Nevertheless, it is truly remarkable that our experiences of spatial extent and of motion through space can be organized so completely. By taking certain figures-line, plane, segment, angle, and circle-as basic ones, and by assuming certain simple facts as axioms, other geometric figures can be constructed and other geometric facts deduced from the assumed axioms. Some of the resulting facts are surprising; for example, the three medians of a triangle meet in a point, as do the three altitudes.
The resulting deductive structure of Euclidean geometry is a model of Mathematical method.
2. Proofs Without Figures
In elementary Euclidean geometry, proofs of various facts about figures are developed from the axioms, with occasional implicit use of intuitively evident properties of the figures concerned. For instance, to find the per
pendicular bisector of a segment
AB
one constructs two circles with centersA
andB
and both with radiusAB;
these circles obviously intersect on opposite sides of the lineA B,
. . .. But what is meant by a "side" ofAB
and why must the circles intersect? (See Figure 1 .) If
all
facts are t o be deduced from axioms, then these plausible observations must also be proved from the axioms, rather than from inspection of the figure. Once this austere requirement was recognized, it turned out to be possible to add suitable new axioms to those of Euclid so that all arguments could be free of any intuitive or pictorial content. This meant that geometry was not really "about" the figures themselves, but was about certain corresponding notions defined exclusively in terms of basic notions to be taken as primitive or undefined. This austere program was systematically carried out by David Hilbert in a book,Grundlagen der Geometrie
(B. G . Teubner, first edition 1 899, twelfth edition 1 977). His axioms were for both plane and solid geometry; we will examine only those for the plane.He presented the axioms in the five following groups:
Group I (Incidence Axioms). These axioms involve only three primitive notions: "point", "line", and "the point P lies on the line I ". They require :
( 1 ) Two distinct points P and Q lie on one and only one line. (Except for the meticulous logical formulation, this is the familiar requirement that "two points determine a line".)
(2) There are at least two points on every line.
(3) There are three points P, Q,
R
not all on a line.With these axioms, one can define a
triangle
(the figure formed by three non-collinear points). However, one cannot yet define "right angle".64 I I I . Geometry
Figure 1 . Do the circles meet?
Group II (Axioms of Order). In Euclidean geometry a line does not have a selected direction, so we cannot describe the order of points P, Q on a line in terms of binary relations such as "P lies to the left of Q" or "P comes before Q" or "P
<
Q". Instead, the description is given means of aternary
relation"B
lies betweenA
andC",
available whenA, B,
andC
are all known to lie on a line. The axioms then read:
( I )
IfB
is betweenA
andC,
thenB
is betweenC
andA .
(2) If
A
andB
are distinct points on a line I, there exists on I a pointC
between
A
andB
and a point D such thatB
is betweenA
and D.(In elementary geometry, one often speaks of "prolonging the line seg
ment
AB
beyondB.
This axiom provides the existence of at least one point D in this prolongation.)(3) If
A, B,
andC
are three distinct points on a line I, then exactly one of these three is between the other two.With these axioms, one can define the
segment AB
forA
oFB,
to consist of all the pointsC
betweenA
andB
on the line determined byA
andB.
One is now in a position to formulate such pictorially obvious facts as '�
line divides the plane into two parts". Specifically, this means that given a line I, all points in the plane but not on I can be put in one of two non
empty disjoint sets U and V (the two "sides" of I) such that any segment
AB
joining a pointA
in U to a pointB
in V meets I in one point, while any segment joining two points of U (or, two points of V) does not meet I.To establish this "intuitively evident" separation property, it is necessary to assume an additional axiom :
(4) Pasch's Axiom. If a line I meets the segment
A B
of a triangleABC
and does not contain
C,
then I meets eitherA C
orBC
(Figure 2).In pictures, this states that a line which crosses into the triangle over side
AB
must again come out, either throughA C
or throughCB
(or through C). From the austere, no-pictures, point of view, it was this axiom which was most sharply missing in the Euclidean formulation. This axiom will show that each triangle divides the plane into two parts, an "inside" and2. Proofs Without Figures
65
c
A B
Figure 2. Pasch's axiom.
an "outside". One can also prove that any closed polygon (which does not cross itself) divides the plane into two parts, and inside and an outside
but the proof is quite involved because the polygon may be convoluted, with many reentrant angles.
With these axioms, one can also define that is meant by a ray or "half line" starting at a point
A .
IfB
is a second point and I the line determined byA
andB,
then the rayr
fromA
containingB
consists of all the points betweenA
andB,
the pointB,
and all points D withB
betweenA
and D.An
angle L rs
between two rays r ands
can then be defined as the figure formed by two (distinct) rays r ands
from the same point. In particular, this defines the (three) angles of a triangleABC.
One also defines straight angle (formed by two rays on the same line).Group III (Congruence). Next one introduces two new undefined terms,
"Segment
AB
is congruent to segment A'B "',
in symbolsAB
==A 'B ',
and"Angle
rs
is congruent to angler 's "',
in symbolsLrs
==Lr 's '
. The corresponding axioms require :(I) Given a segment
AB
and a rayr
fromA
" there is a unique pointB '
on
r
withA B
==A 'B '.
In simpler language, this states that we can "lay off" the length
AB
on a given line fromA
' in a given direction.(2) Congruence of segments is reflexive, symmetric, and transitive.
(3)
AB
==A 'B '
andBC
==B 'C '
implyA C
==A 'C ',
providedB
is betweenA
andC
andB '
betweenA '
andC '.
This amounts to describing the addition of segments.
(4) Given an angle
Lrs
and a rayr '
fromA '
on a line I, there is a unique rays '
fromA '
on a given side of I so thatLrs
==Lr 's '.
This axiom specifies that one can "lay off" the angle
Lrs
from a given rayr '
atA '
and on a given side of the line ofr '.
(5)
If two trianglesABC
andA 'B 'C'
haveAB
==A 'B ', BC
==B 'C '
andLB
==LB ',
then alsoLA
==LA '
andLC
==L C '.
66
I I I . Geometry Given this much about the two triangles, one proves thatA C
==A 'C '
and hence that the triangles are congruent. This is the familiar first congruence theorem (side-angle-side, or SAS) of Euclidean geometry. It is conventionally "proved" by moving one triangle till its parts coincide with the other.
Though we regard such "motion" as lying in the practical origins of geometry, it use in a formal axiomatic proof is not acceptable ; hence this axiom. From the congruence axioms, one can define right angles.
Two lines are defined to be
parallel
when they do not meet (have no point in common). One then requires the following famous axiom :Group IV (The Parallel Axiom). Given a point
A
not on the linem,
there is at most one line throughA
parallel tom.
Group V (Continuity Axioms). Examples show that the axioms stated so far do not yet insure that the Euclidean plane contains all the points that should be there, and no more. To achieve this, two more axioms are needed. To state the first, notice that a segment
AB
on a ray fromA
can be laid off repeatedly to given
pointsB
=BI, B2 , . . . ,Bn
on the ray withAB(
==BiBi+ I
and eachBi
betweenA
andBi+
t. fori
=1 ,
. . . , n - l (Figure3).
(1 ) Archimedian axiom.
GivenC
on that ray fromA,
there is a natural numbern
withC
betweenA
andBn .
That is, multiples of
A BI
will eventually exceed the given segmentA C;
in other words,C,
as measured by the unitAB,
cannot be infinitely far away.(2)
Completeness.
The system of points and lines satisfying the given rela-tions of incidence and congruence cannot be a part of a larger system of points and lines with the same relations satisfying all the same axioms.These are the two axioms of continuity as formulated in Hilbert's book.
However, this pair of axioms can be replaced by a single axiom which concerns the division of a line I into two rays. Each point 0 on I divides I into two rays; if S is the set of all points
A
on one ray andT
that on the other (and 0 is put in neither ray), then 0 lies between any pointA
of S and any pointB
ofT.
The axiom desired is essentially a converse to this:A c
Figure 3 . The Archimedean axiom.