You are on page 1of 10

1.

2 PROOFS IN AXIOMATIC GEOMETRY  11

pixels should be lit. Otherwise, find the neighbor pixel where the distance
from its center point to E is shortest and light up that pixel. In case of
a tie, pick at random. Now continue this process. Carry this out when
the start and end pixels are centered at (3, 2) and (6, 3), respectively. As
you do this, keep track of the number of additions, subtractions, and
multiplications you need to do. (Scan conversion may have to be done
many thousands of times to create a picture — methods that require
a lot of arithmetic are not preferred.) Compare it to the amount of
arithmetic done in the vertical-grid-line-crossing algorithm.
(c) Will the pixels produced by the distance algorithm described in part
(b) be the same as the ones produced by the vertical-grid-line-crossing
algorithm?

18. Let yi be the y coordinate of Ai . Let ri be yi rounded to the nearest integer (if
yi has 0.5 as its fractional part, round up). Define di = yi − ri .
(a) Show how you can find ri+1 from di+1 and m = e2 − s2 /e1 − s1 , the
slope of SE.
(b) Give an algorithm for finding di+1 from di and m.
(c) Combine your answers to parts (a) and (b) to form an algorithm for
finding all the centers s1 + i, ri  of the pixels to light up.
(d) Modify your algorithm so that it contains only integer quantities.

1.2 Proofs in Axiomatic Geometry


As we have seen, we need axioms to start on the process of proving theorems — so
it appears that listing our axioms should be our next order of business. But just as a
tool is designed based on how it will be used, we have chosen our axiom set to fit a
modern style of proof. Before discussing our axioms, in the next section, we pause to
discuss how the concept of proof has evolved in geometry.
In the ancient world a good deal of geometric knowledge was accumulated by
a variety of cultures, including the Babylonians, the Chinese, the Egyptians, and
the Hindus, beginning as early as 2000 b.c. In all of these cultures, up until about
600 b.c., arguments were sometimes offered for geometric assertions. For example, the
Sulvasutras, which date from 600 b.c. or earlier in India, contain a hint of a proof of a
special case of the Pythagorean theorem. In around 1000 b.c. a Chinese work called
the Jinzhang Suanshu offered a sketch of a proof of this theorem in somewhat greater
generality. However, neither of these works presented comprehensive proofs for every
proposition. It appears that many propositions were accepted because they seemed
obvious to the eye or because they had been checked by measurement in special cases.
12  CHAPTER 1 THE AXIOMATIC METHOD IN GEOMETRY

As one might expect, without rigorous standards of proof, a lot of errors and
crude approximations were accumulated along with correct principles. For example,
it was common to compute the circumference of a circle by multiplying the diameter
by 3. This approximate formula even made its way into the Old Testament (I Kings,
7:23). For another example, the Egyptians were accustomed to computing the area of
any quadrilateral, even if it was not a rectangle, by the formula for rectangles.
Around 300 b.c., a Greek scholar named Euclid published a geometry book called
the Elements, which forever put an end to the old style of doing geometry. This great
work was actually the capstone of work done by many geometers over the preceding
few hundred years, beginning with Thales (c. 640–546 b.c.). Not only were proofs
required in this approach to geometry, but the standard of proof was demanding
enough — “rigorous” enough, in the modern jargon — that this standard was
accepted until late in the nineteenth century. Proofs were required to be carried
out by the axiomatic method — something we’ll soon see in detail. This approach
to geometry resulted in a significant expansion of securely established geometric
knowledge.
During the nineteenth century, it became clear that the Euclidean standards of
proof were not as high as they could be: Steps in proofs were often accepted because
they seemed obvious to anyone looking at the picture. A new standard arose, one
we’ll call symbolic rigor. This style of geometry rejects the evidence of our eyes in doing
proofs and demands that every step be justified by an axiom or previous theorem.
A consequence of this is that the symbols of a proof (the words and mathematical
notations) must be convincing by themselves. If a picture accompanies the proof, as
it usually does, it is only a crutch.
Interestingly, the symbolic approach to geometry has some roots in the time of
Euclid, even though it did not really blossom until the work of geometers such as
Moritz Pasch (1843–1930) and David Hilbert (1862–1943). For example, Aristotle (384–
322 b.c.) proclaims: “The geometer bases no conclusion on the particular line which
he has drawn being that which he has described, but [he refers to] what is illustrated by
the figures.” Euclid’s extendibility axiom (previous section) provides a more concrete
example of the early desire to replace visual intuition with verbal reasons in proofs.
This axiom was included in the Elements because, in many proofs, it is necessary to
extend a line segment and Euclid wanted to have an axiom to justify this step. But
suppose the axiom had been left out and, whenever a line segment was extended in
a proof, no reason was given to justify it. Who would object? A reader who believed
that a proof only had to be visually persuasive would surely not object. By including
this axiom, it appears that Euclid was aspiring to the symbolic standard and was not
satisfied with visual acceptability.
But Euclid was inconsistent. Here is a “proof” that depends on something being
visually obvious, which is typical of proofs found in Euclid’s Elements. It would be
regarded as flawed according to the standards of symbolic rigor.
1.2 PROOFS IN AXIOMATIC GEOMETRY  13

Figure 1.4 Base angles of an isosceles triangle are congruent.

ASSERTION 1
If two sides of a triangle are congruent, say, AB = BC in Figure 1.4, then the angles
opposite these sides are congruent, m∠A = m∠C.

PROOF
Let M be the point where the angle bisector of angle B meets segment AC. Now exam-
ine triangles BAM and BCM. These are congruent by “side-angle-side.” Thus ∠A is
congruent to ∠C since corresponding parts of congruent triangles are congruent. 

For many years, this proof appeared in textbooks and was considered totally
convincing. However, mathematicians with the highest standards of rigor would say
the proof has at least one step we have not justified: How do we know that the angle
←−→
bisector crosses line AC ? And if it does, how do we know the crossing M is between
A and C? Visually speaking, these assertions seem obvious. But if we are aiming for
symbolic rigor, the standard of evidence is higher than “it looks like it to me, at least
in this case.” Of course this doesn’t mean that the theorem should be rejected. A
devotee of symbolic rigor would try to fill the gap in the proof or look for another
proof. For this theorem, other proofs are available; we give one in the next chapter.
But there were other theorems for which no airtight proofs could be found by
means of the axioms set forth by Euclid. It became necessary to add some axioms to
14  CHAPTER 1 THE AXIOMATIC METHOD IN GEOMETRY

Euclidean geometry that Euclid had never contemplated. We return to this matter in
the next section when we set forth our axiom system.
If you thought that the argument given for Assertion 1 was convincing, and if
you thought that time spent on showing M exists and lies between A and C is time
wasted, then you may doubt whether symbolic rigor has any merit. Here are some
things in its favor.
Our eyes sometimes deceive us. Just as the magician’s hand is quicker than the eye,
once in a while an invalid proof based on an inaccurate picture can fool the eye. Here
is an example of a ridiculous assertion with a “proof” in which it is hard to find the
fault.

ASSERTION 2
Every point inside a circle, except for the center, is actually on the circumference of
the circle (Figure 1.5).

PROOF −

Let P be any point inside the circle of radius r centered at O. On ray OP, find point Q


so that OQ = r 2 /OP. Choose M as the midpoint of PQ. Erect a perpendicular to OQ
at M, meeting the circle at R.

Figure 1.5 The inside of a circle is really on the boundary.


1.2 PROOFS IN AXIOMATIC GEOMETRY  15

1. r 2 = dd + 2a Construction of Q


2. = a + d2 − a2 Algebra
3. r + a2 = a + d2
2
Transposing in 2
4. r 2 + a2 + b2 = a + d2 + b2 Adding b2 to both sides
5. a 2 + b2 = c 2 Pythagorean theorem
6. a + d2 + b2 = r 2 Pythagorean theorem
7. r 2 + c2 = r 2 Substituting 5 and 6 in 4
8. c=0 Algebra applied to 7
9. P=R Step 8 says their distance apart is 0 

What we have “proved” is obviously incorrect. But where is the flaw? Our eyes
have been fooled by an inaccurate picture. A more accurate picture would show that
Q is improperly positioned and that M lies outside the circle, so R does not even
exist. Thus steps 4, 5, and 6 of our “proof” are invalid since they deal with lengths
and triangles that don’t exist. An adherent of symbolic rigor would not accept these
steps unless a proof could be given that M is inside the circle. Obviously, no such
proof can be given.
People do not always agree on what is obvious or correct. For example, during the time
when the Greek geometers were listing the extendibility axiom as self-evident, Greek
cosmologists were asserting that the entire material world, including the earth, sun,
moon, planets, and stars, was contained in one large sphere. There was nothing
beyond that sphere and no way to move outside that sphere. The universe had an
outer wall. In this view of things, a line segment with an endpoint on that sphere
could not be extended beyond that point.
The contradiction between the extendibility axiom and the “one large sphere”
view of the universe did not seem to bother the Greeks too much. But extendibility
does play a role in one of the sadder episodes of European intellectual history though.
In the year 1600, Giordano Bruno (Figure 1.6) was driven in an oxcart through the
streets of Rome to be burned at the stake in a public square. The path was lined with
onlookers who shouted to him to recant his heretical ideas. What were these ideas
that so enraged the authorities of the day? Bruno was a philosopher, well ahead of
his time, and one of his ideas was to take seriously Euclid’s notion that line segments
could be extended. This appeared to make the universe infinite. Bruno went further
to believe that, since the universe went on forever, there would be other worlds like
the earth that might contain intelligent life. Thus, he was perhaps the first believer in
extraterrestrial intelligence. This was flatly contradictory to prevailing religious beliefs
about the creation of the human race and the earth.
Modern scientists no longer argue whether the entire universe is inside some
big sphere — no one today believes in such a sphere. However, other disagreements
have arisen about the geometric nature of the universe. Do this mental experiment:
Imagine an ant crawling around the equator of a large sphere, like the earth. At any
16  CHAPTER 1 THE AXIOMATIC METHOD IN GEOMETRY

Figure 1.6 This statue of Giordano Bruno stands today in the very spot in Rome where he
was burned at the stake in 1600 A.D. That spot is now an outdoor flower market, the
Campo de Fiori.
1.2 PROOFS IN AXIOMATIC GEOMETRY  17

point the ant can always crawl more, so it seems that an extendibility property of
sorts is true in the ant’s world. But whether the ant can detect this or not, eventually
it will circle back around to where it started. So this is not the sort of extendibility
Euclid meant. A similar conclusion would be true if the ant were crawling on a large
donut. There are no walls to stop a crawl, but the “world” of the donut, like that
of the sphere, is not infinite in size. Some physicists maintain that our universe is a
sort of three-dimensional version of a sphere or donut, in the sense that there are
no walls that a light ray would bump into as it hurtles through space, but the light
ray wouldn’t “go to infinity” because the universe is not infinitely large. In some way
or other, the light ray would “go around in circles.” These scientists claim that we
are unable to notice this strange shape of our universe because the universe is large
and our experience of it is limited. By analogy, think of the ant on the sphere. If the
sphere is very large, how could the ant tell that it was not an infinite plane?
Since people do not agree on what is “obviously” true, we need to justify each
step of a proof — even those that seem obvious to us — by recourse to assertions
that have been agreed on in advance, namely, the axioms.
There are problems where neither looking at pictures nor measuring can help. One such example
was the famous problem of trisecting an arbitrary angle using ruler and compass. The
Greeks understood how to cut any given angle in half — this problem has an easy
solution often studied by beginning students in geometry (see Section 1.2 Exercise 3).
It seemed logical that one also ought to be able to use the ruler and compass to cut
an angle into three equal parts. However, for 2000 years geometers were unable to
find a way to carry out the trisection. Staring at figures and measuring their parts was
not the least bit helpful. In the nineteenth century it was demonstrated that trisecting
any given angle with ruler and compass was impossible. Of course, this does not mean
that there are no trisecting rays for the angle in question. What it means is that the
allowable tools, ruler and compass, are inadequate to construct those rays. This is a
bit analogous to discovering that you cannot drive to Europe from North America.
It doesn’t mean that Europe does not exist — it merely means that the network of
roads, bridges, and tunnels won’t get you there. Returning to the matter of trisection
with ruler and compass, not only was studying a picture not helpful in finding a
trisection method, it was also useless in understanding why no such construction
could exist. The proof of impossibility used abstract algebra, a theory that is notable
for proceeding in a purely symbolic way, with no help from pictures.
Without symbolic rigor, people sometimes disagree about whether a proof is right. There is at
least one important episode in the history of mathematics where the shortcomings
of the Euclidean standard of proof led to uncertainty about an important issue in the
minds of some of the best mathematicians. During a period of 29 years, the eminent
mathematician Adrien Marie Legendre (1752–1833) published numerous “proofs” that
one of Euclid’s axioms, one concerning parallel lines (which we will encounter in the
next section), could actually be proved from the other axioms. Finding such a proof
had long been a sort of holy grail of geometry, easily as famous as that of trisecting an
18  CHAPTER 1 THE AXIOMATIC METHOD IN GEOMETRY

angle. Legendre’s attempts at proof were not widely accepted. This standoff between
Legendre and his doubters would have been avoided if symbolic rigor had made
its appearance during Legendre’s time. We return to this issue in the next two
chapters.
Geometric-theorem-proving software requires symbolic rigor. An interesting side effect of
symbolic rigor, probably not anticipated by the mathematicians of yesteryear who
originally promoted it, is the fact that computer software systems for proving geomet-
ric theorems are based on manipulating symbols in the computer’s memory. For this
reason, an understanding of symbolic rigor is essential for those wishing to devise such
software. This is ironic, because symbolic rigor has often been criticized as complicated
to the point of being impractical, whereas computers are considered very practical.
Surprisingly, despite these numerous arguments in favor of symbolic rigor, it
is little used in research and in applications of geometry. It makes its appearance
mainly in teaching some advanced geometry courses. Some reasons for this include
the following.

1. Seeing is believing. Mathematicians aspire to transcend this proverb, but we are


only human. It is no accident that the word theorem comes from the Greek
word theorein, which means “to look at.”
2. Few or no theorems have had to be recalled. If the grading standard for a course is
raised, more students will probably fail. Analogously, because symbolic rigor
is a tougher standard of proof than the Euclidean standard, one might expect
that some geometric assertions that had been widely regarded as correctly
proven by the Euclidean standard would turn out not to be provable when the
symbolic standard of rigor is applied. This never happened to any significant
degree. What has happened is that new, longer proofs have been proposed
for the old, familiar theorems of Euclid.
3. Symbolic rigor often leads to long, complicated proofs of painfully obvious statements. This
makes geometry time consuming, never a good thing for people who want
to make progress quickly. It also makes geometry hard to learn for all but
the most advanced students.

As our last word on this awkward issue of how much rigor to insist on, the
following quotation from Heath shows that the issue is age-old and was already being
disputed in Euclid’s time. This quotation pertains to Euclid’s proof of the following
theorem, which many would regard as not requiring proof at all.

* T H E O R E M : The Triangle Inequality


In any triangle, each side is shorter than the sum of the other two side lengths. 
1.2 PROOFS IN AXIOMATIC GEOMETRY  19

Figure 1.7 Jack forgets the triangle inequality.

In Sir Thomas Heath’s A History of Greek Mathematics, he reports (Figure 1.7):


It was the habit of the Epicureans, says Proclus, to ridicule this theorem [the triangle
inequality] as being evident to even an ass and requiring no proof, and their allegation
that the theorem was “known” even to an ass was based on the fact that, if fodder is
placed at one angular point and the ass at another, he does not, in order to get his food,
traverse the two sides of the triangle but only the one separating them  Proclus
replies truly that a mere perception of the truth of a theorem is a different thing from
a scientific proof of it and a knowledge of the reason why it is true.
We’ll have a short look at symbolic rigor in the next section as we begin proving the-
orems in Euclidean geometry. After that section, we return to the more relaxed standard
established by Euclid and still used in most research, applications, and teaching.

Section 1.2 Exercises



Marks challenging exercises.
1. (a) Using the notation of Figure 1.5, verify line 2 of the proof of assertion 2
and prove that d + a2 = r 2 + a2 . (Hint: Don’t use the picture, which is,
after all, wrong.)
(b) Use part (a) to prove d + a ≥ r, proving that M is not inside the circle.
20  CHAPTER 1 THE AXIOMATIC METHOD IN GEOMETRY

2. (a) Suppose ABC is a triangle with m∠B = 90 . Does the following con-
struction create a trisecting ray? Assume that each construction step
can be carried out and that each intersection point actually does
exist.
(i) Extend BC to a segment BD, which is twice as long as BC.
(ii) With your compass opened to length BD, place the point at C and


draw a circle that meets ray BA at E. What are the angles of triangle
BEC?
←−→
(iii) Drop a perpendicular from B to line EC, meeting it at F. What is
m∠FBC?


Is BF a trisecting ray for ∠ABC?
(b) If you believe this construction creates a trisecting ray, why does this
not contradict our discussion of trisection in this section?
3. Here is a construction of the angle bisector of ∠ABC. Do you think it meets
the standard of symbolic rigor? Explain.

→ −

(i) Draw a circle of any radius centered at B, meeting BA at E and BC at F.
(ii) Draw circles of the same radius, centered at E and F, and let G be an
intersection of these circles.
Triangles BEG and BFG are congruent by side-side-side, and the measures of
∠EBG and ∠FBG are equal because these angles are corresponding parts.
4. Criticize the construction of a triangle with two right angles shown in
Figure 1.8. Let two circles meet in points A and B. Let AC and AD be diameters
of these two circles from A C being on one circle and D on the other. Let
←−→
E be the point other than C where CD meets the circle containing C. Let F
←−→
be the point other than D where CD meets the circle containing D. Because
∠AEC is inscribed in a semicircle, it is a right angle. Likewise ∠AFD is a right
angle. Thus triangle AFE has two right angles.

1.3 Axioms for Euclidean Geometry


In this section, we describe the axiom set we will use for our study of Euclidean
geometry. (Some of these axioms are also used in the two non-Euclidean geometries
described in Chapter 3.) We then carry out rigorous proofs of the first few deductions
from this axiom set. If time is limited, establishing the spirit of the enterprise is perhaps
more important than studying each axiom and theorem intensively.

You might also like