. There there is at most one line in the plane. A and H. Using B as the vertex and the side of c containing H. and let B be a point of α not on a. Can there be more than one parallel to a through B? The Parallel Postulate says no. B. (Euclid’s Axiom) Let a be any line and A a point not on it. One of its interior angles would be ∠HAB and the constructed angle at B would be exterior. α. Let k be the constructed ray and take a point K ∈ k. this is impossible. a line in a plane. determined by a and A. then A. say D. Since we constructed this angle congruent to ∠HAB. Exercise Show that the parallel relation is transitive. Pick two points on a. IV. and D would form the vertices of a triangle. 35 . If a and k had a point of intersection. It is easy to see that this relation is symmetric. Let c = AB. construct the angle congruent to ∠HAB. Thus a and k are parallel. that passes through A and does not intersect a. It is important to realize that there is nothing in Hilbert’s system that forces uniqueness. If the line a is parallel to the line b write a b.
congruence of the corresponding or alternate angles implies the lines are parallel. Thus ∠B BC ∼ ∠A AC. that is. the angles congruent to ∠A and ∠B because of parallels. which is exterior to ΔABC. use supplements and verticals. Proof. = To get congruence of the other sets of corresponding angles and alternating angles. Theorem 16. Proof. 36 . c. Now suppose that a and b are lines intersected by the line c so that corresponding and alternating angles are congruent. By uniqueness of parallels. Let a ∩ c = A.B K A c H k a Figure 13: Construct a parallel to a given line Theorem 15. Consider the corresponding angles ∠A AC and ∠B BC. We have ΔABC then with interior angle ∠CAB congruent to the corresponding angle. i. Let C be a point on c. Let ΔABC be given. Let a and b be parallel lines intersected at points A and B respectively by a third line. suppose a and b do intersect. The result follows since the constructed angles add up to a straight angle. B ∈ b on the same side of c. Note that we get congruent angles to ∠A. we can construct a line through B parallel to A such that the angle at B corresponding to ∠A AC is congruent to ∠A AC. two right angles. the one congruent to ∠C as a vertical angle. If two parallels are interesected by a third line then the corresponding and alternate angles are congruent and conversely. Let c be the parallel to AB through C. By the discussion above. ∠B. The angles of a triangle add up to two right angles.e. which is absurd. b ∩ c = B and suppose a ∩ b = C. ∠C as shown.. Take points A ∈ a. b must be that line.
as are corresponding angles. c C A B Figure 15: Angles in a triangle add up to a straight line.C b B B a A A Figure 14: Alternate interior angles are congruent. 37 .
Axioms of Continuity 1. (Axiom of Completeness) An extension of a set of points on a line with its order and congruence relations that would preserve the relations existing among the original elements as well as the fundamental properties of line order and congruence that follows from Axioms IIII and V. along the ray from A through B. A thorough treatment of proportion. Theorem 8 shows how to construct a right angle to a given line. constructing parallels. 2. will pass beyond the point B.It is only at this point that Hilbert introduces circles. Exercises 1. Deﬁnition 7.1 is impossible. 38 . as in Book V of The Elements. there are two traditional construction that Hilbert does not do explicitly that we will want in our arsenal while we pursue theorems associated to circles. The last set of Hilbert’s axioms are to do with the geometry of the line. (Axiom of Measure or Archimedes’s Axiom) If AB and CD are any segments then there exists a number n such that n segments CD constructed continguously (endtoend) from A. Give a construction for drawing a perpendicular to a line at a given point on that line. M is the center of the circle. although we do not discuss the associated results in this part of the Grundlagen. Give a construction for dropping a perpendicular from a given point to a given line. 2. Besides the results we have about bisecting angles and segments. etc. We state the continuity axioms here and invoke them later. If M is any point in a plane α then the collection of all points A in α for which the segments MA are congruent to each other is called a circle. calls for these sorts of ideas.. The Axiom of Completeness comes up when we construct the projective plane by adding points to the Euclidean plane. V.
C are equidistant = = from F . they determine the circle centered at F with radius AF .3 The Nine Point Circle We use our work on Hilbert’s axioms to prove some of the theorems we need to discuss the nine point circle. C be noncollinear and consider ΔABC. Bisect AB and AC and call the midpoints of those segments D and E respectively. Our main sources now are [1] and [2]. A parallelogram is formed by 2 pairs of parallel lines. Since A. Proof. an interesting circle associated to a triangle. Draw the perpendicular to AB at D. the perpendicular to AC at E and let F be the point where those two perpendiculars intersect. We start our discussion by considering the circumcircle. Deﬁnition 8. B. There are several circles associated to a given triangle.F C E A D B Figure 16: The circumcircle of a triangle is centered where perpendicular bisectors intersect 2. Let A. the circle determined by the vertices of the triangle. We also = = have SAS for ΔADF ∼ ΔBDF so AF ∼ BF . B. Theorem 17. (Exercise: Why must they intersect?) We get SAS for ΔAF E ∼ ΔCF E implying that CF ∼ AF . Three noncollinear points determine a circle. The 39 .
= Exercise Show that adjacent angles in a parallelogram are supplements. This gives = us AB ∼ DC and AD ∼ BC. alternate interior angles are congruent giving us ∠CAB ∼ ∠DCA. Adding angles we get ∠A ∼ ∠C. Line segments that connect nonadjacent vertices are diagonals. = = Let E be the point of intersection of AC and BD. An altitude of a triangle is the perpendicular dropped from a vertex to the line determined by the opposite side. The diameters of a parallelogram bisect one another. I. The feet of a triangle are the points on the lines determined by the sides of the triangle where 40 . Parallelograms come up in Book I of The Elements and prove an important tool for understanding plane ﬁgures in general. DE ∼ BE. Similarly we can get so ∠ACB = = ∠B ∼ ∠D. using the diagonal BD. = Now ΔABC ∼ ΔADC by AAS. Also.34). since AC is a common side. Proposition 34 from Book I describes their basic features. Proof.D C E A B Figure 17: A parallelogram points where the lines intersect are the vertices of the parallelogram. Deﬁnition 9. AC cuts parallels AD and BC = ∼ ∠CAD. Opposite angles of a parallelogram are congruent. Since AC cuts parallels AB and CD. Theorem 18 (Prop. The opposite sides of a parallelogram are congruent. Similarly. (Why must they intersect?) We have ASA for ΔAEB ∼ ΔCED with corresponding sides = ∼ AE = CE.
Feuerbach circle. In this part of The Elements. You may remember from your high school geometry course that the medians of a triangle intersect in a single point. 41 . starting with a set of propositions from Book I. when we write ΔABC = ΔA B C . we will paraphrase Euclid. or Euler circle of a triangle ΔABC is the circumcircle of the orthic circle of ΔABC. Euclid’s reference to plane ﬁgures being equal means not that they are congruent. I. they are concurrent. it means that the areas of the two triangles are equal. 37. as translated by Heath. that is. To avoid confusion. Theorem 19 (Prop. we need results about parallels and ratios. Deﬁnition 10. The line segment joining the vertex of a triangle to the midpoint of the opposite side is a median of the triangle.35. 38. but that they enclose equal areas. To prove this.C F1 F3 A B F2 Figure 18: The orthic triangle they intersect the altitudes. The orthic triangle of ΔABC is the triangle determined by the feet. Parallelograms (respectively triangles) on the same base or equal bases that lie on a line have equal areas if and only if they are in the same parallels. 39). The ninepoint circle. 36. Also.
Euclid (Book V) and Hilbert (Axioms of Continuity) both set up the tools for discussing comparitive lengths and areas. Proposition 7 (Prop.2). Proposition 6 (Prop.1). If there are positive integers m and n such that m AB ∼ n CD then AB and CD are commensurate and we write = AB/CD = m/n. Being in the same parallels amounts to having the same height.Figure 19: Rectangles in the same parallels The notion of two parallelograms being “in the same parallels” makes sense: bases form one line and opposite sides form another. the ratio of their areas is equal to the ratio of their bases. When we write ΔABC/ΔEF G we mean the ratio of the areas of the triangles and when we write AB/CD we mean the ratio of the lengths of line segments. For example. we can accept the idea of comparing areas. the area of a parallelogram is twice the area of a triangle formed by adjacent sides and a diagonal. if a line cuts two 42 . If triangles (respectively. For example. Then we can compare the lengths of diﬀerent segments. parallelograms) have the same height. we can make sense of the idea of multiplying a line segment by 2 to get a new line segment with double the length of the original. Conversely. VI. VI. Area is trickier since it involves the notion of multiplying line segments instead of adding them but without getting too involved. We state Proposition 1 from Book VI and Proposition 39 from Book I without proof. Two triangles are in the same parallels provided their bases form a line parallel to that formed by the vertices of the opposite sides. A line parallel to one side of a triangle cuts the sides it intersects in equal proportions.
intersecting AB at D and AC at E. ΔCDE/ΔADE = CE/AE. view AD as the base of ΔADE and BD as the base of ΔBDE. It follows that ΔBDE/ΔADE = ΔCDE/ΔADE. show that AB/AD = AC/AE. Proof. By Prop. We need establish that DE BC. Similarly. Consider ΔABC with DE BC. Two triangles are similar provided corresponding sides are in equal proportion. AC be cut so that BD/DA = CE/EA. ΔBDE = ΔCDE. VI. This is the way we usually use the theorem. Exercise In reference to the last theorem. ΔCDE have a common base. Next. It follows that ΔBDE/ΔADE = ΔCDE/ΔADE so ΔBDE = ΔADE.A D E B C Figure 20: Parallel to the base cuts the other sides in equal proportions sides of a triangle in equal proportions. Next is a theorem about similarity. Since ΔBDE and ΔCDE have coincident bases and are in the same parallels. it is parallel to the third side of the triangle. which proves the result. It follows that BD/AD = CE/AE. which proves the ﬁrst part of the result. Euclid’s proof uses a construction in which the two triangles are arranged so that bases lie on a line and one vertex is shared. We have ΔBDE/ΔADE = BD/DA. These two triangles have the same height as determined by the perpendicular to AB from E. ΔCDE/ΔADE = CE/EA. The argument we use is maybe a variant on that preferred 43 . I. As ΔBDE. they must be in the same parallels (Prop. Deﬁnition 11. Now let the sides AB. Similarly. ΔBDE/ΔADE = BD/AD.39).1.
Exercise 1. Now the exercise following Prop. then by uniqueness of angle construction.4). 10. Let A = A. VI. 44 .) Here we use the results we developed in Hilbert’s system. p. This implies the result. and ΔA B C we can make a copy of ΔA B C ∼ ΔA B C . the second ray has to coincide with AC. 202. Show that if corresponding sides of two triangles are in equal proportions. Proof. corresponding sides are in equal proportion. as follows. then the third pair of sides of the two triangles are in the same portion as the others. Let B be on the B side of A so that A B ∼ A B . Show that if two pairs of sides of two triangles are in equal proportions.C C'' A=A'' B'' B Figure 21: Equiangular triangles have corresponding sides in equal proportions by De Morgan (see [2]. Notice that by SAS. We can get another pair of sides in the same ratio by constructing ΔA B C starting with B = B. Given equiangular triangles ΔABC. p. The proof for the following we get from [1]. ∠C = Hilbert’s one theorem on parallels implies that B C BC. for example.2 above gives us AB/A B = AC/A C . VI. so that if we construct a = copy of ∠A using AB for one ray and taking the other ray on the C side of AB. and let C = on AC be chosen so that A C ∼ A C . ∠B ∼ ∠B and = = ∼ ∠C. Theorem 20 (Prop. and the angles between are congruent. 2. In equiangular triangles. then the triangles are equiangular.
Since the diagonals of a parallelogram bisect one another. the Euler line is OG. Exercise Prove the following variant on Proposition 7.) Let O be the circumcenter and G the centroid of the triangle ΔABC.Theorem 21. This makes EF GH a parallelogram. then the medians of ΔABC are perpendicular to the sides. Since EF cuts the sides of ΔABC into the same proportions. Consider ΔOGA and ΔAHG. Theorem 22. (See Figure 22.) The center of the circumcircle is called the circumcenter. Proof. in other words. say the ones determined by the midpoint E of AC and the midpoint F of AB. Consider ΔABC and let D be the point where two medians intersect.e. for example GA. Similarly. A G. we consider the Euler line of a triangle. If the triangle is not equilateral.. Similarly. it is parallel to BC. Using a diﬀerent median. implying that AB ∼ AC. but no more. (The discussion and proof are from [1]. Recall that the segment from vertex to centroid. when we consider ΔBCD. i. There are several diﬀerent “centers” associated to a triangle. Let G and H be the midpoints respectively of CD and BD. Before we prove the theorem. (Recall that O is the point where perpendicular bisectors of sides of ΔABC intersect. we can repeat the = argument to show that the triangle is equilateral.) This gives us congruent triangles ΔAA C. Proof.) If O = G. 45 . The point of concurrency is the orthocenter of the triangle. it follows that the point D trisects segments EB. Show that OA/OA = OB/OB if and only if AB A B . GH = 2OG. We did not prove this. The altitudes of a triangle are concurrent. it must trisect the third median. The next result gives us one more center of a triangle. which implies the result. we see GH BC. p. but it is true that the circle determined by the vertices of a triangle is unique. the circumcenter is unique. The medians of a triangle are concurrent. F C. 17. referring to Figure 23. is twice the segment from the midpoint of the opposite side to centroid. Choose H ∈ OH with O ∗ G ∗ H and OH = 3OG. (That was proved by Archimedes. which is the center of gravity of a triangular plate of uniform density. the perpendicular bisectors of a triangle are concurrent. This is another way of saying that diﬀerent circles can intersect in one or two points. that is GA = 2A G. The point where the medians intersect is the centroid. ΔBA A. As a consequence.
46 . the triangle is equilateral.A B A C Figure 22: If medians and perpendicular bisectors coincide. A B O A B Figure 23: The triangles are similar if and only if AB A B .
the angle thus determined must be a right angle. O. We have not said much about circles in general and angles associated with them. It follows that C is the circumcenter for ΔABC. AH OA . The circumcenter of a right triangle is the midpoint of its hypotenuse so the hypotenuse is a diameter of the circumcircle. Let ΔABC be a triangle with ∠C a right angle. By the exercise. Similarly BH is perpendicular to CA. centroid. OA is the perpendicular bisector of BC. Lemma 4.2. If the two other points are opposite ends of a diameter. The perpendicular bisector to CB is parallel AC so intersects AB at C as well. by Prop. VI. These observations together with the last lemma prove the following. Now use pons asinorum and supplementary angles to complete the proof. 47 . Corollary 4. The circumcenter. C . H. When the rays deﬁning the angle are determined by two other points on the circle. the circumcenter. Proof. We can see this by marking the center of the diameter and joining it to the vertex of the angle. Thus AH is perpendicular to BC. Fix a point on a circle to be the vertex of an angle. and CH to AB. Moreover. this is an inscribed angle. The perpendicular bisector to AB is parallel to CB so intersects AB at its midpoint. Moreover. and orthocenter of a triangle are collinear. It follows that H is the intersection of the altitudes of ΔABC. and the orthocenter. the hypotenuse AB must be a diameter for the circumcircle.Figure 24: The Euler line passes through the centroid G.
and the feet of the three altitudes of the triangle. 48 .e. 18. all lie on a circle.A B' C' C A' B Figure 25: The circumcenter of a right triangle is the midpoint of its hypotenuse. We leave it as an exercise to prove in just one case of the reader’s choosing. This was an easy theorem to prove but is an instance of a wellknown theorem that is a bit more work. The proof of the next is from [1]. The midpoints of the three sides of a triangle. Theorem 23. the midpoints of the three segments joining the vertices to the orthocenter. It is an easy step from there to see that a circle is the locus of points determined by the ends of a diameter. p. i. Exercise Let α be an angle inscribed in a circle.. Theorem 24 (The Nine Point Circle). the vertex of β is the center of the circle and its sides are the radii determined by the points where the sides of α intersect the circle. Prove that β = 2α. An angle inscribed in a circle is a right angle if and only if its sides emanate from points at either end of a diameter. Choose a case to make the proof more accessible. Let β be the associated central angle. together with the vertices of all the right angles with sides passing through the endpoints of the diameter.
The central angle is β. 49 . β α Figure 27: The inscribed angle is α.Figure 26: An inscribed angle is a right angle if and only if its sides are determined by the endpoints of a diameter.
Prop. An argument similar to the one just advanced shows that this is a rectangle as well. Next consider ΔBHC and note that since B and C cut BH and CH in half. C B AH.2 implies C B BC. and C be the midpoint of AB. Considering ΔAHC. we turn our attention to the quadrilateral C A C A . Let H be the orthocenter and let A be the midpoint of AH. the diameters are equal. let D be the foot of the altitude from A to BC. Similarly. are concurrent so are diameters of a circle with center this point of concurrency. This gives us a parallelogram. B the midpoint of BH. so D must lie on the circle with diameter A A . Next.. We claim that in a rectangle. The three diameters C C . C B C B . It shares a diameter with C B C B . The claim follows once we notice that ΔC B C ∼ ΔB C B . which is parallel to B C . B C AH. B be the midpoint of AC. Notice that ∠A DA is a right angle. Let ΔABC be given and let A be the midpoint of BC. Finally. 50 . This completes the proof. Since C B cuts sides AB and AC in half. we see that since B cuts AC in half and C cuts CH in half. Since C B and B C are both parallel to AH which is perpendicular to BC. A A . B C BC as well. B B . the right triangles formed by = the two diﬀerent diameters of rectangle C B C B . VI.A A" C' H B" B D A' C" C B' Figure 28: Proof of the nine point circle Proof. and C the midpoint of CH. we see that C B C B is actually a rectangle. C C . viz. The same argument applies to the other three feet.
51 . nonintersecting equivalence classes. The eﬀect of adding these points is to remove the special status of parallel lines. If not. any two lines intersect. so the line at inﬁnity must be the only line containing a given pair of ideal points. From our point of view right now. construct a parallel to the chosen line through the ordinary point. each class determined by a direction. called an ideal point. Further decree that the collection of ideal points comprise a line. The title of this subsection is a bit misleading: there are many diﬀerent types of projective planes. by construction. If one point is ordinary and the other ideal. we are done. The resulting collection of points and lines is the real projective plane. In other words. If the two points are ideal. we make parallelism an equivalence relation. the ideal line or the line at inﬁnity. We claim that the ordinary point belongs to exactly one line in that class.4 The Projective Plane The material here is taken mostly from [3]. the theorem is just a restatement of two of Hilbert’s incidence axioms. Theorem 26.) Decree that all the lines in the class contain the associated ideal point. pick a line out of the parallel class associated to the ideal point. though. Allowing that a line in the Euclidean plane is parallel to itself. All the lines in given class then have the same direction. If the ordinary point is on that line. call the points of the original Euclidean plane ordinary points. the ideal point identiﬁes a single parallel class. Proof. Two distinct lines meet in one and only one point. to the Euclidean plane for each equivalence class of lines. they both lie on the line at inﬁnity. Every ordinary line in the plane contains exactly one ideal point. the projective plane is what you get when you add ideal points to the Euclidean plane. there is only such parallel. The formal construction is as follows.2. The lines in the Euclidean plane are thus partitioned into nonempty. Add a single point. To prove the claim. If the two points are ordinary points. This proves the claim and the theorem. as developed via using Hilbert’s axioms. Two distinct points determine a unique line. By Hilbert’s Parallel Axiom. (To distinguish them from the ideal points. in a projective plane. Theorem 25.
Axioms of Continuity 1. Two points determine a line. will pass beyond the point B. The critical nature of these two theorems is that they establish once and for all that we need not distinguish between ordinary and ideal points. they belong to the same parallel class so intersect in a single ideal point. 2. (Axiom of Measure or Archimedes’s Axiom) If AB and CD are any segments then there exists a number n such that n segments CD constructed continguously (endtoend) from A. The Axiom of Completeness is particularly germane. If they do not intersect. 52 . V. the ordinary line belongs to a single parallel class so intersects the ideal line in the unique ideal point associated to that parallel class. By our choice of ideal points. If one line is ordinary and the other ideal. (Axiom of Completeness) An extension of a set of points on a line with its order and congruence relations that would preserve the relations existing among the original elements as well as the fundamental properties of line order and congruence that follows from Axioms IIII and V. If the lines are both ordinary. the axioms of continuity. 3. along the ray from A through B. The line determined by two points is unique.Proof. What about the others? I. In the projective plane. nor between ordinary lines and the line at inﬁnity. There exist at least two points on a line. There exist at least three noncollinear points. We start with the last set. It guarantees that we cannot add a single point to a line in Euclidean space without losing fundamental properties. we know that the parallel axiom no longer applies.1 is impossible. all points are created equal and all lines are created equal. Hilbert’s Axioms of Incidence 1. This is a good time to review Hilbert’s axioms with an eye towards which might apply in this setting. 2. then Hilbert’s Theorem 1 says that they intersect in exactly one point or they do not intersect.
The plane determined by three noncollinear points is unique.” A similar problem ensues if we try planting ∞ inﬁnitely far to the left on an ordinary line. This is a logistical problem that does not seem easily resolved as B would presumably occupy a position farther to the right than “inﬁnitely far. If two points of a line lie in a plane. Three noncollinear points determine a plane. z are the real numbers associated to the points A. Think of putting it inﬁnitely far to the right. 7. there is C ∈ AB with A ∗ B ∗ C. 5. 53 . 8. 2. y. on a line containing an ordinary point A. for example. This is part of a course on the real line. If we add a single point to a line. we get that in a Euclidean plane. Two planes with nonempty intersection have at least two points in common.4. If we hang onto the order axioms. we need to ﬁnd a position for it on a line. then we must be able to ﬁnd B so that A ∗ ∞ ∗ B. then every point of that line lies in that plane. B. If we persist. C. it either means x < y < z or z < y < x. it cannot be associated to a number. Every plane contains a point. Hilbert’s Axioms of Order 1. We could think of an ideal point as representing ∞ but for the purposes of order. If A ∗ B ∗ C then A. Putting these together with the axioms of continuity. Of any three points on a line there exists no more than one that lies between the other two. 6. We are safe in assuming that Hilbert’s incidence axioms apply to the projective plane. Given A. there is a onetoone correspondence between the points on a line and the real numbers. The Order Axioms are a diﬀerent matter. B ∈ AB. C are distinct points of a line and C ∗ B ∗ A. B. II. There are at least four noncoplanar points. 3. Nothing here is violated by the addition of ideal points. (There is actually quite a bit more to this than we are explicating here. where x.) Then when we write A ∗ B ∗ C.
that is. Without it. 4. Once we accept that the order axioms have to go. 6. we are stuck with a world that has no rays. the FubiniStudy metric is a commonly used device. The reader should be warned here: there are ways of deﬁning metric systems on the projective plane. no segments. Two distinct points determine a unique line. 3. But there is no natural extension of these concepts as they apply in Euclidean geometry. There is a onetoone correspondence between the real numbers and all but one point of a line. no three of which are collinear.looking for another location on a line for an ideal point. no angles. The notion of congruence is now gone. Exercise Verify that the object in Figure 29 represents a projective plane. we get more general projective planes. Identify the points and the lines. This can be illustrated in the projective plane by restating the axioms switching the words “point” and “line.” When we do that we get: 54 . Axioms for the Real Projective Plane 1. 5. the only positions left are between ordinary points. 2. those points are already spoken for by real numbers. Each line has as least three points. Eﬀectively. How many of each do we have? This is called the Fano plane.4. There exist four points. 2.” and switching the phrases “lie on” and “meet in. in particular. There exists a line.1 Duality The most striking feature of projective geometry is the principle of duality. that it satisﬁes the ﬁrst ﬁve axioms above. The last axiom is what make this the real projective plane. Two distinct lines determine a unique point.
There is a onetoone correspondence between the real numbers and all but one line going through a point.Figure 29: The Fano Plane Dual Axioms for the Projective Plane 1. Exercises 1. There are four lines. no three of which are collinear (c) Two lines. no three of which are concurrent. Note that these could serve equally well to deﬁne the real projective plane. Two distinct lines determine a unique point. (a) The set of all points on a line (b) Four points. 5. There exists a point. Two distince points determine a unique line. and a point on neither line 55 . 6. 3. 4. 2. Write the dual of each of the following statements in the projective plane. Prove that the Axioms and Dual Axioms for the Projective Plane are indeed equivalent. Each point belongs to at least three lines. 2.
Exercise Is a complete quadrangle selfdual? Explain. distinct lines through the given point meet the given line in distinct points. RP 3 as follows. in other words. It is.2 Projective Space We construct real projective three space. Deﬁnition 13. The proof of the theorem relies on embedding the projective plane in projective space. (g) Given a line and a point not on the line. This is not a mere device to facilitate the proof of the theorem. we need some terminology and some information about the structure of projective space. Opposite sides of a complete quadrangle are a pair of sides without a common vertex. These planes are called nondesarguean. triangles and trilaterals are the same objects. Desargues’s theorem is not true. 2. central to the validity of the theorem. The dual of a complete quadrangle is a complete quadrilateral. there are projective planes that cannot be embedded in space and for these.4.(d) The line determined by a given point and the point of intersection of two given lines (e) All lines in the projective plane (f) Distinct concurrent lines have only one common point. The real projective plane is desarguean. Add a point to each parallel class and say 56 . A complete quadrangle is a conﬁguration in the projective plane composed of a set of four points (vertices). It should be clear that triangles are selfdual. In other words. A trilateral is a set of three nonconcurrent lines and the points they determine. in fact. To prove that. together with the six lines (sides) they determine. no three of which are collinear. Deﬁnition 12. A triangle is a set of three noncollinear points (vertices) and the lines they determine (sides). Start again by thinking of the lines of Euclidean three space as partitioned into equivalence classes determined by direction. Desargues’s theorem is major result of projective geometry having to do with relationships of triangles in a plane or in space.
maybe it contains a second ideal point. namely. Theorem 27. If it does. It follows that the ordinary intersection ∩ α is at most one point. As above. The unique ordinary plane determined by and 1 is α1 and the unique ordinary plane determined by and 2 is α2 so α cannot contain . Since each of 1 and 2 is coplanar with . Theorem 28. P . It follows that we are adding one collection of ideal points associated to each parallel class of ordinary planes. Two distinct points determine a unique line.that point lies on all lines in the associated parallel class and on no line from any another parallel class. Since all the points of this ideal line lie in all the planes belonging to a given parallel class. 57 . Two distinct planes meet in a unique line. We conclude that there cannot be a second ideal point in the intersection of the two planes in this case. neither 1 nor 2 is parallel to but 1 2 . An ideal point P lies in a given plane α if and only if α contains a line in the parallel class associated to P . If α is another plane parallel to α. they intersect in a unique ordinary line. Since P ∈ . Note that 1 and 2 determine an ordinary plane. let 1 ∈ α1 and 2 ∈ α2 be ordinary lines in the parallel class determined by P . implying that P1 = P2 . Proof. the ideal line also lies in all the planes of the parallel class. To this. The proof of Theorem 25 is largely valid here. which contains no other points or lines. Say 1 ∩ = P1 . Suppose the planes α1 and α2 are nonparallel and ordinary. The result follows by our construction of RP 3 in this case. the set of ideal points and lines together are to comprise an ideal plane. it contains lines in the parallel class associated to P so contains P as well. as well as in the remaining case where α1 is ordinary and α2 is ideal. α. While it is clear that α1 ∩ α2 contains no other ordinary points. By Hilbert’s system. . we must append the ideal point P that identiﬁes the parallel class of . leading us to a contradiction. We decree that this collection of points comprise an ideal line. The case we must speak to is when both points are ideal since there is more than one ideal line in RP 3. Suppose α1 and α2 are ordinary and parallel. each must meet in an ordinary point. Finally. The collection of ideal points in a given plane comprises a line at inﬁnity or an ideal line. Proof. call the new points ideal points and the points from Euclidean space ordinary points. and 2 ∩ = P2 . that 1 2 .
A line and a plane not containing the line meet in one and only one point. this line would have to belong to α as well. Suppose A is ordinary. But it can only belong to one so this is the unique plane the two lines determine. Let = AB and m = AC. The proof is left as an exercise. Proof. If both lines are ordinary. they determine the unique ideal plane by construction. Theorem 30. then the set of ordinary points in their intersection is either empty or a single point. Suppose one is ordinary and one ideal. this proves the result. If the intersection contains a single ordinary point. C are noncollinear by assumption. the given line cannot be parallel to any lines in the plane so 58 . If both are ideal. The ideal plane contains all ideal lines so line and plane cannot both be ideal. If they belonged to a second ideal line. they determine a unique plane. A. C be the three noncollinear points. Proof. determined by O and Q. Proof. the ordinary line must belong to one of these parallel planes. If two distinct lines meet in a point. The ideal line determines a parallel class of ordinary planes.Let P and Q be ideal points. Since two ordinary lines determine a unique ordinary plane. Since the ordinary line intersects the ideal line. Proof. Theorem 31. It is clear that and m are distinct thus determine an ordinary plane α. Three distinct planes not containing a common line intersect in a unique point. Theorem 32. they determine the unique ideal plane by construction. Now P and Q belong to the ideal line in α. B. Since they contain an ordinary point. Theorem 29. and m are ordinary lines and they are distinct since A. Three noncollinear points determine a unique plane. it is clear that there are no ordinary lines containing both P and Q. B. If both are ordinary. If empty. determined by O and P . By the deﬁnition of ideal points. If all are ideal. then the line is parallel to a single parallel class of lines in the plane so the line and plane share the single ideal point associated to that class of parallels. and m. contradicting the last result. Take any ordinary point O and consider the ordinary lines . they determine a unique ordinary plane by Hilbert. Let A.
Arrange the theorems in the previous subsection in dual pairs. etc. If the dimension of a point is 0. not just real projective spaces. 59 . lines with n − 2dimensional objects. Exercise Prove that if two points lie in a plane. then duality in projective nspace allows us to switch points with n − 1dimensional objects. The given plane is not in that parallel class so intersects any of the planes in it in an ordinary line. If both are ordinary. If the point P is ideal and the line ordinary. Exercise 1. and planes in projective 3space. In projective 3space. like RP 2 and RP 3 . This is the unique plane determined by and P . Now suppose P is ordinary and ideal. etc. Theorem 33. then consider the parallel class of planes associated to the line.there can be no more than the single ordinary point in the intersection of the line and plane. If the line is ordinary and the plane ideal. points and planes are dual. they determine the unique ideal plane. those based on R. There is exactly one ordinary plane in the parallel class determined by that contains P . Proof. the line they determine lies in that plane. This ideal line determines a parallel class of ordinary planes exactly one of which contains . A line and a point not on the line determine a unique plane.3 Duality in Projective 3Space The duality principle applies to all projective spaces. If the line is ideal and the plane ordinary. the result follows by Hilbert. a plane 2. of a line is 1. while lines are selfdual. the line intersects the plane in the unique ideal point associated to its parallel class. consider the ideal line determined by P and the ideal point of .4. In other words. If both are ideal. Now we see that we need not distinguish between ordinary and ideal points. the switch is allowed between objects when the sum of their dimensions is one less than the ambient space. 2. lines. that is. This line belongs to a unique parallel class and the associated ideal point must be the unique point of intersection of the ideal line and the ordinary plane.
If ΔABC and ΔA B C are perspective from a point O since AA and BB intersect. 60 . the planes π. BC ∩ B C = P3 ∈ α ∩ α . AA and CC intersect in a plane π . is an axis of perspectivity. ΔA B C are perspective from a point O provided AA . It follows that P1 . Now AA = π ∩ π .4 The Theorem of Desargues Deﬁnition 14. CC are concurrent at O. As intersecting lines.4. In this case. and BC ∩ B C = P3 are collinear. In projective 3space. It follows that the three lines AA . Write the dual for each of the following. P3 lie on the line α ∩ α . For the ﬁrst part of the proof. π intersect in a point. Say ΔABC determines a plane α and ΔA B C determines α . Similarly. Theorem 34 (Desargues). B. Two triangles ΔABC. Similarly. as statements or objects in projective 3space.2. Proof. AB and A B determine a plane π in which AA and BB intersect. BB . CC are concurrent at O. AC ∩ A C = P2 . O. they determine a plane which contains AB and A B which thus intersect in a point P1 . and CC = π ∩ π . P2 . BB . π . Next suppose ΔABC and ΔA B C are perspective from a line. two triangles are perspective from a point if and only if they are perspective from a line. In this case. we assume that the two triangles are not coplanar. Note that P1 ∈ α ∩ α . and BB and CC intersect in a third plane π . O is a center of perspectivity. The triangles are perspective from a line provided the points AB ∩ A B = P1 . (a) the set of points on a line (b) the set of points on a plane (c) the set of planes containing a given line (d) the set of lines passing through a common point (e) the set of lines lying on a given plane (f) the set of planes in space (g) the set of lines in space (h) the set of all points in space 2. AA and CC determine a plane so that AC ∩ A C = P2 with P2 ∈ α ∩ α . Finally. Since A. C are noncollinear. BB = π ∩ π .
both diﬀerent from O. implying that A. This gives us that A B ∩ AB and A B ∩ A B fall on this line of intersection. The other sides of the two triangles meet on this line as well. not lying in π. Exercise Figure 30 shows what is called a Desarguean conﬁguration. AB and A B intersect there as desired. the planes determined by P and each of the three sides of the triangle ΔABC would coincide. Produce points B . By construction. B . For the second part of the theorem. Let be a line through O. Notice that the points A . Since A B” meets this line just once. Show that any one of the ten points can serve as 61 . while ΔA B C and ΔA B C are perspective from P . C cannot be collinear: if they were.P1 A A' B B' O C P3 C' P2 Figure 30: The theorem of Desargues is true in Euclidean space as long as all the intersections occur. for the same reasons. The lines P A and P A must intersect in a point A since they lie in the plane determined by intersecting lines AA and P P . assume that ΔABC and ΔA B C lie in a plane π. an arrangement of ten points. C in a similar manner. P distinct points on . Choose P . B. ΔABC and ΔA B C are perspective from P . The line of perspectivity in both cases is the intersection of π with the plane of ΔA B C”. We leave the last part of the proof to the reader. We thus have a new triangle ΔA B C . Suppose the two triangles are perspective from a point O ∈ π. C were collinear.
and planes should be and then. Lemma 5. 3 Aﬃne Planes The most familiar algebraic structure is possibly the vector space. 2. 3.. Let u ∈ V . u+W = u +W if and only if for every w1 ∈ W .1. .e. 3. It follows that v ∈ u + W ∩ u + W . In this case. w2 ∈ W with v = u + w1 = u + w2 . that they satisfy Hilbert’s axioms. u + W ∩ u + W is empty or u + W = u + W . Deﬁnition 15. The W coset determined by u is u + W = {u + w  w ∈ W }. Rn . . For any vector space V with subspaces W.2. vn >. Let V be a vector space. if dim V = 2 and dim W = dim W = 1. then there are w1 . then W = W means precisely that {w. lines. 62 .the center of perspectivity for a Desarguean conﬁguration with the same ten points. which is true only if u − u ∈ W . i. 1. W a subspace of V . u + W ∩ u + W = if and only if W = W . w } is a basis for V . vn } by < v1 . . suppose u = αw + βw and u = α w +β w . If v ∈ u + W ∩ u + W . Denote the span of a set of vectors {v1 . and u − v = (β − β)w ∈ W . and vectors u. 1. The results that follow are from [5]. Take v = α w +βw and notice that u−v = (α−α )w ∈ W . our goal is to understand how we can use vector spaces to model Euclidean and projective spaces. . The ﬁrst thing is to establish what points. In this section. u + W = u + W . there is w2 ∈ W with u + w1 = u + w2 which is true if and only if u − u ∈ W . Proof. We take vectors to play the role of points. p. 101. starting with Lemma 3. u ∈ V. Playing the role of lines. If W =< w > and W =< w >. W . . . 2. we have cosets of one dimensional spaces. in particular. u + W = u + W if and only if u − u ∈ W . . .
etc. v + (v − v ) = v ∈ v + W so by Lemma 5. 0 and 1. R is an example of an ordered ﬁeld because aside from satisfying the axioms for a ﬁeld. Example 1. ¯ = {kp  k ∈ Z}. Let a ⊕ ¯ := a + b and let a ¯ = ab. 2. . Proof. a (b ⊕ c) = (a b) ⊕ (a c). i. 0 1 ¯ = {2 + kp  k ∈ Z}.1. Two one dimensional cosets intersect in at most one point.1. a < b implies that a + c < b + c and if 0 < a. v ∈ u + W ∩ u + W . For every a ∈ F . some w1 ∈ W and v = u + w2 . 0 is an identity for ⊕. which implies that v − v ∈ W . i. Suppose v.Lemma 6. Every element a ∈ F . Similarly. Deﬁnition 16. v determine a unique 1dimensional coset in V . 1 is an identity for .e. A ﬁeld is a set F with at least two distinct elements.e. Proof. of integers mod p. Two distinct vectors v. That is. 63 . Lemma 7. as in the next example. 1. thus that W =< v − v >= W . then b < c implies that ab < ac. In particular. Let Zp = {¯ ¯ . 3. . together with two commutative and associative binary operations ⊕ and such that the following axioms are satisﬁed. b. The again by Lemma 5. ˆ 4. p − 1} designate the set of residue classes 0. 0 ⊕ a = a for all a ∈ F . Our favorite examples of ﬁelds besides R are Q. Let W =< v − v >. c ∈ F . v+W = v +W . v − v ∈ W implying either v = v or < v − v >= W = W . 1. . We leave it 2 ¯ b ¯ b as an exercise to verify that ⊕ and are welldeﬁned and that they satisfy the axioms for a ﬁeld... ¯ = {1 + kp  k ∈ Z}. some w2 ∈ W so that v − v ∈ W . Since v − v ∈ W . there is a ∈ F such that a ˆ 5. Example 2. 1 a = a for all a ∈ F . Then v = u + w1 . it has an ordering that respects the ﬁeld operations of addition and multiplication. For all a. C and the integers modulo p. Suppose u+W is another one dimensional coset containing v and v . v + W = v + W = u + W . a = 1. there is a ∈ F such that a ⊕ a = 0. except 0.
the ﬁeld is said to have characteristic 0. β). every line in V must have at least two distinct points. R. there is k such that times 1 ⊕ . multiplication. Let F be a ﬁeld with at least three elements. In particular. When there is one such k. designate this inverse 1/2. Unless noted to the contrary. for any α ∈ F . v ∈ V = F 2 . Let u.Now we are free to consider vector spaces with arbitrary ﬁelds of scalars. When there is no such k. It follows that for any v ∈ u + W . the study of which extends beyond the scope of this course. ⊕ 1 = 0. and Q all have characteristic zero. there must be a minimal value of k for which it is true. Then in F we have 1 ⊕ 1 = 2 = 0 so that 2 has an inverse with respect to . C. β ∈ F with α⊕β = 1. Zp has characteristic p. v ∈ means = u+ < v − u > so that αu + βv ∈ for all α. Let V be a two dimensional vector space over F . we assume V is a vector space deﬁned over some ﬁeld F . That minimal k is called the characteristic of F . . we can coordinatize the points on the line = u + W by (α. where α. k This also happens for certain inﬁnite ﬁelds. These α and β are then the aﬃne line coordinates for v with respect to v and u. For convenience. Does 1/2⊕1/2 = 1 in any ﬁeld F with characteristic diﬀerent from 2? To see that the answer is yes. Theorem 35. which satisﬁes Hilbert’s axioms of incidence for a plane. Then we can deﬁne the midpoint of u and v to be (1/2)(u + v). since F must have at least two elements. We have u. It is not diﬃcult to verify that the characteristic of a ﬁeld is either zero or a prime number. β ∈ F satisfy α + β = 1. Consider that if u+W is a one dimensional coset in V and v ∈ u+W . . we invoke the distributive law to get (1/2 ⊕ 1/2) = 1/2 (1 ⊕ 1) = 1/2 64 2 = 1. On the other hand. . there is a point (1 − α)u + αv ∈ u + W : the set of points on a line in V has the same cardinality as F . In the ﬁnite ﬁelds we have seen. We have proved the following. then v − u ∈ W implies that W =< v − u >. there is α ∈ F such that v = u+α(v −u) = (1−α)u+αv. The collection of vectors and one dimensional cosets of V form the points and lines of the aﬃne plane Π(F ). In other words. Assume now that the characteristic of F is not two.
A median of the triangle is a line determined by one vertex and the midpoint of the line determining the opposite side. {u. w} = {(0. u2 ). This gives us z = (1 − (λ + μ))u + λv + μw. We reconsider Theorem 21 in Π(F ). Exercise Let {u. By symmetry. Using the distance formula to ﬁnd the distance from u to (1/2)(u + v). It follows that every point of Π(F ) has a unique triple of scalars (α. Verify that the medians of the triangle are m1 = u+ < u − (1/2)v − (1/2)w > 65 . w − u} is a basis for F 2 .Next. (0. μ) are (1 − (λ + μ). 1)}. (u2 + v2 )/2). Writing u = (u1 . w}. v. Proof. Lemma8 shows that if we ﬁx any triangle in Π(F ). (1. γ) where z = αu + βv + γw. v = (v1 . there are unique λ. We leave the proof as an exercise. the aﬃne coordinates of a point (λ. A triangle in Π(F ) is a set of 3 noncollinear points and the lines they determine. we want to verify that in Π(R). These are the aﬃne coordinates of z with respect to {u. Exercise Check that when Π(F ) = R2 and {u. v. Deﬁnition 17. w} be a triangle in Π(F ). v2 ) we have (1/2)(u + v) = ((u1 + v1 )/2. the distance is the same from u to (1/2)(u + v). {u. we get d(u. μ). 0). v). v. w} ⊂ Π(F ) is a triangle if and only if {v − u. (1/2)(u + v)) = = (1/2(u1 + v1 ) − u1 )2 + (1/2(u2 + v2 ) − u2 )2 (1/2(v1 − u1 ))2 + (1/2(v2 − u2))2 = 1/2 d(u. then for each z ∈ Π(F ). v. v. μ ∈ F with z − u = λ(v − u) + μ(w − u). β. and an ordering of its vertices. 0). where α + β + γ = 1. (1/2)(u + v) is the same distance from u as from v. Lemma 8. λ. w}.
w} be a triangle in Π(F ) with medians m1 . Letting z = u + v + w. let {u. We conclude that m1 and m2 intersect if and only if the following equations can be solved simultaneously. Similarly. 66 . ξ = δ. To make it easier to think about. In F we have 2 ⊕ 1 = 0 so −1 = 2 and 1 = −2. v. From the ﬁrst equation ξ = −2/3 = δ. Note that z = (1 + δ)u − (1/2)δv − (1/2)δw. there is α ∈ F with u − v = α(v − u) + α(w − u). w − w} is linearly independent. We also have aﬃne line coordinates for z determined by m1 : z = u + δu − (1/2)δv − (1/2)δw. If m1 = m2 . Let z ∈ m1 and say (α. 1/3). 1 + (−1/2)ξ = 1 + δ ξ = (−1/2)δ (−1/2)ξ = (−1/2)δ By the last equation. Since {v − u. It follows that the medians are either identical or parallel. 1/3.m2 = v+ < v − (1/2)w − (1/2)u > m3 = w+ < w − (1/2)u − (1/2)v > . Now suppose the characteristic of F is 3. β = (−1/2)δ. m2 . Let {u. Note that z = u + v + w = −2u + v + w = (v − u) + (w − u). γ) are the unique aﬃne coordinates determined by our triangle with vertices in the order given. We leave the rest of the proof to the reader. β = 1 + ξ. we can rewrite m1 = u+ < z >. α is unique. Proof. We claim they are parallel. It follows that the medians are distinct and parallel. Since 1+δ−(1/2)δ−(1/2)δ = 1. γ = (−1/2)ξ. β. v. Theorem 36. z belongs to m2 if and only if there is ξ ∈ F with α = (−1/2)ξ. Note that u − v = 2(v − u) + 0(w − u). γ = (−1/2)δ. The aﬃne coordinates of the unique point of intersection of m1 and m2 is then (1/3. m3 . impossible as characteristic of F is 3. m2 = v+ < z >. w} be a triangle in Π(F ). m3 = w+ < z >. Then the medians of a triangle in Π(F ) are concurrent. say F = Z3 . Referring to the exercise above. implying that α = 2 and α = 0. Let F be a ﬁeld with characteristic not 2 or 3. it follows by uniqueness of aﬃne coordinates that α = 1 + δ.
2. z are collinear if and only if ⎞ ⎛ α1 β1 γ1 det ⎝ α2 β2 γ2 ⎠ = 0. The best way to understand this is to write out all the permutations on a set of three objects {a. 0). Show that x. γ3 ). 1)} ⊂ Π(Z5 ).1 The Aﬃne Group Deﬁnition 18. In this case we often write GL(V ) = GLn (F ). Example 3. w} be a triangle in Π(F ). We then have 67 . y. one of the big theorems you learn in a beginning course in abstract algebra. What is the point of concurrency of its medians? 3. e(s) = s for all s ∈ S. 0). and if every element in G has an inverse in G then G is a group. v. y. is a group designated GL(V ). This is the content of Cayley’s Theorem. V . Suppose e ∈ G is the identity mapping. In some sense. Let G be a set of mappings on a set S. α3 β3 γ3 3. γ1). b.Exercises 1. 1/3. The set of invertible linear transformations on any vector space. β1 . Example 4. β3 . γ2 ). Complete the proof of the last theorem by verifying that the point with aﬃne coordinates (1/3. (α3 . that is. z be points in Π(F ) with aﬃne coordinates given respectively by (α1 . 1/3) lies on m3 . c} and see what happens if you start composing them. β2 . When dimF V = n. all groups are groups of permutations. Let {u. If G is closed under function composition. In what sense is a vector space a collection of mappings on a set? For any v ∈ V . for all u ∈ V . deﬁne tv : V → V by tv (u) = v + u. (α2 . The most fundamental example of a group is the set of permutations on any set. A vector space V is a group which we also designate V . GL(V ) can be identiﬁed with the set of nonsingular n × n matrices with entries in F . Let x. Consider the triangle {(0. (0. (1. Groups are a critical element in the modern study of geometry. This is easy to check and the reader is urged to do so.
v) : V → V by A(T. z} is a triangle in Π(F ) if and only if {y −x. For any T ∈ GL(V ). y . while function composition is generally not commutative. Notice that unlike GL(V ). Lemma 9. The inverse of tv is t−v . then for any aﬃne transformation A(T. T ∈ GL(V ). tv ◦ tu = tv+u = tu+v = tu ◦ tv . Deﬁne A(T. The aﬃne group is an example of what is called a semidirect product of GL(V ) and V . y. y. The aﬃne group maps lines to lines in Π(F ). T (y) + v. Let u + αw ∈ = u+ < w >. V is a commutative group. where θ is the zero vector. where and are lines. y. There are actually many diﬀerent ways to deﬁne “products” of groups. Deﬁnition 19. v). v)( ) ∩ A(T. We have A(T. v)(u+αw) = T (u+αw)+v = T (u)+αT (w)+v ∈ v +T (u)+ < T (w) >. Lemma 10. v)({x. an element of Aﬀ(V ) can be identiﬁed with an ordered pair (T. There is an aﬃne transformation A(T. The aﬃne group maps triangles to triangles in Π(F ). v)( ). v)(u) = T (u) + v. T (z − x)} = {T (y) − T (x). Fix T ∈ GL(V ) and v ∈ V . and since T ∈ GL(V ). If {x. The aﬃne group Aﬀ(V ) is the set {A(T. As a set.tv ◦tw (u) = tv (tw (u)) = tv (w +u) = v +w +u = tv+w (u) so that tv ◦tw = tv+w . T (z) + v − (T (x) + is also a basis for V . y . in other words. It is clear that if x ∈ ∩ . Lemma 8 implies that A(T. T (w) cannot be zero if w = θ. Let A(T. thus the image of the line under A(T. z −x} is a basis for V . T (z) − T (x)} = {T (y) + v − (T (x) + v). {x. 68 . v) ∈ Aﬀ(V ). v){x. v)( ) = . z } are triangles. v ∈ V . there is a unique A(T. A(T. v) ∈ Aﬀ(V ) with A(T. z }. v){x. By Lemma 8. Any group in which the operation is commutative like this is called an abelian group. y. Let V = F 2 . Aﬀ(V ) is the Cartesian product of GL(V ) and V . T (z) + v}. The identity mapping is tθ . for all u ∈ V . Let and be lines in Π(F ). {T (y − x). v). v) is a line v + T ( ). z} is a triangle. v ∈ V . v) such that A(T. that is. v)(x) ∈ A(T. z} and {x . v ∈ V }. Proof. v) T ∈ GL(V ). z}) = {x . y. we have A(T. Theorem 37. z} = {T (x) + v. Proof. Since T is an invertible linear transformation.
geometry was in a state of disarray. there is a unique element T ∈ GL(V ) that maps {y − x. v. This is typical of the Erlangen Program of Felix Klein. Thus. w. v)(x) = T (x) + v = x . 2. −T (x) + x ){x. At that point in history. Then A(T. The notion that groups could and should serve as a unifying force in geometry turned out to be profound. Let H be the subgroup of Aﬀ(V ) that leaves a given triangle invariant. The lines u+ < w − u > and v+ < x − v > are the diagonals. The collection of H cosets is {g ◦ hg ∈ Aﬀ(V ). Let = x+ < y > and = x + < y > be distinct lines in Π(F ). H is a subgroup of G. The four points are the vertices of the quadrilateral. Going back to our deﬁnition of “group. where the characteristic of F is not two. z − x }. then v is uniquely determined. 69 . v. Show that the diagonals of the parallelogram bisect one another. v)(x+ < y >) = T (x) + v+ < T (y) >= x + < y >. v. x} is a parallelogram if and only if one of its vertices is the sum of the other two minus the third. y. w. x} be a parallelogram in Π(F ). Let {u. that maps the triangle to itself with a permutation of its vertices. The result follows. a hugely inﬂuential idea described in a lecture of Klein’s dating to 1872. and if S ⊂ S. T (y) − T (x) + x . T (z) − T (x) + x } = {x .Proof. z }. It remains at the heart of a great deal of mathematics being discovered today. and the lines they determine. y − x + x . Consider that A(T. no three of which are collinear. Exercises Deﬁne a complete quadrilateral in Π(F ) to be a set of four points.” we see it is easy to verify that if G is a group of mappings on a set S. y. Given triangles {x. y . Elements of H are said to leave S invariant. Let v = x − T (x). These can actually serve as triangles. that is. Choose T ∈ GL(V ) so T (y) = y . 1. Prove that a quadrilateral {u. z} = {x . h ∈ H}. w. then the set H ⊂ G of elements in G that map S → S forms a group inside G. z − x + x If A(T. x} is a parallelogram provided its vertices can be relabeled if necessary so that u+ < v − u > x+ < w − x > and u+ < x − u > v+ < w − v >. z − x} to {y − x . A quadrilateral {u. z} and {x .
We deﬁne division in C by z/w = z w/ w 2. 2)}. therefore the dimension 70 . By the end of the section. b ∈ R}. The quaternions as rotations turn out to be an instance of a group acting on the objects in a geometry. z2 ) = (z1 . S 1 is the one sphere.) Recall further that the complex conjugate of z = a + bi ∈ C is z = a − bi. S 1 is not ﬂat. we should see that it is not as much of a digression as it ﬁrst appears. C = {a+bi  a. 4 The Quaternions This section represents something of a digression. 0). which is one dimensional. usually designated S 1 . Multiplication is commutative (easy to check) and associative (tedious to check. a ¯ z nonnegative real√ number. This gives us the modulus or norm or absolute value of z ∈ C.3.) Prove that ϕ sends lines to lines and triangles to triangles but that it is not an aﬃne transformation. z = a2 + b2 . note ﬁrst that we can multiply elements of C simply using the distributive law and the deﬁnition of i via i2 = −1. Notice that z is the distance from the associated point in R2 to the origin. Hamilton discovered the quaternions after an arduous search inspired by the beautiful connection he discovered between the complex numbers and the geometry of R2 . the aﬃne plane over C. the sphere which is one dimensional. A safe way to think about it is as follows: the dimension of the object is the dimension of its tangent space. In the next section. z2 ). This is signiﬁcant because it allows C and R2 to borrow structure from one another. we will take up the same topic in the real projective plane. Our objective is to describe how the quaternions deﬁne rotations in R3 . if this is ¯ unfamiliar. The tangent space to a circle is a line. This ¯ makes sense whenever w = 0 and now we have that C is a ﬁeld. every parallelogram is the image under some aﬃne transformation of {(0. (See below for more details. Let ϕ(z1 . Consider the collection of unit complex numbers {z ∈ C  z = 1}. (2. where ¯ ¯ z is the complex conjugate. b)  a. that is. Note that this is precisely the unit circle in R2 . 4. (0. so you don’t have a precise idea of what “dimension” means for such an object. b ∈ R} can be identiﬁed with the points in R2 = {(a. 2). 0). Consider Π(C). Recall that the complex numbers. it is not a vector space. (2. We then have z¯ = a2 + b2 . Prove that in an aﬃne plane over a ﬁeld of characteristic not two. To see this.
Deﬁne H = {q = a + bi + cj + dk  a. a ∈ R. Using these relations and the distributive law. au + bv + v × u). If q = (a. b. it is an easy exercise in trigonometry to show that if w ∈ C. v ∈ R3 . we can deﬁne multiplication of quaternions. Then ¯ √ 2 2 2 2 ¯ q q = a + b + c + d and the modulus of q is q = q q . The pure quaternions are those of the form (0. then z = cos θ + i sin θ for some θ. If z ∈ S 1 . ik = −j. the quaternions H. b. kj = −i. Making the identiﬁcation. for q. In fact. jk = i. d). It does. v). this allows ¯ us to deﬁne division: q/w = q w/ w 2 . j. Hamilton wanted to extend this story to R3 and came up empty handed until he realized.w = w (cos θ cos ϕ − sin θ sin ϕ) + i(sin θ cos ϕ + cos θ sin ϕ) = w (cos(θ + ϕ) + i sin(θ + ϕ)). c. however. w ∈ H. c. u) are in H. we have conjugation in H given by q = a − bi − cj − dk. ki = j. Again. d ∈ R} where i. ¯ H can be identiﬁed with R4 .of S 1 is one. v). ji = −k. With z and w in hand. then using the distributive law you should have no trouble verifying that qw = (ab − v · u. we have z. and w = (b. k enjoy no relations among themselves except for the following and their consequences: i2 = j 2 = k 2 = −1 ij = k. Notice that we can designate any quaternion by (a. we write q = a + bi + cj + dk = (a. c. that he needed to view R3 as sitting inside a four dimensional analog to C. b. We see then that multiplication by a unit complex number eﬀects a rotation on R2 : it does not change the distance from a point to the origin but it does change its angle. As in C. at a famous moment when he was crossing a bridge in Dublin. enjoy all the other properties of a ﬁeld so is a skew ﬁeld. This is nice because it lets us write a rather involved looking formulation that actually helps us prove some results. w = w (cos ϕ + i sin ϕ) . Obviously multiplication is not commutative so H is not a ﬁeld. 71 . d).
where i. an object that has inspired extensive research for over one hundred years. sin(θ/2). −v) = (−av · u + av · u. we have v · v = 1. v). sin(θ/2). viewed as a pure quaternion. v2 . v × v = 0. u)(a.v) = 72 . v3 ) and u = (u1 . that is. Notice that there are two choices for v. Then qvq −1 = (cos(θ/2). the surface of a ball. ⎡ ⎤ i j k v × u = det ⎣ v1 v2 v3 ⎦ u1 u2 u3 We can identify R3 with the pure quaternions and then the unit pure quaternions become S 2 . through the angle θ. Bearing in mind that v is a unit vector. If it makes things any easier we can always insist that if v points “up”. Theorem 38. v × u is likewise the usual cross product in R3 . the 3sphere. k now play the roles of the standard unit vectors in R3 . for v = (v1 .v)(0.v be the rotation of R3 about the axis determined by the unit vector v. v)(0.v) really is a rotation of R3 with axis < v > is to see what qvq −1 gives us. the 2sphere. . . is. Also. au + v × u)(a. Where are the rotations? Let Rθ. Rθ. . that is. The mapping u → quq −1 is called conjugation by q. −v) = (−v · u. u2 . An obvious check to whether or not q = (cos(θ/2). by (cos(θ/2). then the rotation is counterclockwise. sin(θ/2). u) and q = (a.v). To prove/understand the theorem. ) where it doesn’t matter what . For u = (0. the zero vector. v · u = v1 u1 + v2 u2 + v3 u3 . j. The unit quaternions can be identiﬁed with S 3 . . − sin(θ/2). ) = (0. since we just wanted to show that u is mapped to another pure quaternion.v is eﬀected by conjugating any element of R3 . v)(cos(θ/2). u3). . . . quq −1 = (a.Here v · u is the usual dot product in R3 . the ﬁrst thing we do is verify that conjugation maps pure quaternions to pure quaternions. . Conjugation comes up frequently in group theory and in geometry.
and that v × u = w. in other words.v) = (− sin(θ/2) cos(θ/2) + sin(θ/2) cos(θ/2).u × v − sin2 (θ/2)w × v) = 0. u → cos θ. but not a proof.v + sin2 (θ/2)v) = (0.w)(cos(θ/2).w).u + sin θ.w) = (0. u.v) = (0. that they are orthogonal.w. and w × v = u. The matrix representation is then ⎡ ⎤ 1 0 0 ⎣ 0 cos θ − sin θ ⎦ 0 sin θ cos θ To see this. think of where the unit vectors in the y and z direction are mapped under the rotation.u + 2 sin(θ/2) cos(θ/2).(− sin(θ/2). w as arranged like the unit vectors in the x. That completes the proof of the theorem.v)(cos(θ/2). we can think of v.w.v)(0.v . u. We leave it as an exercise to show that conjugation by q maps w → − sin θ. have zero dot product.v) = (0. cos2 (θ/2).u + cos θ.u + cos θ. In short. u)(cos(θ/2). A good basis to use here would be {v. cos(θ/2). u × w = v. 73 .u + sin(θ/2).v as a counterclockwise rotation in the yz plane. cos2 (θ/2)u + cos(θ/2) sin(θ/2)w − sin(θ/2) cos(θ/2). v). sin(θ/2). In particular. u. y. we think about a matrix representation of Rθ. w} form a righthanded coordinate system of orthonormal vectors. z directions in R3 . we have v → v. This means that all three basis elements are unit vectors. cos(θ/2). To ﬁnish the proof.u + sin θ. This proves that v is ﬁxed under conjugation by q. This allows us to visualize Rθ. This is promising.w. − sin(θ/2). w} where {v. − sin(θ/2). cos θ. (cos2 (θ/2) − sin2 (θ/2). Now we check: quq −1 = (cos(θ/2). w → − sin θ. Matrix representations depend on choice of ordered basis. − sin(θ/2).
and < (1. 0)}. 1. < (0. Take the rotation that is counterclockwise when viewed along the axis from the ﬁrst octant. β not both zero in F . 1) >. 1. 0). 0 and 1. 1. 0) >. Use q to determine where the point (−1. Theorem 39. 74 . 0. 0). Π(V ). a projective line as a 2dimensional subspace of V . 0. 0) > and < (0. that satisﬁes the axioms for a projective plane. < (0. 0. incident to the same line. This veriﬁes that Π(F ) has four points. Let F be any ﬁeld and consider the vector space V = F 3 . we get a set. about an axis that is normal to the plane x + y + z = 0. 1). We have four axioms to prove: the two that show nontriviality and the two incidence axioms. A similar argument applies to the other sets of three vectors. 0) >. no three of which are collinear. 1. neither of the latter points is collinear with < (1. RP 2 . 1. when does qwq −1 have the same norm as w. In the proof of Theorem 38. that is. 1) nor (1. Give exact coordinates. w pure quaternion? 4. What conditions would we have to impose on q to get that conjugation by q on pure quaternions preserves norm? In other words. 2. Since F contains at least two elements.Exercises 1. and incidence as nonempty intersection. 0. the set of all vectors in V of the form (α. β. why did we not have to show that every vector in R3 underwent a rotation? 3. for q ∈ H. 1. (0. that is. we have the following distinct points in Π(V ): < (1. 0) >. We claim that no three of these are collinear. 5 The Real Projective Plane We consider here one model for the real projective plane. 0) > is the span of {(1. q = 0. α. 0. Proof. 0. Find the explicit expression for q to eﬀect the rotation in R3 through 30◦ . By deﬁning a projective point as a 1dimensional subspace of V . 1. deﬁnes a linear transformation on R3 . Since this set contains neither (0. 2) is mapped to under the rotation. 0) > and < (0. Show that conjugation by q ∈ H. The line determined by < (1. 1) >.
One that can be useful for visualization is the unit sphere. In other words. z). in turn. This is a bijective correspondence between points in the two models. In particular. so are called homogeneous coordinates. a one dimensional subspace of R3 intersects S 2 in a unique pair of antipodal points. with antipodal points identiﬁed. A way to think of this is to realize that you get a circle on the sphere precisely by intersecting S 2 with a plane. which must then contain the antipodes of both original points. < v > and < u + v >. When you intersect S 2 with a plane through the origin. let be determined by the span of {u. notice that for a pair of nonantipodal points on the sphere. thus spanning a 2dimensional subspace of F 3 . a point of RP 2 is realized as the pair of endpoints of a diagonal of the unit sphere in R3 . Note that the span of this set contains lines < u >. that is. y. 5. v} ⊂ F 3 . and intersect in a unique projective point. thus. determines a unique great circle on S 2 . if < u > and < v > determine distinct points in Π(F ). this is a set of three noncollinear points. they correspond to three distinct points on the line determined by u and v. This brief discussion shows right away how the S 2 model and the subspace model interact. then {u.Next we consider a line in Π(F ) and verify that it has three points. you get a great circle. Next. In particular. z] = [x/z. then [u] = [x. This great circle identiﬁes the projective line determined by the two projective points. 1]. Every line in Π(F ) is determined by a linearly independent set of vectors {u. y. S 2 . y. Similarly. the coordinates of [u] are deﬁned only up to nonzero multiples. which determines the unique line in Π(F ) determined by < u > and < v >. v} and be determined by the span of {w. 75 . Since planes through the origin in F 3 must intersect in a unique line through the origin. z] be the projective point corresponding to the nonzero vector u = (x. If you pick two nonantipodal points on S 2 and the origin. It is easy to verify that these are distinct. there are many circles on the surface of the sphere that contain those points but there is only one great circle that contains those points. x}. How is a line realized with this model? To understand this part.1 Homogeneous Coordinates Designate a projective point now as [u]. This plane. If z = 0. v} is a linearly independent set. thus uniquely determines a plane through the origin. let [x. y/z. There are other models for RP 2 .
2. 1] (b) [−2. 2] and [2. 1. 3 − 1] (b) [0. 1. The mapping ϕ : Π(F ) → Π(F ) given by ϕ(a. 1] and [−1. 1. ﬁnd the preimage in Π(R). We leave the veriﬁcation of the next theorem to the reader. b/c) under ϕ. 1]. 0] and [2. < u >=< (a. b. Theorem 40. for each of the following projective points. The ﬁrst six exercises that follow are taken from [7]. 1] (c) [2. The plane z = 1 intersects all lines through the origin in F 3 except those lying in the xyplane. that is a localization. is the projective line corresponding to the space {(a. 1] and [−1. 2. 3. Deﬁnition 20. Now we can think of projective points [a. 3. 0) a. The line at inﬁnity. Our choice of projective line to designate ∞ is arbitrary: we can excise any line from Π(F ) to get a copy of Π(F ).Homogeneous coordinates are especially convenient for localizing. 0] and [2. −3. In other words. c) > is a line in F 3 not lying in the xyplane if and only if c = 0. (a) [1. 0. b. 1] 76 . This gives us a visual aid for understanding Π(F ) as a subset of Π(F ). Describe the line in Π(R). for thinking of points that are somehow “near” one another in RP 2 as belonging to an aﬃne space. b. 1] is a bijection onto the set Π(F ) \ ∞ . b) = [a. if there is one. 2. (a) [2. Its intersection with the plane at z = 1 is the image of the aﬃne point (a/c. 0] 2. 3. 1] as being near each other in the sense that they comprise an aﬃne space. (d) [1. ∞ ⊂ Π(F ). b ∈ F }. 4] (e) [1. Exercises 1. determined by the following pairs of points. b. Using the mapping described in the last theorem. that is.
−x + 7y − 6z = 0 (b) x + 2y − z = 0. Find the projective point where each pair intersects. 2. −3. x − (1/2)y + z = 0 77 . (a) x = 7 (b) y = 8 (c) x + y = 4 (d) −x + 3y + 18 = 0 (e) 2x − 7y = 21 (f) −y + 3x + 5 = 0 5. 4. Find the image of each of the following lines in Π(R) under the mapping ϕ given in the theorem above. (a) x + y + z = 0. −1] (f) [−1. 0. −1] (d) [2. 1/3. −3. −4x + 2y + z = 0. z = 0 (c) 2x − y + 3z = 0. Determine whether each of the following sets three projective lines is concurrent and ﬁnd the points of concurrency. 4x − y = 0. and −x + 2y + z = 0 (b) 2x − y + 3z = 0 and −x + z = 0 (c) x = 0 and y = 0 (d) x = 0 and z = 0 (e) y = 0 and z = 0 (f) −x + 5y − z = 0 and 2x + 7y − 3z = 0 4. 2x − 3y + z = 0. The following pairs of equations describe lines in RP 2 .(c) [−2. 0] (g) [1/2. 6] (h) [1. 6] (e) [2. (a) x + y + z = 0. 0] 3.
3. 3]. in which case ¯ ¯ T [u] = [T (u)] = [αT (u)] = [T (αu)] = [T (v)] = T [v]. I3 is the identity mapping on ¯ ◦ S = T ◦ S. This means that on Π(F ). [w]} ⊂ Π(F ) is a triangle if and only if {u. 0]. 1. 0. GL3 (F ) seems a natural choice for a group of bijections on Π(F ) but we see trouble right away when we notice. GL3 (F ) can be identiﬁed with 3 × 3 matrices over F with nonzero determinant. ¯ The statement about I3 is clear. for instance. [2. Deﬁnition 21. [2. 1/2. (a) [−1. ¯ ¯ Π(F ). −2. 1. w} is a basis of F 3 . then T [u] = T [v] but that is immediate because [u] = [v] if and only if there is nonzero α ∈ F with v = αu. P GL3 (F ). [0. 1]. and T −1 = T −1 . for any α ∈ F . 1].2 The Projective Group Recall that GL3 (F ) is the group of nonsingular linear transformations from F 3 to itself. −2]. 3. −1. . is the collection of map¯ ¯ pings T : Π(F ) → Π(F ) where T ([u]) := [T (u)] for T ∈ GL3 (F ).6. Let T and S belong to GL3 (F ). We put the cart before the horse with this deﬁnition: it really only makes any sense after we prove the following theorem. αI3 is the identity mapping on Π(F ). Which of the following sets of projective points is collinear? For those that are. [3. The projective group. −1]. [1. This suggests that we consider the following as the natural group of bijections on Π(F ). I3 and −I3 deﬁne the same mapping. then T deﬁned by T [u] = [T (u)] is a well¯ deﬁned bijective mapping on Π(F ). v. that both I3 and −I3 ﬁx all projective points. describe the line. we have ¯ ¯ ¯ T ◦ S[u] = T [S(u)] = [T (S(u))] = [T ◦ S(u)] = T ◦ S[u]. 78 . 0] (c) [1. For any [u] ∈ Π(F ). 5. [v]. [1/2. Indeed. 4] (b) [−2. Prove that {[u]. At ﬁrst blush. 2] 7. 0. T ¯ ¯ Proof. ¯ ¯ Theorem 41. If T ∈ GL3 (F ). Moreover. We ﬁrst must show that if [u] = [v].
T ({u. we suppose P is a projective transformation that maps Π(F ) to itself.¯ We leave the proof of the last statement to the reader. that is. α nonzero in F . v ∈ V and A(T. v). v >⊂ ¯ F 3 and a mapping T ∈ P GL3 (F ) corresponds to the collection of mappings αT ∈ GL3 (F ). Recall that an element of Aﬀ(V ) is A(T. The projective groups maps lines to lines and triangles to triangles in Π(F ). y) ∈ F 2 . = TT 00 T (v ) + v 1 . Using block matrix T v 00 1 T v 00 1 A(T. we identify Π(F ) with the plane z = 1 ¯ in F 3 . v ) = T v 00 1 . Theorem 42. It is not diﬃcult to verify thatP must be all nonzero scalar multiples of a matrix of the form ⎞ ⎛ T a b ⎠ P =⎝ 0 0 1 79 . in light of Exercise 7 at the end of the last subsection. The argument for triangles is even easier. v}) is a linearly independent set of two vectors hence corresponds to a two dimensional subspace < T (u). A line ∈ Π(F ) corresponds to a two dimensional subspace < u. ¯ Turning now to P GL3 (F ). v) = We then have A(T. T (v) >=< αT (u). Proof. we can write where T = ∈ GL2 (F ) and v = (x. to a projective line ∈ Π(F ). The fact that T is bijective follows. As usual. v)(u) = T (u) + v. v) = ⎝ c d y ⎠ 0 0 1 a b c d form. T bijective. Next we consider the relationship between the aﬃne and the projective groups. Identifying Π(F ) with the plane z = 1 in F 3 lets us think of the elements of Aﬀ(V ) as matrices with the form ⎞ ⎛ a b x A(T. where T ∈ GL(V ). v) ◦ A(T . αT (v) >.
If not. Show that if we consider a triangle an ordered triple of noncollinear points. 0). Theorem 43. 1) and T (0. by an aﬃne transformation. ¯ (b) Find T [1. 0. Aﬀ(F 2 ) can be identiﬁed as a subgroup of P GL3 (F ) that maps ∞ to itself and that ﬁxes Π(F ) ⊂ Π(F ). 0) = (0. Verify thatP ∈ P GL3 (F ) maps Π(F ) to itself if and only if P corresponds to the set of all nonzero scalar multiples of a matrix of the form ⎞ ⎛ T a b ⎠ P =⎝ 0 0 1 for some T ∈ GL2 (F ). Exercises ¯ 1. ¯ ¯ (d) Does T : Π(F ) → Π(F )? If so. 1) = (1. ¯ ¯ 4. Finish the proof of Theorem 41 by showing that T −1 = T −1 . (a) Verify that T ∈ GL3 (F ). 2. then there is exactly one element in the projective group that sends a given triangle to another.¯ where det P = det T = 0. prove it. we can represent P uniquely with a matrix of this form. Verify that a mapping of the form given in the last problem ﬁxes ∞. 80 . 1. T (0. 3. 0. describe T (Π(F )). 1. that is. In other words. 2. 5. 0) = (0. Let T linear be deﬁned on F 3 by T (1. 0. −1]. ¯ (c) Find T ( ∞ ). 0. Complete the proof of Theorem 42 by showing that P GL3 (F ) maps triangles to triangles. 1).
the pieces that can be described √ as graphs of functions. Instead of thinking of functions. for example. Consider the equation 2x + 3y = 1. (c) Homogenize the equation. an algebraic curve is the set of zeroes of a homogeneous polynomial. in particular. 81 . x − y 2 is an example of an inhomogeneous polynomial of degree 2. We homogenize a polynomial equation in x and y by ﬁrst writing it in the form p(x. we deﬁne a real algebraic plane curve. y = x. you study pieces of curves. If we homogenize x − y 2 . Typically. y = f (x).5. in that context. y) = 0. we get xz − y 2 . that is. To homogenize y = x2 − 2. Example 5. A homogeneous polynomial of degree n is a sum of monomials of degree n. The equation associated to the function y = x would be y 2 = x or x − y 2 = 0. More generally. (c) What does the homogenized equation describe in RP 2 ? 2. multiply each monomial by z d−n . then homogenizing p(x. we are thinking √ of polynomial equations.3 Curves Curves are a source of much comment in calculus. We can understand curves in F 2 by “lifting” to Π(F ) through the process of homogenization. Note that the degree of the polynomial is that of the highest degree monomial. Consider the set of points in R2 given by (t. as the zero set of a polynomial in x and y. (a) What does this set of points describe? (b) Write an equation that gives a relation satisﬁed by the x and y coordinates of the points on this curve. (a) What does this equation describe in Π(R)? (b) Homogenize the equation. To homogenize a polynomial in x and y of degree n. a curve in R2 . y). In Π(F ). rewrite to get x2 − y − 2 = 0 then homogenize the left hand side to get x2 − yz − 2z 2 = 0. t2 /2). Exercises 1. where d is the degree of the monomial.
(d) Describe the set of points in Π(R) that satisfy the homogeneous equation using parameters s and t. (Hint: Think of embedding the points (t, t2 /2) into R3 by putting them in the plane z = 1. Now homogenize the resulting parametric description.)
5.4
Conics
Recall that in R2 , we can deﬁne a conic section to be the set of zeroes of any degree two polynomial. ax2 + by 2 + cxy + dx + ey + f. When the constant term, f , is not zero, and at least one of a, b, c is not zero, the conic is nondegenerate, that is, it is a bona ﬁde parabola, hyperbola or ellipse, as opposed to one or two lines or a single point. If we homogenize our deﬁning polynomial, we get ax2 + by 2 + cxy + dxz + eyz + f z 2 It may be hard to see what this deﬁnes. Note that we can get rid of cross terms by rotations, though. To help you believe that, we note that if we send x → x + y and y → x − y then xy → x2 − y 2. Now look at the xyplane, and compare the x axis to the line x + y, and the y axis to the line x − y: we rotate through an angle of 45◦ to eﬀect the transformation. This lets us see that we lose no generality if we consider that the deﬁning polynomial for a conic in RP 2 is given by ax2 + by 2 + f z 2 . Next note that scaling in the x, y, z directions separately is deﬁned by a diagonal matrix which determines a projective transformation. To understand how to think of a conic in RP 2 , all we care about are the zeroes of the polynomial. It follows that the only thing that matters here is the ratio of signs of the coeﬃcients of x2 , y 2, and z 2 . This means that we can think of a conic section in the projective plane as associated to the surface given by x2 + y 2 − z 2 = 0, that is, a cone. Diﬀerent conic sections then arise via a transformation of the cone that amounts to a rotation about the origin in R3 or scaling in various directions, neither of which change the cone into a diﬀerent type of surface. This makes the following theorem transparent. Theorem 44. Under projective transformations, all (nondegenrate) conic sections are the same. In other words, given any two conic sectionsin RP 2 , there is a projective transformation that sends one to the other. 82
5.5
Degree
¯ Our goal here is to show that the degree of a curve in Π(F ) is actually a geometric feature of the curve, not just an attribute of one polynomial that may deﬁne the curve. We need a few results about homogeneous polynomials to get there. This material draws from Walker’s classic book on algebraic curves [10]. Theorem 45. The factors of a homogeneous polynomial are homogeneous. Proof. Suppose F = f g, where F is homogeneous and f is not homogeneous. Write f and g using homogeneous components as follows: f = Fi + . . . + Fi+j g = Gk + . . . + Gk+ where Fi , Fi+j are nonzero,homogeneous with degrees respectively i and i+j, and Gk , Gk+ are nonzero homogeneous of degrees k and k + . Attempting to write f g using homogeneous components we get f g = Fi Gk + (Fi Gk+1 + Fi+1 Gk ) + . . . + Fi+j Gk+ where Fi Gk and Fi+j Gk+ are both nonzero with degFi Gk = i+k < degFi+j Gk+ , contradicting F homogeneous. Let Fn designate a homogeneous polynomial of degree n. Fn can have any number of variables. Say it has k variables, that is, Fn = Fn (x1 , . . . , xk ). Let f be the nonhomogeneous polynomial in k − 1 variables that we get by letting xi = 1, for some i ∈ {1, . . . , k}. We say Fn and f are associates. The next three results, variations on a theme, follow immediately from the theorem. Corollary 5. The factors of Fn are associates of the factors of f where Fn and f are associates. Corollary 6. F is irreducible if and only if its associates are irreducible. A ﬁeld is algebraically closed provided all polynomials in a single variable with positive degree factor into linear factors. Our favorite example of an algebraically closed ﬁeld is C. Note that R is not algebraically closed as, for example, x2 + 1 does not factor in linear factors if we restrict to coeﬃcients in R. 83
Corollary 7. Let D be algebraically closed and Fn homogeneous degree n in D[x0 , x1 ]. There are a = 0, ai , bi in D with F = aΠn (ai x1 − bi x0 ). In other i=1 words, F factors into linear factors. ¯ Now consider an algebraic curve in Π(F ). This is the set of projective points that satisfy Fn (x, y, z) = 0, where Fn (x, y, z) is homogeneous of degree ¯ n. Let be any line in Π(F ). Think of as determined by a two dimensional subspace spanned by vectors u and v in F 3 . We can think of as the set of all nonzero points su+tv, where s, t ∈ F . In other words, [s, t] are homogeneous coordinates for a point on . Consider what it means for a point to lie on both C and . Because it lies on , it is su + tv, for some (s, t) ∈ F 2 \ (0, 0). Because it lies on C, Fn (su + tv) = 0. A point of intersection of C and then corresponds precisely to a solution [s, t] to Fn (su + tv) = 0. We are now thinking of s and t as variables so that Fn (su + tv) is homogeneous degree n in two variables. Our corollary above allows us to say that, counting multiplicity, there are n ¯ solutions to Fn (su + tv) = 0 in Π(F ), as long as F is algebraically closed. This proves the following theorem. Theorem 46. Assume F is algebraically closed. The degree of a curve in ¯ ¯ Π(F ) is the number of intersections it has with a line in Π(F ) The deﬁnition of degree as a geometric property of a curve is the ﬁrst subplot in the story of intersections and singular points of curves. An important tool in the telling is resultants. We end these notes with a brief discussion of this idea. Deﬁnition 22. Let f (x) = a0 + a1 x + a2 x2 + . . . + an xn , g(x) = b0 + b1 x + b2 x2 + . . . + bm xm , ai , bi ∈ F , an , bm = 0. The resultant, R(f, g) is the determinant a0 a1 . . . 0 a0 a1 . . . b0 b1 . . . bm−1 bm . . . 0 b0 . . . bm where there are m rows of ai s and n rows of bi s. 84 an . . . an−1 an
Note that the resultant is the determinant of an m + n × m + n matrix. Example 6. Let f (x) = x2 − 2x + 1 and let g(x) = x − 1. Then R(f, g) = 1 −2 1 −1 1 0 0 −1 1 = 0.
The resultant of a polynomial and its ﬁrst derivative is called the discriminant of the polynomial. We are used to seeing the discriminant deﬁned for a quadratic polynomial f (x) = ax2 + bx + c as b2 − 4ac. If you think about what the discriminant gives us, you start to get a sense of what resultants are good for more generally. In particular, when the discriminant of a quadratic polynomial is zero, the polynomial has a repeated root, that is, a root in common with its derivative. A quick calculation reveals that the deﬁnition we have here deﬁnes the discriminant of f (x) = ax2 + bx + c as −a(b2 − 4ac). Obviously, this is zero if and only if b2 − 4ac = 0. As is typically the case with determinants, we only care whether the resultant is zero or not zero so for what we need, the two deﬁnitions are close enough. Theorem 47. R(f, g) = 0 if and only if f and g have a common nonconstant factor. Proof. We sketch the proof. Start by noting that f and g have a nonconstant common factor if and only if there are ϕ(x) and ψ(x) with degree ϕ(x) < degree g(x), degree ψ(x) < degree f (x) and f ϕ = gψ. Now think of ϕ(x) = α0 + α1 x + . . . + αk xk , ψ(x) = β0 + β1 x + . . . + β x , where αi , βi are unknowns. After doing the multiplications to write f ϕ = gψ, we see that we have equations giving us a0 α0 = b0 β0 , a0 α1 + a1 α0 = b0 β1 + b1 β0 , etc. Essentially, the resultant is the determinant of the (transpose of the) coeﬃcient matrix for this large homogeneous system of equations. The system has a nontrivial solution— αi s, βi s not all zero— precisely when the determinant of this matrix, the resultant, is zero. In the projective plane, an algebraic curve is deﬁned by a homogeneous polynomial in three variables. Two such polynomials with a nonconstant common factor deﬁne curves with a common component. How can we use resultants to pick up on these common components? 85
We can think of f as belonging to D[xk ]. 4. . . Theorem 48. . then R(Fn . an−1 an . . 5. xk ) is then a polynomial in k − 1 variables. This idea allows us to deﬁne resultants for polynomials in several variables. . . z]. xk−1 ]. + an λn .Let f (x1 . y) = x − y is? Now do the calculation. Gm ) = 0 if and only if Fn and Gm have a common nonconstant factor.. . xk ) and g(x1 . A resultant for f (x1 . Let f (x. m in F [x0 . the coeﬃcient of x is −2y and the coeﬃcient of x2 is −1. D = R[y. Exercises 1. z) = x2 + y 2 − z 2 and the line x + y + z = 0. D is an example of an algebra. Example 7. Find all of them. . (Note that there are three possibly diﬀerent resultants. . 2.. ﬁnd the discriminant of x3 + px + q. where D = F [x1 . .) 3. Then we have f = z 2 − (2y)x − x2 . that is. Find the resultant of the two polynomials from the last exercise. Using the deﬁnition of discriminant given in the text. How can you tell. Consider the determinant a0 a1 a2 −λ 1 0 0 −λ 1 ··· ··· ··· 0 0 0 . without doing the calculation.. . Gm are homogeneous of degrees n. . . . z) = 2xy − x2 + z 2 . y.. ··· ··· ··· . so the “constant” term is z 2 . x1 ]. Find the intesection points of f (x. a vector space that is closed under multiplication. −λ 1 Argue that this must be equal to a0 + a1 λ + a2 λ2 + . . 0 0 . 86 . what the resultant of f (x. . 0 0 . If Fn . . . y) = x2 − 2xy + y 2 and g(x. y. We can think of f as belonging to D[x]. . xk ) be a polynomial with coeﬃcients in a ﬁeld F . ...
T. H. W. [2] Euclid. “Algebraic Curves.References [1] Coxeter. Dover Publications. Inc. 1 (Books I and II). [9] Stanford Encyclopedia of Philosophy.” Second Edition.” Second Edition. John Wiley and Sons. 1989. Scotland [7] Malkevich. Joseph.mcs. John Wiley & Sons. “The Thirteen Books of The Elements..” Marcel Dekker. 1999. 87 .edu/ malk/mycourses/math244/realprojectiveplane. Open Court.edu/ [10] Walker. Robert J.” SpringerVerlag New York. New York. [5] Loustau. “Projective and Euclidean Geometry. [3] Fishback. Inc.html. “Introduction to Geometry. “Linear Geometry with Computer Graphics. “Hilbert. 1991. Andrews.M. David.uk/history/index. Translated with introduction and commentary by Sir Thomas L. New York. LaSalle. “The Real Projective Plane. “Foundations of Geometry. Second Edition.. Revised and Enlarged by Dr.. 1978.stanford.html [8] Reid.S.ac. Illinois.” http://wwwhistory. New York.. 1969. http://plato.” Vol. [6] “MacTutor History of Mathematics. Paul Bernays. 1956.cuny.york..” Second English Edition.” web notes. Inc.standrews. [4] Hilbert. Constance. University of St. http://www. New York. Inc. Translated by Leo Unger.” Springer Verlag.. Heath. John and Meighan Dillon. New York. 1996.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.