Veblen, Youn - Projective Geometry

PROJECTIVI
GEOMETRY
Oswald Veblen and John Wesley Youn
HUNT
PROJECTIVE GEOMETRY
Projective
BY
Geometry
II
VOLUME
OSWALD VEBLEN
OSWALD VEBLEN
Late Professor of Mathematics, Princeton University
JOHN WESLEY YOUNG

Late Professor of Mathematics, Dartmouth College
BLAISDELL PUBLISHING COMPANY DIVISION OF GINN AND COMPANY

New
York
Toronto
London
A,
1L
A,
Ji,
J1
A, &, f
or A,
Ji,
r,
fying A, E,
(the properly projective spaces)
Among the spaces satismay be mentioned the

and
modular
spaces, the rational
nonmodular
of
space, the real space,
be specified categorically by adding the proper assumptions to A, E, P. The passage from the point of view of general projective geometry to that of the particular spaces is made in the first chapter of this volume.
the complex space.
Any
one
these
may
Having
groups
(1) the
fixed attention on
of transformations to
any particular space, we have a set of each of which belongs its geometry.
For example,
group
in the complex projective plane
we
find
among
others,
of all-continuous one-to-one reciprocal transformations
geometry),
(analysis situs), (2) the group of birational transformations (algebraic (3) the projective group, (4) the group of non-Euclidean
(5) a
geometry,
(cf.
sequence
of
groups connected with Euclidean geometry
The groups (2), (3), (4), and (5) all have analogues in the other spaces mentioned in the paragraphs above, and consequently
54).
it is
in such a
desirable to develop the theorems of the corresponding geometries way that the assumptions required for their proofs- are
put in evidence in each case. This will be found illustrated in the chapters on affine and Euclidean geometry. The two principles of classification, (a) and (6), give rise to a
double sequence of geometries, most of which are of consequence in present-day mathematics. It is the purpose of this book to give an elementary account of the foundations and interrelations of the
more important
of these geometries (with the notable exception of (2)). I venture to suggest the desirability of other books taking account of this logical structure, but dealing with particular types
May
of geometric figures
such books should be not merely to prove every theorem, rigorously but to prove it in such a fashion as to show in
ideal
of
The
which spaces it is true and to which geometries it belongs. Some idea of the form which would be assumed by a treatise on conic
sections written in this fashion can be obtained from
83 below.
type of exposition would be feasible at the present time are quadric surfaces, cubic and quartic curves,
for
this
Other subjects
which
rational
curves,
configurations, linear
line
geometry, collineation
groups, vector analysis. Books of this type could take for granted the fouudational and coordinating work of such a book as this one, and thus be free to
use all the different points of view right from the beginning. On the other hand, a general work like this one could be much abbreviated if there were corresponding treatises on particular geometric figures
(for
example, conic sections) to which cross references could be made.
OSWALD VEBLEN
BBOOKLIN, MAINE AUGUST, 1917
CONTENTS
CHAPTER
SECTION
1. 2.
FOUNDATIONS
FAQB
1
8.
4.
6.
6.
Plan of the chapter List of Assumptions A, E, P, and Assumption K Double points of projectivities
3
6
Complex geometry Imaginary elements adjoined to a real space Harmonic sequence 8. Assumption H 9. Order in a net of rationality *10. Cuts in a net of rationality
7.
6
7
9
11
18
14 16
21
of the assumptions
*11.
Assumption
of continuity
and independence complex geometry Ordered projective spaces *16. Modular projective spaces
*14. [Foundations of the *15.
17. Recapitulation
*12. Chains in general *13. Consistency, categoricalness,
...
23
29 32 33 36
CHAPTER
II
ELEMENTARY THEOREMS ON ORDER

18.
19. 20.
Direct and opposite projectivities on a line The two sense-classes on a line
37 40
43
Sense in any one-dimensional form

Separation of point pairs
21.
44
45 47
22.
23.
Segments and intervals

Linear regions
24. Algebraic criteria of sense 25. Pairs of lines and of planes

26.
49
50 52
The
triangle
and the tetrahedron

Cross ratios of points in space
27. Algebraic criteria of separation.
....
55
58 59
61
28. Euclidean spaces
Assumptions for a Euclidean space 30. Sense in a Euclidean plane *31. Sense in Euclidean spaces
29.
63
*32. Sense in a projective space
64
CHAPTER
SECTION
34.
III
THE AFFINE GROUP IN THE PLANE

PAGE
to a given
The geometry corresponding

Euclidean plane and the
group of transformations
...
70
71
35.
affine
group
36. Parallel lines

37. Ellipse, hyperbola,
38.
72
parabola
73 74
The group
of translations
39. Self -conjugate

40.
subgroups.
Congruence
78
Congruence of parallel point pairs Metric properties of conies 42. Vectors

41.
48. Ratios of collinear vectors 44.
45.
80
81 82
85
Theorems
of Menelaus, Ceva,
and Carnot
89
92 94
Point reflections
46.
47.
Extension of the definition of congruence
The homothetic group

Equivalence of ordered point triads
95
96
triads
48.
49. 50.
Measure of ordered point The equiaffine group
99
106
*51. Algebraic formula for measure.
Barycentric coordinates
106
*52. Line reflections

*53. Algebraic formulas for line reflections 54. Subgroups of the affine group
109 115
116
CHAPTER IV
EUCLIDEAN PLANE GEOMETRY
55.
56.
57.
Geometries of the Euclidean type Orthogonal lines
119
120
123 126 129
131
Displacements and symmetries. Congruence
58. Pairs of orthogonal line reflections

59.
60.
The group
Circles
of displacements
61.
62.
Congruent and similar triangles Algebraic formulas for certain parabolic metric groups
order relations
134 135
63. Introduction of
64.
138 140
142
The
real plane
65. Intersectional properties of circles
66.
67. 68.
69.
70.
The Euclidean geometry.

Distance
set of
assumptions
144
147
Area The measure of angles The complex plane

Pencils of circles
149
151
154
157 163
71.
72.
Measure
of line pairs
CHAPTER V
ORDINAL AND METRIC PROPERTIES OF CONICS
SBCTION
74.
75.
7(5.
PAGE
170 174
177
One-dimensional projectivities Interior and exterior of a conic
77. 78. 79.
80.
81.
Double points of projectivities Ruler-and-compass constructions Conjugate imaginary elements Protective, affine, and Euclidean Foci of the ellipse and hyperbola Focus and axis of a parabola
180
182
classification of conies
186 189 193 196

199 201
82. Eccentricity of a conic

83.
84.
Synoptic remarks on conic sections Focal properties of collineations
Homogeneous quadratic equations in three variables 86. Nonhomogeneous quadratic equations in two variables
85. 87.
202 208
Euclidean classification of point conies
210
212
88. Classification of line conies
*89. Polar systems
215
CHAPTEE VI
INVERSION GEOMETRY AND RELATED TOPICS
90.
Vectors and complex numbers
219
.
91.
92.
The
Correspondence between the complex line and the real Euclidean plane inversion group in the real Euclidean plane
222
226
93. Generalization
94.
95.
231 by inversion Inversions in the complex Euclidean plane 235 a the real Euclidean and between plane complex pencil Correspondence
of lines
238
241
96.
The
real inversion plane
97. 98. 99.
Order relations in the real inversion plane Types of circular transformations Chains and antiprojectivities
244 246 250 253 267

259
262 264 268 271
100. Tetracyclic coordinates 101. Involutoric collineations

102. The projective group of a quadric 103. Real quadrics 104. The complex inversion plane
105. Function plane, inversion plane, and projective plane 106. Projectivities of one-dimensional forms in general *107. Projectivities of a quadric *108. Products of pairs of involutoric projectivities
273
277
281
109. Conjugate
imaginary lines of the second kind
CHAPTER
SECTION
111. Afflne
VII
AFFINE AND EUCLIDEAN GEOMETRY OF THREE DIMENSIONS

PAGE
geometry
lines
287
288
112. Vectors, equivalence of point triads, etc

113.
The parabolic metric group. Orthogonal
and planes
293
114. Orthogonal plane reflections 115. Displacements 116. Euclidean
295
297 301
and symmetries. Congruence

of three dimensions
geometry
*117. Generalization to
118. Equations of
n dimensions the affine and Euclidean groups

volume, angular measure
displacement into orthogonal line reflections
304
305
311
119. Distance, area,

120.
The sphere and other quadrics
315
121. Resolution of a
....
317 321
122. Rotation, translation, twist

123. Properties of
displacements
325
124.
Correspondence between the rotations and the points of space 125. Algebra of matrices
126. Rotations of
....
328
333 335
337
an imaginary sphere
127. Quaternions 128. Quaternions and the one-dimensional projective group *129. Representation of rotations and one-dimensional projectivities by
339 342
points
130.
Parameter representation
of displacements
344
CHAPTER
VIII
NON-EUCLIDEAN GEOMETRIES
131. 132.
Hyperbolic metric geometry in the plane
350 352 355
Orthogonal lines, displacements, and congruence 133. Types of hyperbolic displacements 134. Interpretation of hyperbolic geometry in the inversion plane 135. Significance and history of non-Euclidean geometry
136.
....
357
360
362
Angular measure
formulas for distance and angle
of three dimensions
137. Distance 138. Algebraic
364
365
*139. Differential of arc

140. Hyperbolic
366 369
371
141. Elliptic plane
142. Elliptic
143. 144.
145.
geometry geometry. Definition geometry of three dimensions Double elliptic geometry Euclidean geometry as a limiting case of non-Euclidean Parameter representation of elliptic displacements
373 375 375 377
CHAPTER IX
THEOREMS ON SENSE AND SEPARATION
SECTION
147.
PAGK
of the
Plan
chapter
385
148.
149.
Convex regions
Further theorems
385
on.
150.
Boundary
of
convex regions a convex region
388 392
151. Triangular regions 152. The tetrahedron

153. Generalization to
154.
396
397
155.
156. 157.
158. 159. 160.
n dimensions Curves Connected sets, regions, etc Continuous families of sets of points Continuous families of transformations Affirie theorems on sense Elementary transformations on a Euclidean
400
401
404
405 406
407
line
409
Elementary transformations in the Euclidean plane and space Sense in a convex region 162. Euclidean theorems on sense 163. Positive and negative displacements
101.
....
411
413
414
416 418 419
421
164. Sense-classes in projective spaces 165. Elementary transformations on a projective line 166. Elementary transformations in a projective plane
Elementary transformations in a projective space *168. Sense in overlapping convex regions

167.
423
424
425
*169. Oriented points in a plane *170. Pencils of rays
429
433 435
*171. Pencils of segments and directions *172. Bundles of rays, segments, and directions
*173. One- and two-sided regions 174. Sense-classes on a sphere
436 437
437 438
441
Order relations on complex lines 176. Direct and opposite collineations in space 177. Eight- and left-handed figures 178. Right- and left-handed reguli, congruences, and complexes *179. Elementary transformations of triads of lines *180. Doubly oriented lines
175.
443 446 447

451
*181.
182.
183.
More general theory of sense Broken lines and polygons
454
457
A theorem on simple polygons Polygons in a plane 185. Subdivision of a plane by lines 186. The modular equations and matrices 187. Regions determined by a polygon
184.
458
460 464
467
188. Polygonal regions and polyhedra 189. Subdivision of space by planes 190.
191.
473
475
The matrices v The rank of H~
H H
z,
and
JET 8
477
479
SECTION
192.
193.
PAGK
480
482 483 484 489 490 493 496
601
Polygons in space Odd and even polyhedra
194.
Kegions bounded by a polyhedron The matrices E l and E2 for the protective plane 196. Odd and even polygons in the projective plane 197. One- and two-sided polygonal regions 198. One- and two-sided polyhedra
195.
199. Orientation of space
INDEX
PROJECTIVE GEOMETRY
CHAPTER
I
FOUNDATIONS
1.
Plan
of the chapter.
In the
first
volume
of this
book
we have
been concerned with general protective geometry, that is to say, with those theorems which are consequences of Assumptions A, E, P. In but most of the many cases we also made use of Assumption
theorems which
(though
to
trivial)
we proved by the aid of this assumption remain true when this assumption is false. The class of spaces
is
which the geometry
assumptions used
of that space in
of Vol. I applies is very large, and the set of therefore far from categorical.
of course, to serve as
The main purpose of geometry is, which we envisage
a theory
ourselves
and external nature.
This purpose can be accomplished only partially by a geometry based on a set of assumptions which is not categorical. therefore pro-
We
ceed to add the assumptions which are necessary in order to limit attention to the geometry of reals, the geometry in which the number
system is the real number system of analysis. These assumptions are stated in two ways, the one on the theory
of
3)
dependent
the real number system and the other ( 7-13) also state the assumptions 14, 15, 16) independent of it. necessary for certain other geometries which are of importance
We
(5,
because of their relations to the real geometry and to other branches of mathematics. At the end of the chapter we give a summary of
the assumptions for the various protective geometries which
considering.
we
are
FOUNDATIONS
"a point is
[CHAP.
"
on a
line
"
or " a line
is
on a point
means
that the point
belongs to the line
(cf. p.
16, Vol.
I).
ASSUMPTIONS OF ALIGNMENT:
A
loth
line
1. If A and B A and B.
are distinct points, there
is
at least one line
on
A 2.
If
A and B
A and
C
(D
C,
are distinct points, there
is
not more than one
on both
B. are points
A3. If
not
all
A, B,
the
on
same
=
line,
and
and
E]
are
points such that B, C,
on a line and on a
line,
A,
is
D E
F
are
are
there
such that
a point A, B, F are on
are
a line and also D, E, on a

line.
ASSUMPTIONS OF EXTENSION
E 0. E 1. E 2. E 3.
E 3'.
There are at least three points on every There exists at least one line.
line.
All points are not on the same line. All points are not on the same plane* If S 8 is a three-space^ every point is on S 8
:
ASSUMPTION OF PROJECTIVITY
P.
If a
projectivity leaves each' of three distinct points of

it
line
invariant,
leaves every
point of the line invariant.^
ASSUMPTION
H
fl
The diagonal points of a complete quadrangle are noncollinear.

explained
As was
when Assumption P was
first
introduced, this
assumption does not appear in the complete list of assumptions for the geometry of reals, but is replaced by certain other assumptions
from which
of
it (as
well as
can be derived as
a-
theorem.
The
list
assumptions for this geometry will consist of Assumptions A, E, and the new assumptions.
3.
Assumption K. The most summary way

is to
of
completing the
list
:
of
assumptions for the geometry of reals
introduce the following

Vol. I)
is
K.
geometric
with the real
number system (Chap. VI, number system of analysis.

list
isomorphic*
Thus
A, E, K.
a complete
of
assumptions
for the
geometry of reals
is
Thd use of Assumption K implies a previous knowledge of the real number system. f Its apparent simplicity therefore masks certain real difficulties. What these difficulties are from a geometric point of view will be found on 7-13, where K is analyzed into independent statements H, C, R. reading
These
sections,
however,
may
be omitted,
if
desired,
on a
first
reading.
Since a geometric number system, in one one-dimensional form is isomorphic with any geometric number system in any one-dimensional
form in the same space, it is evident that the principle of duality valid for all theorems deducible from Assumptions A, E, K.
In order that the results
of reals,
it
is
of Vol. I
be applicable to the geometry

is
must be shown that Assumption P
a logical conse-
quence
tive
Assumptions A, E, K._ Since multiplication is commutain the real number system, this result would follow directly
of
7,
from Theorem
Chap. VI, Vol.

It is
ever, incomplete.
I. The proof shown (Theorem 6, loc.
there given
cit.)
is,
howholds,
if
that
if
multiplication is commutative; multiplication is commutative, be made as follows

:
but
it is
not there proved that
is satisfied.
The needed proof may
THEOEEM
1.
Assumption
is
valid in
any space satisfying

is
and E and such that multiplication Assumptions. in a geometric number system (Chap. VI, Vol. I).
Proof. It
is
commutative
obvious that the

the3,
two choices of Theorems 1 and
number systems determined by any fundamental points H^H^H^, are isomorphic (cf. Chap. VI, Vol. I), so that we may base our argument
on an arbitrary choice of these points. We are assuming that multiplication is commutative, and are to prove that any projectivity II
* This term
t
is
defined in
52,
is
Vol. I.
The
real
number system
rest ultimately on.the positive integers (cf. Pierpont,
Variables, pp. 1-94
postulates (of. Society, Vol. VI (1905), p. 17\.
thought of either as defined in: terms which Theory of Functions of Real or ; Fine, College Algebra, pp. 1-70) or by means of a set of E. V. Huntington, Transactions of the American Mathematical
to be
definition, II is the resultant of a
sequence
of perspeetivities
where
Chap.
[Jf]
Ill, Vol. I, this
denotes the points of the given line. By Theorem 5, chain of perspeetivities may be replaced by
three perspeetivities
Moreover, by Theorem 4, Chap. Ill, Yol. I, the pencils [P] and [Q] may be chosen so that their respective axes pass through two of the x and given fixed points of II. Let us denote these points by y
Em.
respectively and let H<* he the third fixed point.

tion of
their
By
another applicaso that
Theorem 4 the pencils [P] and [Q] may be chosen

point
common
since
is
on the
line
SHn
Now,
collinear.
Hm
T,
(fig.
2).
is
transformed into
is fixed, T,
,
itself, S,
Hx
and
must be
Since
Since
point
P of the line HXR forms P to a point Q of
H HH
y
J^
is fixed, S,
V,
it is
any point of the line transformed by the perspectivity with S as center to a

y
and
IIX and
must be
is
collinear.
are collinear.
If
the perspectivity with as center transthe line RH~y the perspectivity with U as
;
center transforms
to
back
to a
point
H
X
of the line
H H We
X
V
.
have
show
that
H'
= H.
Let JfQ he the trace on the line

of
HH
of
PT;
Q
let
Jf1 he the trace
RT;
and H'
is
the trace of UQ.
The complete quadrangle

and hence (Theorem
3,
TRSP
determines
I)
(JfQ
HH
v
ir
Chap. VI, Vol.
in the scale

and hence
in the scale
TRQU determines Q (H HxH

Q
lf
Hm H H'}
v
ffx
Since multiplication
is
-Hv = H'.
commutative,
H= H', which proves the theorem

P.
reader will find no difficulty in. using the construction above to prove that the validity of the theorem of Pappus ( 36, Vol. I) is necessary and
sufficient for the
The
commutative law of multiplication and for Assumption

projectivities.
4.
Double points of
DEFINITION.
A projective trans-
formation of a real line into

or elliptic* according as
It
it
be hyperbolic, parabolic, has two, one, or no double points.

itself is said to
I,
was proved in
58, Vol.
that the determination of the double
points of a projective transformation t
px
i
= ax + ox
1
'
depends on the solution

(2)
of the
equation
f-( A = ad
be.
where
its
This equation has two real roots

,
if
and only
if
discriminant
(a
is positive.
-\-dy- 4 A
,
7X2
Hence we have
the
If
A < 0,
transformation
(1) is hyperbolic.
For an
elliptic
or
parabolic projectimty
is
always
positive.
* These terms are derived from the corresponding types of conic sections 37). In a complex one-dimensional form a somewhat different terminology used (cf. 98). t In this volume we shall generally write homogeneous coordinates in the form (X& xj, whereas in Vol. I we used (a^, xn ).
(see
is
and hence
In case the projectivity (1) is an involution, a 4 A is the discriminant of (2). Hence
d(
54, Vol.
I),
An
involution
is
elliptic
or hyperbolic according as
is
positive
or negative.
of these theorems with the theory of linear evident on comparison with the first sections of Chap. II. deduction of the corresponding theorems from the intuitive concep-
The intimate connection

is
order
tions of order
is
to be
found in Chap.
IY
of the Geometria Projet-
tiva of Enriques.
EXERCISE
projectivity for which projectivity for which A,

5.
A>
:
is
is a product of two hyperbolic involutions. a product of three hyperbolic involutions.
Complex geometry.
of
of
a.
many problems
Assumption provides for the solution construction which could not be solved in
problem
net of rationality. But even in the real space the fundamental of finding the double points of an involution has no general
solution.
it is only necessary to set up an involution for which Take any involution of which two pairs of conjugate points AA' and BB' form a harmonic set H(AA', BB'). If the scale
To
see this
A>
,
0.
P P P x
l}
is
chosen so that
is
P
xx'
A'
=
1.
,,
B=P
19
then B'
(
= P_
and
the involution
represented
by the
bilinear equation
54, Vol. I)
The double points

equation
of this involution,
if
existent,
would
satisfy the
x2 =
real roots.
of
1,
which has no
An
replace
effect
Assumption
If,
is
thus
to
deny the
possibility of
solving this problem.

it
and negate Assumption by properly chosen other assumptions, we are led to a

however,
is
we
geometry in which this problem
always
soluble, namely,
the
which the geometric number system is isomorpbic with the complex number system of analysis. Although this geometry does not have the same relation to the space of external
geometry
of the space in
nature as the real geometry,
it is
extremely important because of
its
relation to other branches of mathematics.
One way of founding this geometry is to replace Assumption by another assumption of au equally summary character, namely,
J.
yeomdric number system
is
isomorpMc with
the
complex num-
ber system of analysis,
Since this
number system obeys

1 apply. J.
the commutative law of multisatisfies
plication, the corresponding
geometry
Assumption
P,
and
all
the theorems of Vol. plex geometry

is is
Thus, a set of postulates
for the
com-
A, E,
of
The problem
projectivity
finding the double points of a one-dimensional completely solvable in the complex geometry for
;
any such projectivity

( vo
may
,
be represented by the bilinear equation

j
,
54, Vol. I) '
cxx' -f- dx'

its
ax
* o
v,
and therefore
double points are given by the roots of
which exist in the complex number system. The analogous result holds good for an w-dimensional
In this case the problem reduces to that algebraic equation of the nth degree.
6.
it is
projectivity.
of finding the roots of
an
Imaginary elements adjoined
to
desirable to think of another point of
a real space. In this connection view which we may adopt
toward the complex space. Suppose we are working in a real geometry on the basis of A, E, see below). It is a (or of A, E, H, C, R theorem about the real number system* that it is contained in a
number system
equation
(the
are of the form ai
+1
complex number system) all of whose elements where a and 5 are real and i satisfies the
;
2
+l=0.
Hence
a theorem about the real space that it is contained in another space which contains the double points of any given involution.
it is
This
may
be seen in detail as follows
By
the theory of homo-
geneous coordinates the points of a real projective space S are in a correspondence with the ordered tetrads of real numbers (XQ xv xz x s),
,
,
except (0, 0, 0, 0), such that to each tetrad corresponds one point, and to each point a set of tetrads, given by the expression (mxQ) mx v
* This same question
is
discussed
from the point
of
view
of a general space
and
mentioned above, the

contained in the
set
set of all ordered tetrads of real
numbers
z
s)
is
of
all
ordered tetrads
(ZQ ,
z
1}
z2 ,
where
z
s
are
complex numbers.
Let us define a complex point as the class of all ordered tetrads of complex numbers of the form
(Jcz
,
Jcz
jfej,
2,
Jcz
B)
where
for a
all
given class ZQ , %1} %z , za are fixed
and not
all
zero
and k
takes on
classes
complex values different from zero. Let the satisfying two independent linear equations
set of these
be called a complex
that the set of all
line.
With
these conventions
it
is
easy to see
complex points and complex lines satisfies the assumptions A, E, P, and thus the complex points constitute a proper projective space. Let us call this space S c
.
The space S c contains the

(fCX
set of all
,
complex points
,
of the
form
KX^,
KXZ
KX^j
where XQ
points S r
tions of
x^,
x 2 xs are
,
any the form

.
If
set of
(3)
Let us call this subset of complex complex points of S r which satisfy two equawith real coefficients be called a "real line," we
all real.
have, by reference to the
homogeneous coordinate system in
S, that
the complex points of S r are in such a one-to-one correspondence with the points of S that to every line in S corresponds a " real line"
in S r ,
and conversely.
is
Thus, S r
a real projective space and
is
contained in the complex
projective space
Sc
Obviously S
may
where
also
S'
in a complex projective space S'
be regarded as contained consists of the points of S
together with the points of Sc which are not in Sr , and where each line of S' consists of the complex points of S' which satisfy two
equations of the form (3) together with the points of S whose coordinates satisfy the same
two equations.
DEFINITION. Points of the real space S are called real points, and points of the extended space S', complex points. Points in S' but not
in
are called
imaginary
points.
This discussion of imaginary elements does not require a detailed
knowledge or study
fact,
of the
complex number system
as such.
It
is,
in
a special case of the more general theory in Chap. IX, Vol. I (cf. particularly 92), which applies to a general projective space. It serves in a large variety of cases where it is sufficient to know
merely the existence of the complex space S' containing S and satisfying Assumptions A, E, P. It is a logically exact way of stating the point of view of the geometers who used imaginary points before the
advent
of the
modern function
theory.
There are problems, however, which require a detailed study of the complex space, and this implies, of course, a study of the complex number system and such geometrical subjects as the theory of chains
(see
11, 12, below,

is
and
later chapters).
a very elegant and historically important method of introducing imaginaries in geometry without the use of coordinates, namely, that due to von Staudt.* It depends essentially on the
properties of involutions
There
and
74-75
of this
volume.
which are developed in Chap. VIII, Vol. I, The reader will find it an excellent
exercise to generalize the
stated in Proposition
Staudt theory so as to obtain the result Chap. IX, Vol. I. 7. Harmonic sequence. We shall now take up a more searching
Von
2,
study of the assumptions of the geometry of

it
reals.
In Chap. IV, Vol.
I,
was proved that every space satisfying Assumptions A, E contains 3 a net of rationality R and that this net is itself a three-space which satisfies not only Assumptions A and E but also Assumption P (Theorem 20). To this rational subspace, therefore, apply all the theorems in Vol. I which do not depend essentially on Assumption H For example, every line of R 3 is a linear net of rationality and may
,
.
be regarded (with the exception of one point chosen as oo) as a commutative number system all of whose numbers are expressible as rational combinations of and 1.
Throughout VoL I we
It
left
the character of this net indeterminate.
might contain only a finite number of points or it might contain an infinite number. We propose now to introduce a new assumption which will fix definitely the structure of a net of rationality.
* Cf K. G. C. von Staudt, Beitrage zur Geometrie der Lage, Niirnberg (1856 and J. Luroth, Mathematische Annalen,Vol.VIII(1874), p. 145. Segre, Memori rlfilln. R Annarlpmia. rlplle srianyfi Hi Tnrinn (%\. Vnl XXXVTIT
. .
1857).
aria leu Jx
ue a point ui
iiiuej.seuu.uij.
tw.
points of the line h by [//] and those of the line K^H^ by let II be a projectivity defined by perspectivities as follows
[./f],
:
and
The
set of points
H^,
Hv H
,
--,
H^
JTi+
such that TL(HJ)
= ITi+1
together with the set
such that II (.HI ._J = If_ i}

-ffoo
is
called a
is
harmonic sequence. The point
is
not in the sequence but

projectivity II is
evi-
called its limit point.

-^
The
dently parabolic to sr O im
and
carries
THEOREM
2.
The middle one
* of any three consecutive points
of a harmonic sequence is the harmonic conjugate of the limit

point of the sequence with re-
gard
to the
other two.
.
Proof.
By
construction -we have
* This term refers to the subscripts in the notation Ej.
57,8]
HARMONIC SEQUENCE
to
11
the
COROLLARY. All points of a harmonic sequence belong

net of rationality.
same
THEOREM
by
Q>
I}
3.
Two harmonic
M M M
Proof.
sequences determined by
N Hv H*, and
Q,
x are protective in
any
projectivity II by which
By Theorem 3, Chap. IV, Vol. I,

sets of points into
the projectivity II transforms
harmonic
8. to
harmonic
sets.
Assumption H. By reference to fig. 3 it is intuitively evident most observers that in any picture which can be drawn representing points by dots, and lines by marks drawn with the aid of a straightedge,
no point
which can be accurately marked

the other hand, there
is
will ever coincide
with Hj(i =/).
On
and E to prove that Hf 3= J} because (Introduction, 2, Vol. I) these assumptions are all satisfied by the miniature spaces discussed in 72, Chap. VII, Vol. I, and if the number of points on a line is Thus we are led to finite, the sequence must surely repeat itself.
nothing in Assumptions
make
a further assumption.
exists,
ASSUMPTION H.* If any harmonic sequence contains only a finite number of points.
not every one
The
existence of a harmonic sequence determined by
any three
and E. That any two points follows directly from Assumptions sequences are projective follows from Theorem 3. Hence Assumption
H gives
at once
4.
THEOREM
Any
three distinct collinear points
0>
H^,
HM
deter-
mine a harmonic sequence containing an infinite number of points and and as consecutive points and v as the limit point. having
The principle of duality deducible from Assumptions A, E, H. Proof. This principle has been proved
5.
THEOREM
is
valid for
all
theorems
in
77^
,
Chap.
n
I,
Vol. I, for all
theorems deducible from
and E.
If
on a line
projection
I,
let a line
I
/'
meet them
^ are any three planes in H H^, H respectively. The

rj Q
,
n is
harmonic sequence determined on I' by Q H^ the space dual of a harmonic sequence of points. Since the
by
of the
evident on the basis of Assumptions and E alone that the transx a is a parabolic projectivity. Denoting it by a, it formation x'
= +
if
,
is
clear that
there
Jc
is
n any integer n such that a
is
m then a"* +
= am
is
the identity,
period, there
Hence, if a has a finite only a finite number of points in a harmonic sequence,

wi being
and
any
integers.
contrary to
Assumption H. Hence
THEOREM 6. parabolic projectivity never has a finite period. In other words, if of three points determining a harmonic sequence the
limit point is taken as oo in a scale
and two
consecutive points as
and
1, then the
sequence consists of
-1
1+1=2 2+1=3 3+1=4

that
is,
-l-l=-2 -2-l=-3 _3-l=-4
of zero
COROLLABY
of zero and
1.
and all positive and negative integers. The net of rationality determined by
numbers of
the
0, 1, oo consists
all
form
where
and n
are positive or
negative integers.
Proof. By Theorem 14, Chap. VI, Vol. I, the net of rationality determined by 0, 1, co consists of all numbers obtainable from and 1 by the operations of addition, multiplication, and
subtraction,
division,
(excluding division by zero).
COROLLARY 2. The homogeneous coordinates of any point in a linear planar or spatial net of rationality may be taken as integers. x v % 2 , % a are the homogeneous coordinates of a point the net, they are defined, according to Chap. VII, Vol. I, in terms of the coordinates in certain linear nets. Hence be taken they
Proof. If XQ
,
in
may
977
in the
form
or
where
m
Q,
and n l are
lf
integers.
&
If
is
the product
of their
denominators,
mx mx
inx z
mx
are integers.
The
first of
these corollaries enables us to obtain the following
simple result with regard to the construction of any point in a net of to rationality. Let ff be the harmonic conjugate of n with regard
l
H
is
n
l
and Jf_ r The sequence

5
H-%> H-fr
5)
H-v
-&
-#i>
H> &$ '"
projective
(fig.
with
and therefore must be harmonic.

a harmonic sequence
The points
H H
Q,
lf
Jfa determine
"
By
Cor. 1,
any point
of the net
of rationality is
contained in a
sequence of the last variety for some value of n. 9. Order in a net of rationality. DEFINITION. If
of
A and B are points
from //, A is said to precede B with respect to the scale x if and only if the nonhomogeneous coordiH^, Q nate (cf. 53, Vol. I) of A is less than the nonhomogeneous coordinate
R (H^H^Ha,}
different
,
A precedes B, is said to follow A. the corresponding properties of the rational numbers there follow at once the fundamental propositions With respect to the scale Jf , H^ x (1) if A precedes B, does not precede A] (2) if A
of B. If
From
precedes
and
B precedes
R (H
C,
distinct rjoints of
77"
then A precedes C (3) if A and B are V then either A nrecedes B or B nTftnedes A.

;
The use of the properties of numbers in the argument above and in of geometry is dependent analogous cases does not imply that our treatment on analytical foundations. Every theorem which we employ here is a logical
consequence of the assumptions A, E, H alone. The argument which is involved in the present case may be stated as follows The coordinates relative to a scale //, v m of the points
:
-.,
H_ v H_ v
H H H Hv H
Q, 2,
harmonic sequence, when combined according to the rules for addition and multiplication given in Chap. VI, Vol. I, satisfy the conditions which are known to characterize the system of positive and negative integers (including conditions (the axioms of the system of positive and negazero). From these
of a tive integers) follow theorems which state the order relations among these the order relations among the rational integers, and also theorems which state
numbers, the latter being defined in terms of the integers. But by Theorem 6, r Cor. 1, the rational numbers are the coordinates of points in R(T fl 1 fl M ).
r
Hence the points of R (////!//) satisfy the conditions given above. It would of course be entirely feasible to make the discussion of order in a
jt.innolit.ir
witlirmt +!io IXSQ
r>
Two
subsets, [A]
ffx ) constitute a cut (A,
B]
nd only
3
if
they satisfy the

.//, is
net except
Hm every point of [A] prein [A] or in [B] such cedes every point or |y;j. it there is a point that every point of [A] distinct from precedes it and every point of follows it, the cut is said to be closed and to have [B] distinct from
as its cut-point;
in [A]
otherwise the cut
is said to
be open.
The
class
[A]
is said to
and [B] such that every and precedes every B. These sets precedes or is identical with as cut-point. Not of points are therefore a closed cut having every cut, however, is closed, for consider the set \A~\, including all points whose coordinates in a system of nonhomogeneous coordinates havsets of points [A]
With R (HÎT^H^} determines two
be the lower side and [B] to be the upper side of the cut respect to the scale JIQ H^, llm any point 0(0 ^ff*,) of a net
,
x as the point oo are negative and the set squares are less than 2
ing
or, if positive,
[B~],
such that their
including all points whose
* An asterisk at the left of a section number indicates that the section may be omitted on a first reading. have marked in this manner most of the sections which are not essential to an understanding of the discussion of metric geometry in nhans. TTT si.nrl TV.
We
io]
CUTS IN A NET
and have
their squares greater than
2.
15
Since
coordinates are positive
no rational number can satisfy the equation
this equation is not satisfied

net.
by the coordinates
of
any point in the
The
sets [A]
and
[/*]
constitute an open cut.

,
precedes
of
DEFINITION. With inspect to the scale JT JfJ} 7/, an open cut all the points of its upper side and is preceded, by all points its lower side. A closed cut precedes all the points which its cut-
point precedes and is preceded by all points by which its cut-point is cut (A, If) precedes a cut (C, >) if and only if there is a preceded.
point
preceding a point C.
7.
THEOREM
(1)
If a cut
(A,
B) precedes a cut
(C, D), then (C,
D)
does nut precede (A, B).

(2)
If a cut
the
(A,
(A, 13} jrrc.cedfis (C,
B) is not the same as the cut (C, D}, then D) or (C, D) precedes (A, B), or loth cuts are
cut-point.
(C,
either
closed
and have
(3)
1
same
// a cut (A, B) precedes a cut

).
D} and
(C,
D) precedes a
cut
(JE,
7'
then (A, B) precedes (E, F).
above and
Proof. These propositions are direct consequences of the definition of the corresponding properties of the relation of precedence
between points.
said to be between
DEFINITION. With, respect to the scale X) a cut (A^, A 2 ) is I} Q two cuts (Bv B^ and (C^ (72 ) in case (Bv B2 ) precedes (A l} AJ and (A^, A 2) precedes (Clf C2 ) or in case (C^ C2 ) precedes If any one of these cuts is (A 1} A 2) and (A l} A 2) precedes (B^ B^. closed, it may be replaced by its corresponding cut-point in this defi,
H H H
nition.
between any point (Thus, for example, any open cut is upper side and any point of its lower side.)
of its
An
open cut (A, B)
is
said to be algebraic
if
there exists an equation,
a xn
a^z"-
+
,
an
,
0,
with integral
all
coefficients,
\_A~\
and two points A
such that the coordinates of
points of
between
AQ
and
all
B make
the left-hand
member
of this
equation greater than zero and

less
,
than zero.*
If it is
assumed
A and B make it points of [B] between that this equation has a root between A and
B this is equivalent to assuming that there exists a point corresponding to the cut (A, .B) on the line A^B but not in the given net.
For the purposes 01 geometric constructions it wouia oe sumcieuu MJ USSUJLUO the existence of cut-points for all algebraic open cuts (see Chap. IX, Vol.. I). For many purposes, indeed, it would be desirable to make the assumption
referred to on
p. 97,
Chap. IV, Vol.
I,
and which we here put down
for refer-
ence as Assumption Q.
ASSUMPTION Q. There
is
not more titan one net of rationality on a line.
But it is customary in analysis to assume the existence of an irrational number corresponding to every open cut in the system of rationals, and it is
convenient in geometry to have a one-to-one correspondence between the points of a line and the system of real numbers. Hence we make the assumption which follows in the next section.
must not be supposed that in the assumption which follows we are new points in any respect different from those already considered. What we are doing is to postulate that a space is a class of points having certain additional properties. The assumption limits the type of space which
It
introducing
we consider
it
cedxire is not parallel to the genetic
does not extend the class of points. In this respect our promethod of developing the theory of
irrational numbers.
EXERCISE
The
scale
If
H H
,
points of R (H^H^H^, together with the open cuts with respect to the v //,, constitute a set [JT] of things having the following property
:
[5] and [T] are any two subclasses of [X~\ including or a every S precedes every T, then there is either an
other 2"s and
is
all
X's and such that
which precedes
all
preceded by
all
other S's.
* 11. Assumption of continuity.

.
We shall denote the cut-point of a
closed cut (M, N) by J^ V) In the following assumption it is not stated whether the cuts (A v A n ), (B J} It n ), and (D^, Z>2) are open or closed. If
one of
them
is
closed, therefore, the corresponding one of the

P(Dl
,
symbols
Â,A)> ^BI, BS)>
ail d
z>,)
must be understood in the
sense just defined.
ASSUMPTION
C.
If every
I
points, then on one line
net of rationality contains an infinity of in one net ^((H^H^H^ there is associated
IIX a point with every open cut (A, B}, with respect to the. scale, Q H^ which is on I and such that the following conditions are satisfied: %A,B)
, ,
(1)
Â,S->
If two open
cuts (A,
B) and
(C,
D) are
distinct,
the points
ana> -%c,) are distinct;
(2) If (A^ A^J (Bl} 2) cut between two points and
and
J5
are any two cuts
and
(C^,
C2 any
)
open
of
R^^H^H^}, and
if
is
a projec-
tivity
such that
associated with some cut
then
(A v
T (J^ Cj) is a point AJ and (^ J3J.

)
(D^
2)
between
lows
.f
_//
is
called positive,
and one associated with a cut which precedes

ly Assumption 0, with an not a point of R
is
called neyativi'.
,
THEOKEM 8. The point JÂ<B) open cut (A, B} of R( 7^7/^1),

'Proof.
associated',
is
(H^H^.
The associated point could not be H^, because there are projectivities of R (H H^) which leave II invariant and change the given cut into different cuts, and therefore, by Assumption C,
change the associated point. Now suppose a point D, distinct from x but in R(ZT //1 // ), to be associated with some open cut. Since the given cut is open, there must be a point A between D and the
[D
If B is a point on the opposite side of the cut from D, A and B with respect to the scale JIQ H^ H^, both precede or both follow The transformation which changes every point of I into its harmonic has, when regarded as a transx and conjugate with regard to
cut.
formation
of
,
//j,
Hn
is
the points of ^((H^H^H^) with respect to the scale the equation

00
Zi CL
^
\Gy
where d
the coordinate of D.
It therefore transforms rational points
which follow
into rational points
which precede
it,
and vice
versa.
Hence
precede cede D.
and
if
B are transformed into two points, A' and B', which A and B follow D, or which follow D if A and B pre(2),
By Assumption C
the point
D which
D
is
associated with
an open cut between A and B is transformed into a point D' associated with a cut between A' and B'. By Assumption C (1), D' is distinct
from D, contrary to the hypothesis that

transformation.
is
a fixed point of the
THEOREM
9.
The points of ^(H^H^H^}, excluding
Hx
>
reference to the scale in which If
0,
-7^=
1,
#"
form, with a number sys,
tem isomorphic with

Proof.
the real
number system of
I,
analysis.
The
definitions of Chap. VI, Vol.
operations of addition and In that place we derived all the fundamental laws of operation, except
give a meaning to the multiplication for all points of the line L
18
FOUNDATIONS
[CHAP. I
the commutative law of multiplication, on the basis of Assumptions A and E. We have also seen in the present chapter (Theorem 6, .Cor. 1)
that the coordinates of points in
R^//^) are the ordinary rational
numbers.
remains to show that the geometric laws of combination as applied to the irrational points otC(S' JIl ffv,) are the same as for the ordinary irrational numbers.
Hence
it
The analytic
(x
(
1
definition
If
of addition of irrational
6 are
is
numbers*
may
be stated as follows:
,
y) and
(,
2,
y a ),
a and then a + b
two numbers defined by cuts the number defined by the cut

satisfies this
Vi+yd' To show that our geometric number system

8,
+*
b
condition
in
C^/^-frJ, suppose
first
and
an irrational point.
a rational point of The projective transformation

is
that a
^(H^H^H^)
(4)
x*=x+a
of points [ajj into
it.
changes the set

as [> 2
the set [xa +
a],
which
is
the
same
+ zj.
,
Similarly,
changes [yj into [y 2 + yj. Hence, it changes
(j a yg ) into (3^+ a;,, yx 4- ya), and hence, by Assumption C (2), changes b into a point determined by. a cut which lies between every Therefore b is changed into the point assopair Bj+aJj and y^-\-y z ciated with the' cut (^4- aig yx 4- y,)- But the transform of & is a 4- &. Hence the geometric sum a + b is the number defined by the .cut
the cut
K+av
?i
+ ^)'
Next, suppose both a and & irrational. The transformation (4) changes [X] into the set of irrational points [ 2 -fa], & into 6-f-,
and [yj into [y2 + a]. By the paragraph above, the cut which defines any #2 + a precedes the cut which defines any y2 + a. Hence, by^ Assumption 0(2), the cut which defines any point % +a precedes
2
the cut which defines b
+ a,
and
this precedes the cut
which defines
,
yz +
a.
Any point ^4- x2
of the lower side of the cut
a, precedes the cut defining one of the points xz above, and hence precedes the cut defining b 4- a. Similarly, any point of the upper side of this cut follows the cut defining & -f a. Hence
(x^+ x2 y^+ y ) by the paragraph.
(^ 4-
2,
y^ 4-
y2)
is
the cut defining

of
& 4- a.
Thus we have
identified
geometric addition
points in
^(H^H^H^
with the addition, of
ordinary real numbers.
be stated as follows
cuts (x v
y/j)
If
a and
let
I are positive
numbers defined by the

,
and
the
(,/;,
[x[]
Then ab
tive
is
number denned by the
be the set of positive values of x r cut (^.KZ y,?/ 2 ). If a is negaIf
and
(a
I
(
positive, ab
??)).
ab
(a)
b.
is
positive
and
(
I negative,
If
both a and
0.
are negative, ab
a)
or
i =
0,
ab =
&).
If
Consider the transformation

x'
= ax.
is
If
is
positive
[,;;J
and
rational while b
positive
and
2 ]-
irrational, this
It al so trans-
transforms
into [a#J,
[?/ 2 ]
which
into
is
the same as
JX
forms
[ayj, which is the same as [yy~\. Hence, by Assumption 0(2), ab is the number associated with
b into
ab and
(4*2> 2/1^)'
If
b,
both a and
are irrational
and
positive,
we
and [yj transformed into
[aajj, a&,
and
[a?/,],
again have [x], where, as in the
denning
analogous case of addition, the cut defining ax z precedes the cut ab, which in turn precedes the cut defining ay Moreover,
.
any x[xz precedes some ax 2 and any yj/ z follows some ay y Hence, by the same argument as in the case of addition, (x[xs y^) is the cut with which ab is associated.
,
The transformation
tJG
~-
J.
5?
changes the cut (x^ a?2 ) defining the irrational number a into the open cut ( x2 xj, which therefore defines an irrational a'. But since
,
xz may be any negative rational and a? 2 x1 may be any positive rational, the sum of a and a', which has been proved to be determined by the cut (jr 1 a? a 2 a?^, must be zero. Hence we have that (1) a a such that is the irrational a -f a = 0.
ojj
,
The transformation
is
_ a/= :(!)
.
the same as x'
1) x for all rational points.
Hence, by Assumption
C (2),
By
tive
these transformations are the same for all points of

for points of
Hence,
C^H^), (l)x = x(
on Assumptions
,
C(H H H^}.
1).
it is
the associative law of multiplication (which,
to be
if
rememis
bered, depends only
A and
,-,
E) we have,
nega-
and
b e positive,
ab
a) 6,
and
if
both,
a,
and
ab
b are negative,
= (- 1) (- a) (- 1) (- &) = (- a) (scale in
b
b).
COROLLARY. With respect to a we have ab = ba whenever a and
THEOREM
into
10.
Any
projectivity
Ha =^>,HQ,H=\ are in C (H H Hx which transforms H H and Jfm v

which
}.
Q,
into points of the
chain
(^(H^H^)
transforms any point of the chain
a point of the chain.

WP. have seen that x r
a
ax and x
=x + a,
for rational or
,s
of a, are projectivities
which change
Ha into
itself
oints of
^(Hyff^S^
I),
I/a? is a projectivity
Vol.
"
Tr
-- J
The which interchanges m and and by Theorem 9 it changes every point
into points of the chain.
-'-to
'.
a point of
C^H^).
I,
hap. VI, Vol.

'-.W e
!
it
follows that
three points of the chain by

types.
Moreover, any
of
ormation
'ransforms
>,
C (HflJlJ
Q)
H^
II
and
II',
the prod-
liemselves.
Hence, by
invariant every point it would leave invariant
mid be the
me
'ct of
identity for as II' for all points
ging
Q,
projecfcivities of
ffv ffn into the
lese transform the chain
'~"r
invariant three points of
f the chain invariant.

leaving the given points, say
o}.
Then
PHP-
activity snch that T?(S

1
leaves ff , o
H^
&
n)
invariant
andhence
leaves all points of the chain invariant, as
shown
in the proof of the
theorem.
Hence
2.
II leaves all points of the chain invariant.
COROLLARY
is
Any projectivity
,
of the chain
a,
C (H H^Hn }
Q
into itself
of the form
px(
= CX
Q -f-
dxv
where the
coefficients
are real numbers.
* 12. Chains in general. DEFINITION. If (A, B) is an open cut in any net of rationality R (K^K^K^) with respect to the scale Jf K^ n let be a projectivity transforming R into R(H and
,
(K^K^K^) Q This projectivity transforms (A, B) into a cut (C, D) in m into x If is the point m R(// //1 // ) with respect to the scale O H^,
H
as
H HJ
associated by
Assumption C with
is
(C, D}, the point II~
(X) = X'
is
called the irrational cut-point associated with (A, B).
The point X'

II'
independent
of the particular projectivity II.
For
let
be any projectivity changing (A, B} into a cut (E, F) in R (H^Hfl^} with respect to the scale Q and let Y be the point associated I} H^,
H H
,
with (E,F) and Y' = I\.'- l (Y}. Then II II'- 1 changes (E,F} into (C, D) and hence, by Assumption C (2), must change Zinto X. This
can take place only
ciated with
(.4, B~)
if
Y' = X', that
is,
only
if
the cut-point
X'
asso-
unique. By projecting any net of rationality into R (H^H^Hn ] it is shown that the cut-points associated with it satisfy the conditions stated for
is
the points associated with the cuts of R (H^H^H^ in Assumption C. Hence the theorems of the last section also apply to any chain whatever, a chain being
denned as follows
DEFINITION. The totality of points of a net of rationality R(ABC), together with all the irrational cut-points denned by open cuts with
respect to the scale A, B,
in
R(ABC),
is
called the chain defined
by A, B, C and
denoted by C(ABC). The irrational cut-points are said to be irrational with respect to R(ABC).
is
Thus we have
THEOREM
(2)
11.
(1)
The protective transform of a chain
is
chain.
Every open cut in any

If two such cuts with
net of rationality defines
a unique
the
irrational cut-point collinear with, but not in, the net.

(3)
respect to the
same
scale
and in
same
22
(4)
FOUNDATIONS
If two open cuts are homologous in a projectivity,
LCHA*. I
their cut-points
are homologous in the same projectivity.

(5)
Any
three points A, B, projectivity which transforms
into
three points of the

iffito
chain
C(ABC)
one
transforms any point of the chain

one chain containing three dis-
a point of
12.
the chain.
THEOREM
tinct points
There
line.
is
and only
of a
Proof. Let A, B,
C(ABC]
such that
that
into
which C
HJEI^HÂBC.
be the given points. They belong to the chain (H^H^) is transformed by a projectivity By Theorem 11 (5) any projectivity such
all points
ABC-^BAO
transforms
of
C(ABC)
into points of
C(ABC). But by definition such a projectivity transforms C(ABC] into C(BAC\] hence C(BAC) is contained in C(ABC}. In like man's contained in C(BAC}. Hence C(ABC}=(BAC} =
4,
B,
C to
iectivity
be points of some other chain C (PQR}. By such that* PQRA-^QPAR changes
ito points of
C(PQR). But by definition it Hence C(QPA) is contained in

changes
ictivity
C(QPA)
into
C(PQR).
In
like
manner C(QPA)
= C(PBA) =
COROLLARY.
cut in
A chain contains the irrational
cut-point of every open
any
net of rationality in the chain.

13.
THEOREM
any
have
THE FUNDAMENTAL THEOREM: OF PROJECTIVITY FOR

C,
A CHAIN. If A, B,
(A, B, C,
three distinct points
are distinct points of a chain and A', B', C' of a line, then for any projectivities giving
D)
(A
B', C',
and
(A, B, C, JD)
^ (A
1 ,
B', C', D[)
we
D'=D' r
H^ be
the two pro j activities mentioned in the theorem. C(ABC) fixed; for it leaves every
Proof. Let II,
n^
!! then leaves every point of
point of
R(ABC)
fixed,
and hence, by Theorem 11
(4),
must leave
every irrational cut-point of an open cut in R (ABC] fixed. But 1 LT^ !! is then the identical transformation as far as the points of
This theorem
may
also be stated as follows
Any
Our
protective correspondence between the points of two chains is
uniquely determined by three pairs of homologous points.

list of
r
assumptions for the geometry
of reals
may now be com-
pleted bj the following assumption of closure.
ASSUMPTION
R.
On
at least one line, if there is one there is not
more than one chain.

by Theorem 12, that every line is a chain. It also an argument strictly analogous to the proof of Theorem 5, that the dual propositions of Assumptions C and R are true. Hence
It follows at once,
follows, by
we have
THEOREM
deducible
14.
from Assumptions A,
The principle of duality is valid for E, H, C, R.
all
theorems
* 13. Consistency, categoricalness, and independence of the assumptions.
Let us
now
duction (Vol.
I) to
apply the logical canons explained in the Introthe foregoing set of assumptions.
C,
THEOREM 15. Assumptions A, E, H, number system of analysis is existent.
are consistent if the real
(%
Proof. Consider the class of all ordered tetrads of real numbers #, # 8 ), with the exception of (0, 0, 0, 0). Any class of these
,
x^
ordered tetrads such that

its
if
one
other
members
is
are given
of its members is (a a^ a z Q by the formula (ma Q ma^ ma 2

, ,
,
8)
,
all
8 ),
where
any
real
number not
zero, shall
be called a point.
tetrads satisfy
Any
two
class consisting of all points
whose component
independent linear homogeneous equations
Vo+ Vi + Va+ U x a=
3
>
V O XQ
shall be called a line.
satisfy the assumptions
+ vft + v x + v x =
z
The
class of all points

C,
and
lines so
denned
A, E, H,
(cf.
4,
Vol.
I).
THEOREM
16.
Assumptions A, E, H,
I, it
C,
R/brm
a categorical
set.
has been proved that the points of a space satisfying Assumptions A, E, P can be denoted by homogeneous coordinates which are numbers of the geometric number system of
Proof. In Chap. VII, Vol.
Chap. VI, Vol.
I.
Since
is
a logical consequence of A, E,
H,
C,
number system in question

system
of analysis.
if
is
isornorphic with the real
number
homo-
Now
two spaces S t and S 2 satisfy A, E, H,
0,
It,
consider a
in each space and let each point of S x geneous coordinate system coordinates. This correspond to that point of S 2 which has the same
correspondence
It
is
evidently such that
if
three points of S a are collinear,
their correspondents in S n are collinear.

IP
worthy
of
up in
as
many ways
17.
remark that the above correspondence may he as there are collineations of S x into itself.
set
THEOREM
H,
C,
Assumptions
A 1, A 2, A 3,
0,
1,
E 2, E
3,
E 3',
are
an independent
of
set.
Proof.
The method
Inm/iai r.onsRmipnr.A nf
1.
proving that a given assumption is not a the other assumptions was explained in the I. Suppose there is given a class of objects
of If we call each x a point and subclasses a line, then each of our
.^classes of [x\.
class
* assumptions, wiien thus interpreted, will be either true or false with but the one are If all true to this assumptions interpretation. respect
cannot be a logical consequence of the others consequence of true statements must be true. In the sequel we shall call the objects, x, pseudo-points, and the subclasses of [x\ which play the r61e of lines, pseudo-lines.
and the one
is false, it
;
for a logical
1.
plane TT together with one other point 0. the lines of TT. A 1 is false because there
ing 0.
plane.
The pseudo-points shall be the points of a real projective The pseudo-lines shall be is no pseudo-line containis
is
A2 A3
true because
it
is
true because the only sets of points A, B, C,

TT.
satisfy its hypothesis are in

is
by the ordinary projective which D, The only pseudo-plane is TT, and there
satisfied
no pseudo-space.
Hence
it is
evident that
E 0, E
C,
1,
E 2, E 3
are true
and
3' is
vacuously true.
Assumptions H,
are evidently true.
* If the hypothesis of a statement is not verified, we regard the statement as true. Following the terminology of E. H. Moore (Transactions of the American Mathematical Society, Vol. Ill, p. 489), we shall describe statements which are
true in this sense as "vacuously true" or
"vacuous."
It is possible to put any or all of the assumptions into a form such that they are vacuous for the ordinary real space. For example, Professor Moore has pointed out that Al could be replaced by the following proposition, which is vacuous for
ordinary space. A 1. Let A he a point and
B be
on B, then
A=B
a point.
If there
is
no
line
which
is
on
A and
13]
INDEPENDENCE PROOFS
The pseudo-points
25
shall be the points of a real protective three-space S g together with one other pseudo-point 0. The pseudolines shall be the lines of S each pseudo-line, however, containing 0.
8,
A 2.
Thus any two pseudo-points
are collinear with
a pseudo-plane
is
an ordinary plane together with 0\ a pseudo-space is S together 8 with 0. Hence it is evident that A 2 is false and A 1, A 3, E 0, E 1,
E 2, E 3, E 3'
points,
are true.
of
some
which
There exist harmonic sequences of pseudoare ordinary harmonic sequences. Hence
is true. By reference to the definition of a quadrangular set and harmonic conjugate it is clear (because every line contains 0) that any pseudo-point is harmonically conjugate to
Assumption
with regard to any two pseudo-points which are collinear with P.
Hence a
linear net of rationality contains all the pseudo-points of a
The operations of addition and multiplication are not unique, however, and hence the definition of order does not apply there are no open cuts, and Assumptions C and R are vacuously
pseudo-line.
true.
A 3.
The pseudo-points
,
shall be the points of a real projective
with the exception of a single point 0. The pseudo-lines shall be the lines of S g except that in case of those lines which pass
,
space S g
3 is false the pseudo-lines do not contain 0. Clearly whenever the pseudo-points A, B, C, D, are chosen so that the
through
lines
AB
and
obviously true.
DE meet in 0. A 1, A 2, E 0, E 1, E 2, E 3, E 3' are A harmonic sequence and a net of rationality of
pseudo-points can be found identical with an ordinary harmonic sequence and net of rationality on any line not passing through 0.
Hence H, C, and E are also true. E 0. The pseudo-points shall be
the vertices of a tetrahedron, and
the pseudo-lines the six pairs of pseudo-points. Thus the pseudoplanes are the trios of pseudo-points, and a pseudo-space consists of
all
four pseudo-points.
because
we may have
EA and JD=.
A1
and
A2
are obviously true.
A3
is
true
1,
E 2, E 3, E
3' are true.
H,
C,
E
false
1.
are vacuously true. There shall be one pseudo-point and no pseudo-line.

all
E1
is
and
E 2.
the other assumptions are vacuously true. There shall be three pseudo-points and one pseudo-line con-
E 3. The
are true
pseudo-points and pseudo-lines shall be the points and
lines of a real projective plane.
A 1, A 2, A 3, E 0, E
1,
2, II, C,
and
E 3'
is
vacuous.
E 3'. The
lines of a real
pseudo-points and pseudo-lines shall be the points aud four-dimensional projective space. E 3' is false and all
shall be the points
(cf.
the other assumptions are true.
lines of
H. The pseudo-points and pseudo-lines any modular projective three-space

All the assumptions
and
16,
72, Vol.
I,
and
below).
and
are true,
is false,
and C and
are vacuously true.

C.
The pseudo-points and pseudo-lines
shall
be the points and

net
of rationality in
linear nets of rationality of a three-dimensional
an ordinary real projective space.

G,
All the assumptions are true except
is false. E is vacuously true. The pseudo-points and pseudo-lines shall be defined as the points and lines in Theorem 15, the coordinates, however, being ele-
which
ments
of the
system
of
ordinary complex numbers.
All the assump-
tions are true except R,
which
is false.
the others,
Assumption 0, which is more complicated in its statement than is, however, such that neither of the two statements into
it
which
is
separated
may
:
be omitted.
This result
is
established in
the following theorem
THEOREM
tion
18. Assumption C(l) is not a consequence of AssumpC(2) and all the other assumptions. Assumption 0(2) is not a
consequence of
(1)
:
and of
If a
the other
(1) the
o
following
projectivitij
assumptions even if we add to transforms ITX into itself and
_fiT
(A,
(A,
and H^ into points of ^(H^H^H^), and transforms an open cut B) into an open cut (C, D}, it transforms the point associated with B] into the point associated with (C, D}.
Proof* (1) Any real number x determines a class x of numbers form ax-^-'b where a and & are any ration als. x is trie same as for all rational values of a and 6. Hence, if x and ax+f> y are two are either identical or mutually exclusive. Thus irrationals, x and y
of the
K K
the class of
* This
all real
numbers
falls into
a set of mutually exclusive
argument makes use of portions of the theory of classes which could not be treated adequately without a long digression. Hence we assume knowledge of the methods and terminology of this branch of mathematics without furthei
explanation.
numbers
If
or the
torm ai + b, where a and
b are rational
and 1=.
1.
take as pseudo-points and pseudo-lines the points and lines of a three-space based (as in the proof of Theorem 15) on this number
system,
If
it is
we
clear that all the
we
also take as the pseudo-points
assumptions except C are satisfied. H^ those having the 0> I}
H H
coordinates
(0, 1, 0, 0), (1, 1, 0, 0), (1, 0, 0, 0),
'R(H^H1H^)
(js,
consists of
the net of rationality
and the points whose coordinates are
pseudo-point (ai -f
Suppose now that we associate the with every cut in this net which in the ordinary geometry would determine an irrational point (aJc -f- 6, 1, 0, 0). Every point is thus associated with an infinity of cuts, contrary to
1, 0, 0),
where x
is rational.
1, 0, 0)
&,
Assumption 0(1). Moreover, the cuts with which any point is associated occur between every two pseudo-points and hence between every
two cuts
of
R(Zr
c)J
Sr -H 00 ).
i
Therefore Assumption
C (2)
remains true
in this space.
of the theorem the pseudo-points and (2) For the second half pseudo-lines shall be the points and lines of a three-space based on a commutative number system whose elements are the ordinary rational
numbers and
the ordinary
rationals
all
combination shall
open cuts in the rational numbers. The laws of .be such that addition is precisely the same as for
system, and multiplication is the same between rationals and irrationals, but different
irrationals.
number
and rationals or
between irrationals and

bers associated with
Thus the product

rule.
of the
two open
x
1
cuts will not, in general, be the
numnumber
associated with the cut given

jective transformation
by the usual
Hence
the pro-
= ax will
false.
Assumption C
(2)
must be
not preserve order relations, and On the other hand, C (1) and the
other assumptions are obviously true.

* We do not show how to set up the correspondence. The assumption that this correspondence exists is a weaker form of the assumption used by Zermelo (Mathematische Annalen, Vol. LIX, p. 514) in his proof that any class can be -well ordered. Our proof of the second part of the theorem is dependent on the validity of Zermelo's result that the continuum can be well ordered. The whole theorem is therefore subject to the doubts that attach to the Zermelo process because of the lack of explicit methods of setting up the correspondences in question.
The existence of the required new number system can be inferred from Hamel's theorem* that there exists a well-ordered set of real
numbers
(5)
flj,
a,
o8
-,,
such that every real number can be given uniquely by an expression of the form
(6)
aQ 4
a
:
^4
4 an a
in ,
containing only a finite
number
of terms, the a'a all being rational.
The ordinary
rules of combination for cuts determine a multiplication

;
table for the a's

(7)
that
is,
a set of rules of the form

ti
aft
/3's
+ â. 4 /^ +
+ pm akm
:
where the
are rational.
numbers
by the
system in general to be added or multiplied in the form (6)

rules for addition
The laws of combination for the number may now be stated as follows Express the two
;
add or multiply
and multiplication of polynomials, reducing
the result in the case of multiplication

table for the
a's.
by means
of the multiplication
JSTow suppose
(8)
we denote by
<<
,*:,-
the same set of numbers [a] arranged in a different order of the same type as (5). Such an order would be obtained, for example, by intertherefore a one-to-one correspondence
changing a 1 and a 2 and leaving the other a's unaltered. There is in. which every a { corresponds
to the a[ having the same subscript. Moreover, since the set of all a's includes the same elements as the set of all a"s, every real number
is
expressible in the form

(9)
ao 4
a^+ a a'^----z
{
(-
an a(n
new law
of multiplication,
which we
shall denote
by x
is
now
denned by setting up a multiplication table for the

the rule that
(10)
a' t
a''s
according to
a!
4 a^ H---- 4 #<
whenever
* Mathematische Annalen. Vol. "LT
n, 45ft
numbers is obtained by expressing each in the form (9), multiplying according to the rule for polynomials, and reducing by the multiplication table for the a"s.
Since the set of all expressions of the form
a
forms a
+a a +a a
1
ii
+...
fi
number system, the
set of all expressions of the
form
forms a number system isomorphic with the first. For if we let each a correspond to the a( with the same subscript, the sum of any two elements of the first number system corresponds, by definition, to
{
the
sum
of the corresponding
two elements
in the second
number
Similarly for the product of a rational by a rational or of a rational by an irrational. The product of two irrationals in the first
system.
system corresponds to the product of two irrationals hi the second, because the two polynomials in the a's are multiplied by the same
rules as the
two in the
a"a,
and are
also
reduced by corresponding
shall be distinct
entries in the respective multiplication tables.
We may
insure that the two

a's,
number systems
.
selecting the
in the first place, so that a x
= V2
by and a 2 =V3, and
then choosing the a' 'a so that a[ = a z * 14. Foundations of the complex geometry. Let us add to Assumptions A, E,
H, C the following assumption ASSUMPTION E. On some line, I, not all points

:
belong to the
same
chain.
Let P PV PV, be three points of I. The geometric number system Q determined by the method of Chap. VI, Vol. I, by the scale P P v is vP
, ,
commutative
for all the points in the chain
P 1 X) C(PP
but not neces-
sarily for other points.
However,
x' = = or
it is
clear,
without assuming the
commutativity
of multiplication, that
,
a/= or 1 #'= x +
define projectivities.
x'
a,
ax,
x'= xa
(a
= constant)
54, Vol. I; for
For
x'
this follows
from
=x+a
Let
it
reduces to Theorem
2,
Chap. VI, Vol. I; and for the other

let
(1),
two
cases, to
Theorem
4,
J be
any point
of
).
Chap. VI, Vol. I. not in C(P PIi), and

1
[X] be the
set of
all points in
PP C(P Q 1
a>
Then, by Theorem 11
the set of points
30
FOUNDATIONS
is
[CHAP.
[X+J]
with
a chain.
l
a3
This chain has no point except

if
P n
in
commo]
C(PPP ),
because,
m X+J=X'J= P
it
would follow tha

l
a> ).
X'
X= J, and thus J would be

[X+J]
to
:
a point of
C(PP P
Let us denot
the chain
by
C'.
In order
continue this argument
we need
the following assuim:
tion of closure
ASSUMPTION
I.
Through a point
I
of any chain
C
C.
of the line
and any point

of
I
J on
but not in C, there is not more than one chai
which has no other point than

let
P in common with
r>
Now
1
exist, because, for
PP K] Q C(P
different
C'. Such point be any point of I not in C(PP^P a} } or does not coincide wit Q P^J) example, the chain C (P or C'. The chain C(PJP X ) has, by Assumption I, a poin
from
P m
,
in
common with C (JfjJJ'J.g).

X' = X+J(P 1
Let
he
this poin
In case
(12)
X^ P
the projectivity
-X-
-X)
transforms
forms
P into J, X^ into M X )=C(P X P C(PPP

Q
itself,
and I
into itself.
Hence
it
tram
into
is
C^JT^), and
in particular P,
If
Cfjar/i). Hence every point c of the form X+JX", where JiTan
X"
are in [X].
X^
=P
the projectivity
(13)
X'=JX
which contains a C(P P P^) into C(PJP P is of the form JX. Thus we have
,
transforms
this case
1}
],
P. Hence,
LEMMA
where
1.
Every point of the line
is expressible
in the form
and
2.
LEMMA
in
1
C(P P P Two points A + JB and A' + JB where

are in
l
a> ).
!
,
A, B, A!,
at
PP X C(P Q
For
if
')
are identical if
and only if
A' and
1
B = B'.
B*B',A + JB = A'+ JB'

J"
implies
if
and thus
that
would be in
C(P PP
1
a,}
and
B = B',
J= (A -A)(B~ B'}~
it
implies direct!
A=A'.
14]
COMPLEX SPACE
of
31
l
Each
forms
the projectivities
1
a>
X'=(J^
1
C(PP P )
J)X and X' = X(P - </)

Hence,
if
transof
into
C(P (P -J)I^.
C(P P I^
1
be any point
where A"
is also in I) it
>
).
By
the distributive law (Theorem
5,
Chap. VI, Vol.
follows that
A-AJ=A"-JA".
By
(14), this reduces to
By Lemma this we can
2, it
follows that
A-JA' = A"-JA". A=A" = A'.
Hence
AJ=JA. From
deduce, by the elementary laws of operation,
= AC + AJD + JBC + JBJD = CA + CJB + JDA + JDJB + JB)

Hence the geometric number system determined by any
is
scale
on
Since chains are transformed into chains by any protective transformation, it follows that the geometric number system determined by any scale on any line in a space satisfying A, E,
commutative.
H, C, R, I satisfies the commutative law by Theorem 1,
of multiplication.
Hence,
THEOREM 19. Assumption P Assumptions A, E, H, C, R, I.

in the form
(15)
is
satisfied in
any space
satisfying
Since every point in the geometric
number system
is
expressible
A + JB, we B
are in
have
JÂ^JB^
and
where
C (J?J?^,). Thus J is
one of the double points
of the involution
(16)
XX'- %B (X+X')-A =
X C(PP P
1
0,
which transforms which are

tively into
itself.
Any two points of C(PP^) conjugate in this involution may be transformed projecP and .g by a transformation which carries C(PP P ) into
)
into
itself.
a>
This reduces the involution to
J-LJ.U.OU
double points are not in C(P PI^). The Q now reduces (17) to vr' P
1
transformation A'
=V
.-/AT'
AA =
x[
and
tlius transforms J" to a point satisfying the equation
J-=-P.
Hence we have THEOREM 20. The
Assumptions A, E, H, C, R, I system, of analysis, i.e. with z i ~ 1 and a and b are real.

*15. Ordered projective
protective spaces which
spaces
geometric nu.mlcr system in any space satisfying is isomorphic with the complex number the system of numbers a -fib, where
spaces.
There
is
an important class of
may
be referred
to as the ordered
projective
and which are characterized by the Assumptions S given below. This class of spaces includes the rational and real projective spaces and many others. The set of assumptions, A, E, S, is not categorical but it may be made so by adding a suitable continuity assumption or by some other assumption of (.'Insure. These assumptions introduce a new class of undefined elements
called senses,* in addition to the points and lines which are the undefined elements of Assumptions A and E. The senses are denoted by symbols of the form S(ABC), where A, B, C denote
points.t
1.
For any
three distinct cullinear points A, B,

three distinct collinear points there
C
is
there is
a,
sense
S(ABC). S 2. For any
not
more than
one sense S(ABC). 53. S(ARC)= S(BCA). 54. S(ABC)* S(ACB). So. If S(ABC)=S(A'B'C') and S(A'B'C'}^ S(A"B" C"} S(ABC} = S(A"B"C"}. = then S 6. If S(ABO) X(BCO), S(ABO) = S(ACO). S 7. IfOA and OB arc distinct lines, and 8(OAA ) =
1
then
OBBt
then
StfWB = S(fWB.
t<
S(OAA
and
* Sets of assumptions m->re nr Itss r.-late.l these have been given bv A T? Schweitzer, American Journal .f MatheiiiJUirs. Vol. XXXI, p.Sfif), and \ N Wl*> head, The Axioms of Project! vt- GffiTiu-try. Camlirirlw Tracts. Camhridê "IQOB
f
With
II.
respect to the intuitional basU
of these
assumptions, cf '"flk
.
Chap.
'=-.
6-12
It'
below in
>S(A/>C) be identified with, the sense-class which is discussed 1 9, Chap. II, it will be seen that S 1 and S 2 are immedi,
ately verified and 8 M,
S 7 reduce
to
Theorems 2-6, Chap.
II.
This
real
shows that the assumptions S are

protective space.
satisfied
by a rational or a
These assumptions are capable, as
is
shown
in Chap. II, of serving
as a basis for a very complete discussion of geometric order relations.
Assumption P
is
not a consequence
of
A, E, S alone.
EXERCISES
Prove tli at Assumption II is a consequence of A, E, and S. 2. Prove that with a proper definition of the symbol < (less than) the geometric number system in an ordered projective space satisfies the following
1.
conditions
(1) If a
(2) If a (3) If a (4) If a (5) If
(G) If (Cf. E. V.
and b are distinct numbers, a < b or b < a. < b, then a jt b. < b and b < c, then a < c. < b, there exists a number, x, such that a < x and x<b. < a, then b < a + b for every b. < a and < 6, then < b.
Huntington, Transactions of the American Mathematical Society,
Vol.
VI
(1905), p. 17.)
3. Introduce an assumption of continuity and with this assumption and A, E, S prove Assumption P. 4. Prove that P is not a consequence of A, E, S alone.
* 16.
Modular projective
space satisfying Assumptions

are projective.
of points, every
Hence, if such sequence contains an infinity
We have seen ( 7) that, in any and E, any two harmonic sequences one harmonic sequence contains an infinity
spaces.
of points,
and
by
8 these points are in one-to-one reciprocal correspondence with
the ordinary rational numbers. On the other hand, if one harmonic sequence contains a finite number of points, every other harmonic sequence in the same space contains the same finite number of
points.
Hence the
two
classes
and E fall into spaces satisfying Assumptions and those satisfying those satisfying Assumption
the following:
ASSUMPTION H. If any harmonic sequence
exists, at least
one con-
ing
H nonmodular.
Theorem
5,
It follows, just as in
that the principle of duality
is
true for any
modular
space.
invariant point.
n be Let II be any parabolic projectivity on a line, and let If be any other point of the line, the points J/j,
its
form
harmonic sequence, by
definition.
If this is to
contain only
a finite
number
If
of points,
),
there
must be some
positive integer
n such
that !!*(#)
= n m (//
where
is
zero or a positive integer less
than
n.
m = k,
we have
and hence
rP = 1.
the harmonic sequence are contained in the set
Hence
all the points of
In case k
integers,
is
not a prime number, that

7c
is, if
there exist two positive

Jc
k 1}
different
from unity such that
=k
&a
let
us con-
sider the parabolic projectivity
Ii\ The
points
a harmonic sequence. Since any two harmonic sequences contain the same number of points, it follows that the given sequence could not have contained more than k^ points. In case
satisfy the definition of
the same argument shows that the given harmonic sequence could not contain a number of points larger than either factor. This process can be repeated only a finite number of
breaks
up
into
two
factors,
times and can stop only
when we
arrive at a
prime number. Hence
we have
THEOREM 21. The number of points in a harmonic sequence is prime. The points of a harmonic sequence may Je denoted by
77" TT / 7T \ TTP )> '"' ô' LL (
where II
is
projectivity is
a parabolic projectivity. a prime number.
The period, p, of any parabolic
With
reference to a scale in
which Jf
is oo,
point of the harmonic sequence

x'
0, II (ITQ ) 1, and the limit II has the equation
= x-\-l.
harmonic sequence are
Hence the coordinates
of the points in the

0,
1,
2,
p-l,
respectively,
IP =
1,
where 2 represents 1 + 1, 3 represents 2 + 1, etc. Since we must have that p = 0, p + 1 = 1, np + Jc = k, etc. In other
words, the coordinates of the points in a harmonic sequence are elements of the field obtained by reducing the integers modulo p, as
explained in
72, Vol.
I.
By Theorem
14, Chap. VI, Vol.
I,
the net of rationality determined
by the points whose coordinates are 0, 1, oo consists of the point oo and all points whose coordinates are obtainable from and 1 by
the operations of addition, subtraction, multiplication, and division (except division by zero). Since all numbers of this sort are con-
tained in the set

0,
1,
...,i-l,
we have
THEOKEM
22.
modular space
is
The number of points in a net of rationality in a p + 1, p being a prime number constant for the
space in question.
Obviously, if Assumption Q ( 10) be added to the set A, E, H, the number of points on any line must be p 1, p being prime. shall be called a rational modular space. space satisfying A, E,
The problem of finding the double points of a projectivity in a rational modular space of one or more dimensions leads to the consideration
of
to the relation
modular spaces bearing a relation to the rational ones analogous which the complex geometry bears to the real geometry.
The existence
of such spaces follows from the considerations in Chap. and n ). The geometric number systems IX, Vol. I (Propositions 2 for such spaces may be finite* (Galois fields) or infinite.t
*E. H. Moore, The Subgroups of the Generalized Finite Modular Group, Decennial publications of The University of Chicago, Vol. IX (1903), pp. 141-190 L. E. Dickson, Linear Groups, Chap. I. t L. E. Dickson, Transactions of the American Mathematical Society, Vol. VIII
;
*/-.
IT,
S OO
17. Recapitulation. The various groupings of assumptions which we have considered thus far may be resumed as follows: A space
satisfying
Assumptions
A,
E
P
is
is
is is
a general projective space

a proper projective space;
A, E,
A, E,
A, E,
H H
Q Q
in
t
a nonmodular projective space a modular projective space

; ;
A, E, S A, E, H,
A, E, H,
is is
is
.
an ordered projective space a rational modular projective space a rational nonmodular projective space
;
A, E, H, C,
is
,. a real projective space
A, E, H, C, E,
or A, E,
I\
J
f is
a complex projective space,
The
sets of
first
six sets of
are, categorical.
The
set of
assumptions are not, and the remaining ones theorems deducible from any one of these
a projective geometry, and the various
assumptions
is called
geometries may be distinguished the corresponding spaces.
by the adjectives applied above
to
CHAPTER
II

18. Direct
and opposite
projectivities on a line.
In
lt
9 a point
if
was said
nate of
to precede a point
in this scale
relative to a scale
PP P x
Q,
the coordi-
was
to
less
than the coordinate

of
of B.
Supposing
the coordinate of
to be a
and that
changing
(1)
P to A
and
P
:
x'
B and = (b
leaving
a}x
P + a.
of
the projectivity x fixed has the equation

b,
to be
In this transformation the
coefficient of
is
positive
if
and only
if
precedes B.
(2)
But the transformations

ss'=ax
the form
/3,
where a
of the
is positive,
evidently form a group.
This group
is
a subgroup
group
of all projectivities leaving

all
P x
invariant, for the latter

0.
group contains
transformations (2) for which a 3=
The group
of transformations (2) for
which a
is positive is,
by
-what
such that whenever a pair of points A and B are transformed to A' and B' respectively, A precedes B if and only if A'
we have
just seen,
precedes B'.
The discussion
of order relative to a scale could therefore
be based on the theory of this group. The order relations denned by means
ever, a special relation to the point
of this group have all, howand they can all be derived by specialization from a more general relation defined by means of a more
P x
extensive group.
this larger group,
We
shall therefore enter first into the discussion of

(
and afterwards
23) show
how
to derive the relaof "sense."
tions of
"precede" and "follow" from the general notion
The
definitions for the general case, like those for the special one, will
be seen to depend simply on the distinction between positive and

negative numbers.
projective transformation of a line
may
be written in the form
xi
fTio
ft
a w xo
+ a ux
i>
'a
QT<i
TmTYiKoT-a
r\f fli^
sumption H, the
integers.
be taken (Theorem 6, Cor. 2, Chap. I) a a^.'s may The discussion which follows is valid on either hypothesis. DEFINITION. The projectivities of the form (3) for which A > are called opposite. are called direct, and those for which A < Since the determinant of the product of two transformations (^
is
the product of the determinants, the direct projectivities form subgroup of the projective group. The same transformation (3) cannc be both direct and opposite, for two transformations (3) are identic^
only
if
the coefficients of one are obtainable

all
by multiplying them
2
from those by the same constant /?; but

dependent on the choice
of
the othf
this merel
changes A into /o A. In form, the definition
is
of the coordinat
system which
nition is
tivity
used in equations (3). Actually, however, the del independent of the coordinate system, for if a given projei
is
has a positive
with respect to one
scale, it has a positive

:
with respect to every scale. This may be proved as follows Let the fundamental points of the scale to which the coordinate
in (3) refer be
P P Pa
,
1}
and
let
Q Q Q x be
Q
,
lf
the fundamental points

y^, y^
<
any other
scale.
By
56, Vol.
Q
,
I,
the coordinates
are such that
,
of
any point
,
Qfl x the projectivity whic v P Suppose that, relative to the scale PQ, P P transforms Q Q Q I} Qx to P P x respectively has the equations :
I}
,
with respect to any scale
Q Q Qw
,
yjy^ =
fy(Q<a Q
2/0
=
0.
(4)
Thus any point

are
(aj
,
E whose coordinates relative to the scale Q) v j transformed by this projectivity to a point R' whos coordinates relative to the scale P^, P* are (y , ^).
a^)
P P
is
Since cross ratios are unaltered by projective transformations,
Hence
it
follows that if XQ
and
relative to the scale
P^, P*,,
the coordinates of any point x^ the corresponding values of y and Q y^give
are
* It is, in fact, valid in any space satisfying Assumptions A, E, S, P. The pure! ordinal theorems are indeed valid in any ordered projective space ( 15), but thoi regarding involutions, conic sections, etc. necessarily involve Assumption P alsi Of. the fine print at the end of 19.
w]
DIRECT PROJECTIVITIES
39
ly (4) are the coordinates of indicate (4) by (y Q yj T(x

,
R
Q>
relative to the scale
0) x() xj, and (3) by (x' S(xQ xj. Now a direct transformation (3) carries a point whose coordinates relative to the scale x are (X Q xj into one whose coordinates rela, l}
,
Q Q Q Qn
,
I}
Let us
PPP
tive to the
same
scale are (x'
x[),
where
(x' , x[)
= S(x
,
Q>
xj.
The
are (y
coordi,
nates of these two points relative to the scale
QQ Q Q x
I}
T(xQ
ajj)
and
Q , y[) (y'
= T(x'
y^
x[} respectively.
Hence, by substitution,
Q > 2/0 (y'
1 where T~ indicates, as usual, the inverse
of T.
The determinant
of
the transformation
TST~
is
where
is real (or rational),
and A' therefore has the same sign
as A.
Thus the
definition of a direct projectivity is independent of the
choice of the coordinate system.
This result can be put in another form which

the sequel:
is
important in
DEFINITION.
lent
Two
figures are said to be conjugate
under or equiva-
with respect to a group of transformations if and only if there exists a transformation of the group carrying one of the figures into
the other.
THEOREM
1.
direct projectivities
If two sets of points are conjugate under the group of on a line, so are also the two sets of points into
~by
which they are transformed

Proof.
any
projectivity of the line.
be a direct projectivity changing a set of points [A] into a set of points [JB], and let T be any other projectivity on the
Let
line,
and
let
and T(B)=JB', it follows that above shows that TST~ l is a

r
T(A)=A' and T(B)=B'. Since T~ (A'}=A, S(A)=B, TST~ (A')=B'. But the discussion
l l
direct projectivity.
Hence [A
and
[B
are conjugate under the group of direct projectivities, as was

39, below), the group
to be proved.
According: to the definition in
75, Vol. I (see also
EXERCISES
1.
Within the field of all real numbers the positive numbers may be define as those numbers different from zero which possess square roots. Generaliz this definition to other fields, and thus generalize the definitions of direc case determine how far the theorems on sense an projectivities. In each 72, Vol. I). order in the following sections can be generalized (cf. of The 2. projectivities which transform a net of rationality int group itself has a self-conjugate subgroup consisting of those transformations whie
are products of pairs of involutions
rationality.
is
having their double pointr, in the net c This group contains all projectivities for which the determinar the square of a rational number.
*3. Work out a definition and theory of the group of direct projectivitie independent of the use of coordinates. This may be done by the aid c 69 and 70, below). theorems in Chap. VIII, Vol. I (cf.
19.
The two sense-classes on a
line.
DEFINITION. Let
JB
Q,
6'
any three distinct points of a line. The class of all ordered* of points ABC on the line, such that the projectivities
triad
are direct,
is
called a sense-class
and
is
ordered triads in the

or to be in the
same sense-class
are said
denoted by S(A B CQ ). to have the same
Tw
sent
same
sense.
Two
collinear ordered triads not in the
sam
sense-class are said to have opposite senses or to be in opposite senset
One
sense-class
chosen arbitrarily
may
be referred to by a particular nam<
as riyht'handed, clockwise, positive, etc.f The term " sense," standing by itself,
"
might have been defined as follows
The
senses are
any
set of objects
is
with the sense classes." This

in
42.
in one-to-one and reciprocal correspondenc analogous to the definition of a vector give:
When
there
is
may
serve as the
two
senses
question only of one line, any two objects whateve for example, the signs -f and This agree
.
with the definition of sense as "the sign of a certain determinant." Whe: dealing with more than one line, it is no longer correct to say that there ar
two senses
* "
there are, in fact,

is
two senses
for each line.
a logical rather than a geometrical term, just as in the defi nition of "throw" ( 23, Yol. I). It i^ a device for distinguishing the elements o a set. For example, when we say that cannot be transformed into b; any transformation of a given group, it is a way of saying that the group contain no transformation changing into A, B into C, and C into B.
Order," here,
ABC
ACS
A partial list of
: ;
would include
VIII, p. 248
references on the notion of sense in one and more dimension; Mobius, Barycentrische Calcul, note in 140 Gauss, Werke. Vol von Staudt, Beitrjige zur Geometric der Lage, 3, 14 Study, Archr
; ;
19]
SENSE-CLASSES
is
41
When
there
one adopts, as we do, the symbol S(ABC) to stand for a sense-class, no occasion for attaching a separate meaning to the word " sense." Tt
in the
of the calculus.
may be
regarded as an incomplete symbol,* like the

2.
dx
dx
THEOREM
then
If the ordered triad

Jt ('
l)
ABCis in
the sense-class
S(AHC)=ti(A
o ).
If S(AIi(,')
= S(A'B'C")
S(A Q B CQ ), and S(A'B'C')
8(A"B"C"),
G'
then ti(AHC)
= ti(A"B"C").
A"
FIG. 6
ABO
form a group.
3.
B"
C"
A'
B'
Proof. Both statements are consequences of the fact that the direct
projectivities
THEOREM
then
If S (ABC)
3=
S(A'B'C') and S(A'B'C')^ S(A"B"C")
S(ABC) = S(A"B"C").
0"
C'
B'
ABC
FIG. 7
A"
B"
A'
Proof.
opposite.
of
If
S(ABC)^
Hence the theorem
S(A'B'C'}, the projectivity ABC-Â'B'C' is follows from the fact that the product
two opposite
projectivities is direct.
sense-classes
COROLLAKY. There are two and only two
on a
line.
THEOREM
4.
If A, B, C are
=
distinct
collinear points,
S(ABC)
= S(BCA)
Then
and S(ABC]
S(ACB}1
(1,
1), (1, 0), (0,
Proof. Let A, B,
be taken as
1) respectively.
is
invariant.
an opposite projectivity interchanging Hence X(ABC) S(ACB). In

=
that
S(ACB}^X(BCA}.
'
It follows,
and C and leaving A manner, we can prove by Theorem 3, that S(ABC)

like
= S(BCA).
* The term incomplete symbol appears in Whitehead and Russell's Principia Mathematica, Vol. I, Chap. Ill, of the Introduction, together with a discussion of
its
"
logical significance.
t
This
may
be expressed by the phrase "Sense
is
preserved by even and altered
THEOEEM
Proof,
5.
If S(ABD}
= 8(BCD\
so
of
then
S(ABD) = S(ACD}.
1),
Choose the
coordinates
that _D=(0,
to
A=
(\,
0),
B=
(1, 1).
The transformation
fl!
ABD
BCD may
o
be written in
the form
-- aj"
-O
O
c
FIG. 8
B
and
(1, 0) goes to (1,
if
because
(0, 1) is
invariant
if
1).
This transformatrans-
tion will be direct
and only
> 0.
The point C, being the
form
of
(1,
1), is
(1, 1
+ a).
The transformation carrying
ABD
to
x[=(l+a}x v
which
is
direct because (1
+ a) > 0.
i
As an immediate consequence of Theorem 1 we have THEOREM 6. IfS(ABC)=S(A 1 B 1 Cl] andABCA l B 1 C{7;A'B'C'A! B' l C(,
PIG. 9
Theorems 2-6 contain the propositions given in 15, Chap. I, as Assumptions S. Theorem 6 is slightly more general than S 7 but is directly deducible from it. The developments of the following sections will be based entirely on these propositions, and hence belong to the theory of any ordered projective space, except where reference is made to figures whose existence depends
on Assumption. P
A, E, P, S.
Theorems
of the latter sort hold in any space satisfying
These propositions have the advantage, as assumptions, of corresponding to some of our simplest intuitions with regard to the linear order relations. The
reader
may
(of. fig's.
verify this by constructing the figures to which they correspond 6~9\ Each nronosition will VIP, fnrmrl t,n p.nrrfisnrvnrl to a -nrmnhftr of
then the ordered triad 123
is
said
to
have the same sense as 1'2'3'

set of all
if
and only
if
S(ABC]
of
= S(A'
is
the same sense as 123
ordered triads having called a sense-class and denoted by (123).

]$'(.:'}.
The
independent of the choice an immediate corollary of the definition that the plane and space duals of Theorems 2-6 all hold good (cf. figs. 10 and 13).
6 this definition is
It is
In view
Theorem
of the points A, B, C, A', B', C'.
By the definition of a point conic there is a one-to-one correspondence between the points [P] of the conic and the lines joining them to a fixed point 9 of the conic. We now define any statement in
C
terms of order relations among the points of the conic [P] to mean that the same statement holds for the corresponding lines [J^-P]. By
Theorem 2, Chap. V, Vol. I, it follows The independent of the choice of the point definitions of the order relations in the line conic, the cone of lines, Theorem
6,
above, together with

is
that this definition
and the cone
of planes are
made
dually.*
propositions with regard to sense are perhaps even more evident intuitionally when stated with regard to a conic or a flat pencil than with regard
The
to the points of a line (cf. figs. 10
and
11).
* These definitions are in reality special cases of the definition given above for any one-dimensional form, since the cones and conic sections are one-dimensional forms of the second degree ( 41, Vol. I) and since the notion of projectivity between one-dimensional forms of the first and second degrees has been defined in 76, Vol. I. However, at present we do not need to avail ourselves of the theorems in Chap. VIII, Vol. I, on which the latter definition is based.
21. Separation of point pairs.
DEFINITION.
Two points A and B of

same line if and by the symbol AB CD.
of the
II II
a line are said to separate two points C and = only if S(ABC] S(ABD). This is indicated
The relation AB CD implies the relations CD AB and AB D C, and excludes the relation A C BD. (2) Given any four distinct points of a line, we have either AB CD or AC\\ BD or AD \\J3C. (3) From the relations AH\\CD and A D\\ BE follows CD and ABCJD-Â'B'C'D', then the relation AD\\ CE. (4) If AB
THEOREM
II
7.
(1)
II
II
II
II
A'B
1
II
C'D
Proof. (1) If
(5)
AB
\\
CD, we have
S(ABC}
^ S(ABD),
AB
II
which, by the definition of separation, implies 2~6 we obtain successively, from (5),
DC. By Theorems
the last of which implies CD \\AB. The relation is excluded because it means S(ACB} =h S(ACD], which contradicts the second of
II
AC BD
the equations above.
Theorem 3 we have either S(ABC] * AB\\CD} or S(AJ3C) = S(ABD). In the latter case either S(ABC) 1= S(ADC) or S(ABC) = S(ADC). The first of these alternatives is equivalent to S(ACB} S(ACD} and yields AC \\JBD; the second implies S(ADC) = S(ABC} = S(ABD)
(2) By S(ABD]
the corollary of
(in
which
case
-f=S(ADB], and thus yields

(3)
AD\\C.
=
=f* S(ABD) and S(ADB] S(ADE}. S(BCA} = S(DBA), which, by Theorem 5, implies S (DBA) = S(D CA}, and thus 8 (ADB} = S (AD C). Hence, by the second hypothesis, S(ADC} S(ADE}, and therefore AD CE.
The hypotheses give S(ABC)

of these gives
The
first
II
(4)
This
is
a direct consequence of
Theorem
6.
*The properties expressed in this theorem are sufficient to define abstractly the relation Of Separation. Cf Va.ilfl,t.i Tf.flVliP. rift MnthiSma.+.innfta Vnl -nn 7ft IRS.
to
THEOEEM C and D,
Proof.
8.
If A and
are harmonically conjugate with regard
they separate
C and
D.
II
By Theorem
II
7 (2)
AD\\BC.
imply
"We also have
either AB CD or AC \\BD or ABCD-^BACD. Hence AC\\BD would
we have
BC AD, contrary to Theorem 7

contrary to Theorem 7
9.
(1).
(1)
and
BD \\ACy
Hence we must have
AD BC would imply AB CD.

II II
THEOREM
An
involution in which two pairs separate one another
has no double points.

Proof. Suppose that the given involution had the double points M, N, and that the two pairs which separate one another are A, A' and B, B' respectively. Since the involution would be determined
by the projectivity
in which,
by Theorem
8,
it
into
follow, by Theorem 6, that every ordered triad was carried an ordered triad in the opposite sense. Since the involution carries AA'B to A'AB', we should have
would
S(AA'B}
and hence
* S(A'AB')
S(AA'B] = S(AA'B'),
:
contrary to hypothesis. This theorem can also be stated in the following form
COROLLARY
1.
An
involution with
doiMe points
is
such that no two
pairs separate one another. COROLLAKY 2. If an involution is direct, each pair separates every other pair. If an involution is opposite, no pair separates any other
pair.
22.
distinct points of a line.
Segments and intervals. DEFINITION. Let A, B, C be any three The set of all points such that
S(AXC}=S(ABC}
is
called a segment
called the ends of the segment.
ends,
is
interior
ABC. The points A and C are The segment ABC, together with its called the interval ABC. The points of ABC are said to be to the interval ABC, and A and C are called its ends.
and
is
denoted by
COROLLARY COROLLAEY
1.
A segment does not

If D
is
contain
its ends.
2.
in
ABC,
then
COROLLARY
3.
If
is
in
ABC,
then
and D
are not separated
and
C.
THEOREM
of which
Proof.
nionically.
10.
If A and
B are any two distinct points of a

and
also two
line, thi
are two and only two segments,
and only two
intervo
and
B
is
are ends.
Let
If
C and
any point
be two points which separate A and B h of the line distinct from A and J3, eitl
S(AXB)
or
= S(ACB)
ADB.
In one case
is
in
AGB, and
in the other case in
Either of the two segments (or of the two interva whose ends are two points A, may be referred to as a segment j
DEFINITION.
an interval AB}. The two segments or intervals be complementary to one another.

(or
AB
are said
COROLLARY. If A, B,
line consists
are
any
and
three distinct points of

to
line,
of the three segments complementary

C.
ABC, BCA, CA
one
together with the points A, B,
Proof.
tions
Any AC\\BX
point
or
X distinct
II
from A, B,
satisfies
of the re
AB CX
,
or
,
AX\\BG.
THEOREM 11. If A v A 2 A n is any set of n(n> 1) distinct poii of a line, the remaining points of the line constitute n segments, ea A n as end points and of which has two of the points A^ A 2 two of which have a point in common.
,
Proof.
The theorem
1 0.
is
true for
2,
by Theorem
If
is,
Suppose
it
true for ti'k.
+1
points are given, the point

for the case
Ak +
Ic,
by the theorem
on
one of the k segments determined by the other Je points, say on the segment whose
ends are
and
A5 By
.
the corollary to
,
Theorem 10, this segment consists of A k+l together with two segments whose ends
are respectively for
At + v A
and
= k+1
if
valid for
n=
A k + v Ay Hence
Jc.
the theorem
is
is vaJ
'
Hence
the theorem
established
mathematical induction.
DEFINITION.
finite set of collinear points,
(i
= l,--,
n), is
set
of points in the order {AÂ Z
A^
is
also
in the orders
{A.2 A 8
4,4} and {A nA n _,
A^}.
EXERCISES
1.
If
2. 3.
A CUBE, then CD BE. The relations A B CD, AB\\ CE,AB\\DE are not possible simultaneously. two Any points A, B are in the orders {AB} and {BA}. Any three
II
AR CD
and
II
II
collinear points are in the orders
{ABC}, {ACB}, {CAB}.
Linear regions. The set of all points on a line, the set of all points on a line with the exception of a single one, and the segment are examples (cf. Ex. 1 below) of what we shall define as linear regions
23.
on account
of their
analogy with the planar and spatial regions con-
sidered later.
DEFINITION. region on a line is a set of collinear points such that (1) any two points of the set are joined by an interval consisting entirely of points of the set and (2) every point is interior to at least
one segment consisting entirely

to be convex
if it satisfies
of points of the set.
A region is
said
also the condition that (3) there is at least
one point of the line which is not in the set. DEFINITION. An ordered pair of distinct points AB of a convex region R is said to be in the same sense as an ordered pair AB' of R
and only if S(ABA< ) S(A'B'A ai ), where A* is a point of the line not in R. The set of all ordered pairs of R in the same sense as AB is denoted by S(AB) and is called a sense-class. The segment compleif
mentary to
val
is
AA X B
is
called the segment
called the interval
AB.
set of points of
AB. The corresponding interR is said to be in
the order [AÂ Z

is
n}
if
they are in the order
{AA
A n A n }.
If
separated from
to
C,
and B, C is between A and B with respect to follow R. If S(AB) = S(CD), then C is said to precede D, and in the sense AB. If there is a point Bx other than A*,, which is not in the convex
A* by
region R, the sense
S(ABA^)
is
the same as the sense
S(ABB
ai
}>
and
the segment
AA^B is
the same as the segment
AB^B. Hence
48
[CHAP. II
THEOREM 13. For a given convex region R the above definition has the same meaning if any other point collinear with R but not in R be
substituted for
Am
1.
COROLLAEY
If
8 (AS)
= S(A'B'} & S(A'B'}
and S(A'B'} and S(A'$]
= S(A"B"),
then
S(AB] = S(A"B"}. COROLLARY 2. If S(AB)

S(AB] = S(A"B"). COROLLARY 3. 8 (AS)
COROLLARY
4.
* S(A"B"},
then
If S(AB}
* S(BA). = S(BG],
then
S(AB)
= S(AC).
These corollaries are direct translations of Theorems 2~5 into our

present terminology.
Theorem 7
:
translates into the following state-
ments in terms
of
betweenness
THEOREM
and
or
0.
is
14.
(1)
If C
is
between
A and B,
then
is
not between
(2)
If three points A, B,
are distinct,
is
between
and
E.
A B
between
A
"
and
and
A and C or A is between C and B. (3) If C is A is between B and E, then C is between B and
between
precede
Theorem 7 translates into the following statements in terms " and " follows."
15.
(1)
of
THEOREM
precede
in this sense.
(3) If,
G.
If C precedes B in the sense AC, then B does not (2) In the sense AC, either B precedes C or C
precedes B.
in the sense
AB, A
precedes
C and
E precedes A,
AB
then
E precedes
DEFINITION. If A and are any two points of a convex region R, the set consisting of all points which follow A in the sense is called the ray AB. The point is called the origin of the ray. The
ray consisting of all points which precede A in the sense AB is said to be opposite to the ray AB. The set of all points which is sometimes called the precede A in the sense prolongation of
AB
the
segment
AB
beyond A.
EXERCISES
region on a line is either a segment or the set of the line with the exception of one point.*
1.
A convex
all points on.
2. If three points of
the order
3.
{CBA} but not
a convex region are in the order {ABC}, they are in in the order {ACS} or {CAB}.
4.
if A is between Between any two points there is an
In a convex region,
B and
C,
it is
between
C and
B.
infinity of points.
V*.
^JJWWOiiJg,
c*
OJQU\-1U
1
\J
uvjJJ.iJ.WUJ.wgcja.CUUO lÛul
U-1J_L|O.UCO
ILL
YYI1HJJ.1 fl
<o
IO OO,
show that the sense yl/s is the same as the sense A'B' if and only if B A is of the same sign as B' A'; also that two point pairs have the same sense if and only if they are conjugate under the group
x'
where a
> 0.
ax
+
If
b,
24. Algebraic criteria of sense.
(a
,
(>
a a ),
B = (b
Q,
6 t ),
and
(7=
(C Q , Cj)
are
any three
distinct points of the line, the transformation
(6)
changes
only
if
(1, 0), (0, 1),
and
(1, 1)
into A, B, and
respectively
if
and
p Q and p l satisfy the equations
that
is, if
Pi
With
is of
this choice of p^/p l the determinant of the transformation (6)
the same sign as
By
definition the projectivity

if
is
&i)
direct if
and only
c[]
if
is positive.
Now
A'
= (a[,
a[],
B'
= (&
',
C'
= (c'
Q
are
any three points
of
the line, and

S'
=
If S' is of the
two cases are

in
possible.
same sign
as S, the projectivities
which
(7) (8)
(1,0) (0,1)
(1,0)
(1,1)^^,
lÂ'B'C'
projectivity in
(0, 1)(1,
are both direct or both opposite,

/9)
and hence the
which
is direct.
If S' is opposite in sign to S,
one of the projectivifcies

(9)
(7)
and
direct (8) is
and the other opposite, and hence

16.
is
,
opposite.
Hence
(a' , a[],
THEOREM
'
Let
'
A = (a
a,j,
J5
= (6
bj,
C = (o
^ vj
A'
'
C Q , c[) be
collinear points.
Then S^-AEO) = S(A'J3'C')

V
if
and only
if the expressions
and
'0
1' '\
^0
l
c!
ai
have the same sign. COROLLARY 1. Three points given by the finite nonhomogeneous coordinates a,
to
1),
are conjugate
under
three points given by the finite
respectively, if
1
(c
a')
and only if (a have the same sign.

Set a
the group of all direct projectivities nonhomogeneous coordinates a' b', c', 1 b } (b a) and (at, c) (c b)(b c')
} 1
1
Proof.
a^fa Q)
=b
/b
<)
= cJc
Q,
and etjply the theorem.
COROLLAEY 2. Two points given by the finite n&iilwrnogeneous coordinates a and b are conjugate under the group of all direct projectivities
leaving the point oo of the
to the
nonhomogeneous coordinate si/ stem invariant

a'
two points given by the finite nonhomogeneous coordinates
and b'
respectively if
and only if a
and
C
Q
a'
b liaw
th same
sign.
Proof. Set a
aja^
bjb^
= 0, c = 1, aad apply the theorem.

l
THEOREM
17. A,
is
separate C,
D
B
if
and onfy
if
the
cross ratio
B (AS, CD)
Proof.
negative.
By
the last theorem, A,
separate
C,
JD
if
and only
d
if
and
are opposite in sign.
k
these
But the quotient of
same sign as Tb(AB, CD] (cf. p. With the aid of this theorem the proof much more simply than in 21.
the
two expressions has 165, Chap. VI, Vol. I).

of
Theorem
can be made
18. The pints of space fall into two classes sttch that two points Olt Oz of the same class are not separated t>y th& points in which, the line of differ* Oj02 meets the planes a and (3, while two jioints 0, ent classes are separated by the points in which the t-im meets a and ft.
25. Pairs of lines
and
of planes.
THEOREM
not on either of two planes
a and
/3
OP
Proof.
By
<x8 are separated
the space dual of Theorem 10 the planes of the pencil by a and B into two segments. Let \O~] be the set
JU.UU
UJJ.
the line a 13.
The two planes (o rr of the pencil a IB which are on any two and P are points
and
separated by a and ft. Hence, by Theorem 7
and 20, the points in which the line OP meets a and ft are
separated by
In.
and P.
FIG. 13
like
manner, any
two points Jt 2 determine with the line a/3 a pair of planes (or a single plane) not separated by a and ft, and hence the line 0^0Z meets a and ft in points (or a single point) not separated by O l and 2 By the same reasoning, any line P^PZ meets a and ft in points (or a point) not separated by P and P z
.
lL
COROLLARY
plane which
m are two coplanar lines, the points of the If are not on lor m fall into two classes such that two points
1.
I
and
0^ O2 of
the
same
class are not sepa-
rated by the points in which the line 0^0Z meets I and m, while two points 0,
P of different
the points in
classes are
separated
I
"by
which
2.
OP meets and m.
is
COROLLARY
pair of classes
the
There
only one
FIG. 14
lines).
[0]
and [P]
conditions of the
satisfying above theorem
(or its first corollary)
determined by a given pair of planes (or
points in different classes (according to Corollary 1) relative to two coplanar lines are said to be separated by the two lines otherwise they are said not to be separated by the lines.
DEFINITION.
Two
points in different classes (according to Theorem 18) relative to two planes are said to be separated by the two planes; otherwise they are said not to be separated by the planes.
Two
EXERCISES
1. If
^ and
are
two coplanar
lines
and
any point
of their
common
plane, all triads of points in a fixed sense-class
S l on
/j
are projected from
into triads in a fixed sense-class

of the plane,
it is
S 2 on Z 2 (Theorem 6). If is any other point by ^ and 12 if and only if triads in the sense separated from into triads in the sense S' 2 S^ are not projected from in terms of the sense of pairs of points in can be stated also This problem the region obtained on l { or 1 2 respectively by leaving out the common point.
it
30. In the form stated in Ex. 1 in this form is generalized in has the following generalization. 2. If ^ and lz are two noncoplanar lines, and o is any line not intersecting them, all triads in a fixed sense *S\ on l^ are axially projected from o into
The theorem
triads in a fixed sense

l
52 on
:
(Theorem
6).
The
lines not intersecting
^ and
fall
into two classes
triads in the sense
S2>
those by which triads in the sense S i are projected into and those by which triads in the sense S^ are projected
into triads in the sense opposite to

3.
Sy
Obtain the definition of separation of two coplanar lines by two points

Prove that
if
as the plane dual of the definition of separation of

lines.
separate the lines.
State
two points by two coplanar two coplanar lines separate two points, then the points and prove the corresponding result for pairs of points
and
of planes.
26.
The triangle and the tetrahedron.

19.
THEOREM
line
If a
line
ABC meets the sides BC,
AB m which meets the segments BAfi, CB A also meets the segment AC B. A Proof. Suppose first that m
CA,
t(
not passing through any vertex of a triangle in A Cl respectively, then any other lt B^
passes through
then
B
and hence,
separate
B^ and Bz do not and C, Cl and (72 do

if
not separate the theorem
A
is
and B.
true
if
Similarly,
passes
through B^
If does not pass through A^ or V let m' be a line joining A : to the point in which meets CA. By the argument above we have
first
that
in'
meets
all three
then that
segments BA^C, CBÂ, and AC^B, and
meets them.
Let us denote the segment
IC^B by
7,
BA^C by a, and CBÂ by
&
two
and
aoove uieorem unen gives tne iniormation tnau every line wnicn meets of the segments a, /3, 7 meets the third. Any line which meets a
ft meets 7, for, as it does not pass through A or Jl, it meets either 7 or 7; hut if it met 7, and by hypothesis meets a, it would meet @. Hence the theorem gives that a, fi, 7 are such that any line meeting
two
segments meets the third. By a repetition of this argufollows that every line of the plane which does not pass through a vertex of the triangle meets all three segments of one of
of these
it
ment
the trios a/3y, a@y,
a/3<y, afiy,
and no line whatever meets

&firy,
all
three
segments
in
any
of
the trios a/37, a fiy>
a Pj-
The lines of the plane, exclusive of those through the vertices, therefore fall into four classes:
(1) (2) (3) (4)
those which meet a,
/3,
7,
7,
those which meet a, @,

those which meet a,
/3,
7,
those which meet a,
/?, 7.
No two
1?
lines
1
l^,
of the
same
class are separated "by
any
pair of the
vertices of the triangle, while any two lines joining the point ljn to the of different classes are separated by two of the lines joining lines /
the point I 1 This result
to the vertices.
is
perhaps more
intuitively striking
when put
:
into the dual form, as follows
THEOREM
20.
The points of
sides
a plane not on the

triangle fall into
of a
four
classes
such that no two points L I} L n of the same class are separated

1y any pair of the points in which the line meets the Z
L^L
FIG. 16
the
of the triangle, while any two points L lt M^ of different classes are separated by two of points in which the line LÎ^ meets the sides of the triangle.
sides
DEFINITION.
is
Any
one
of
called a triangular region.
the four classes of points in Theorem 20 The vertices of the triangle are also
called vertices of the triangular region.
The property and
of
the triangle stated in
Theorem 19 can
also ser
as a basis for a discussion of the ordinal' theorems on the tetrahedr<

for those of the (n
+ l)-poinb
I}
in ?t-space.
,
Suppose. \vc have a

its
;
teti
hedron whose vertices are
A A 2 ,A S A f
,
Let us denote
faces by
the face a 1 being opposite to the vertex A I} etc. let us dene the edge a^ being the line A { . the edges by a 12 M a^ 84 42 , 18
a2 # 8
,
4,
Each edge a., is separated by the vertices which we shall denote by cr and o^. Let
.
A
TT
i}
Aj into two segmen
through any vertex b y %' '"> 42 ancl

;
be a plane not passij the six segments which it meets may be denoti
<;
"
>
tlie
complementary segments by c^,
%,
-,
Then
as a corollary of
Theorem 19 we have
that any plane
which
nonco-
meets
three
planar segments of the

set
u'
all
>
^
the
-,
meets
the rest of
them, and, moreover,
no plane meets
^2- If
all
segments 3^, 5^,
we observe that
any plane not passing

through a vertex must meet the edges a ,
lz
FIG. 17
it
a is> a ^n three distinct u points,
becomes clear that the planes n
passing through any vertex fall into eight classes such that two plan of the same class are not separated by a pair of vertices, where
two planes of different classes are separated by a pair Under duality we have
of vertice
the faces of a tetrahedron fa two points of the same class are not sep rated ly the points in which the line joining them meets the faci whereas two points of different classes are separated ly two of it
THEOREM
21.
The points not upon
into eight classes such that
points in which their line meets the faces of the tetrahedron.
DEFINITION.
is called
Any
one
of
a tetrahedral region.
the eight classes of points in Theorem 1 The vertices of the tetrahedron a:
HJ.IU
uu
juuve uuau
is
tiie
uuuuuxuy
ui tiny uiiu 01 uiu Glasses 01 points in
Theorem 20
composed of A, B, C and three segments having the property that no line meets them all. We shall defer this discussion, however, to a later chapter, where the results will appear as special
cases of
more general theorems.

Cross ratios of points in space.
of inter-
27. Algebraic criteria of separation.
The
classes of points determined
(Theorems 18-21) by a pair
secting lines, a triangle, a pair of planes or by a tetrahedron can he
discussed by means of some very elementary algebraic considerations. As these are similar in the plane and in space, let us carry out the work only for the three-dimensional cases.
points
Suppose that the homogeneous coordinates of four noucoplanar A I} A 2> A s AI are given by the columns of the matrix,
,
(10)
&
an
and
let (X Q
, point X. matrix obtained by substituting XQ xlf x2 xs respectively for the elements of the first column in the matrix above by a^, x, 3 a^ the
, , , ;
,
x2 x 8 ) be the homogeneous coordinates of any other x^, Let us indicate by x, a z a g a^ the determinant of the
, ,
determinant obtained by performing the same operation on the second column, etc. The expressions \y, 2 a s a^ etc. have similar mean,
Y. The ings in terms of the coordinates of a point (y y^ y z y a )= following expressions are formed analogously to the cross ratios of
,
,
four points on a line
(cf.
58, Vol. I):
y
X,
(11)
tAJj
be defined analoClearly there are twelve numbers \. which could gously to these and if the notation A^ A 2 A &) A t X, Y be permuted among the six points, 720 such expressions are defined. Each number k^
; , ,
coordinates of any point be multiplied by a constant or it an six points be subjected to the same linear transformation. If Y be not upon any of the planes determined by the points A I} A z
,
A A^
s,
there exists a projectivity
the points A^
A A A
2, a>
into
and the points represented by the columns of

which
carries
(1, 1, 1, 1)
Y into
(12)
Let (JT Xj, X^,

,
s)
be the point into which (XQ x lt %2 x s )

, ,
is
carried
by
this projectivity.
By
substituting in (11)
<>, '
we
see that
\*
x,
if
~~
i, V
,
if
34
From
this it follows that
a,
a 8 &J,
\a^, %,
taken as the homogeneous coordinates with respect reference whose vertices are A A 2 , A^.
I}
to the
a g , oj,etc. could be tetrahedron of
A^,
The
line
(JT0> JTlf
2,
Zj ~ \ (1
1, 1, 1)
meets the planes determined by the four points represented by (12) = X^ \ = X^, \ = JT8 in four points given by the values X = , \
The
cross ratios of pairs of these points

.
with (XQ
X^
24 ,
z,
JTJ and
(1, 1, 1, 1) are
ratios of
Hence A & are cross XyjT,, XJX# and XjX X and Y with pairs of points in which the line joining them
z 14 ,
A;
84
meets the faces of the tetrahedron AÂÂÂ.^ and are separated By Theorem 17, the points
by the planes
and AÂÂ^ if and only if &14 is negative. They will be separated by AÂÂ^ and AÂÂ Z if and only if is negative, and by 04 & AÂyA.^ and AÂ 2 A S if and only if 3i is negative. Hence, by Theorem 21,
3
/fc
A 2 A A^
we have
THEOREM
J-
22. IVie points
respect to the tetrahedron

positive.
X and Y will be in the same Jc A,A n A A, if and only if &, <// 1284 14' 24'
a
1'
class with
Te
84
are all
COROLLARY. The eight regions determined by the tetrahedron A

, ,
AA
A^
are those for which the algebraic signs of Tc u Jc 2i k &i appear in the following combinations : (+, -f , +), (+, +, ), (+, ~, +), (-, +, +),
27]
CKITEKIA OF SEPARATION
r,
57
Recalling that
(of.
az a s
,
=
4
is
the equation of the plane
70, Vol.
I),
we
(x)
see that
if
ax
are the equations of

ratio of
two points
line
Xy
formula given above for the cross with the points of intersection of the with these planes becomes
planes, the
two
and
a(x)
(13)
P(x)
Thus two points
are in the
same one
of the
two
classes determined
by the planes a(x) and /3(x) if and only if this expression is positive. This result assumes an even simpler form when specialized somewhat
with respect to a system of nonhomogeneous coordinates. Suppose that XQ = be chosen as the singular plane in a system of nonhomogeneous coordinates; then the same point is represented nonhomogeneously by (x, y, z) or homogeneously by (1, x, has the equation represented above by a (x) =
y, z),
and the plane
be the plane x If /3 (x) written above becomes
=
,
0,
N
the expression for the cross ratio
which reduces in nonhomogeneous coordinates, when (#o x^ xz a?8 ) and a"), to (y Q Vv 2/25 2/ 3 ) are replaced by (1, x', y' z') and (1, x",
, , > ,
'
aX'
+a
y+a
4-
Hence two
points (x',y',z'} and (a",

o^se
gular plane, and
,y 4-
a 3z
y",") are separated by the sin= if and only if the numerator

For reference we shall
:
and denominator
state this as a
of (14) are of opposite sign.
theorem in the following form
THEOREM
Theorem 18,
system
23.
Ity
The two
classes
the singular
-\-
of points determined, according to plane of a nonhomogeneous coordinate

cz
and a plane ax
by
}-
+d=
are respectively the points
EXERCISES
1.
Carry out the discussion analogous to the above in the two-dim eusio: case. Generalize to n dimensions. 2. How many of the 720 numbers analogous to k u are distinct?
28. Euclidean spaces.
jective
DEFINITION.
The
set of all points of a

of those
space*
of
n dimensions, with the exception

Thus, in particular, the set
on a
siuj
(n
1)
-space S" contained in the %-space, is called a Euclidean sp(

of all
of n dimensions.
but one
of
points of a protective line is called a Euclidean line,

all
and the
set
the points of a protective plane, except those
on a single
line,
called a
Euclidean plane.
DEFINITION.
(n
1)
The pro jective

I")
(n
1) -space
S"
is
called the singul

33
-space or the (n
-space at infinity or the ideal (n
V)-spi
associated with the Euclidean space.

ideal or to be at infinity,
is
Any
figure in S
is
said to
?i-sp
whereas any figure in the Euclidean
said to be ordinary.
a Euclidean line
The ordinary points of any line in a Euclidean plane or space foi and thus satisfy the definition ( 23) of a line
convex region.
The
definitions
and theorems
of that section
m
Th
therefore be applied at once in discussing Euclidean spaces.

if A.
and
B are
"
any two ordinary
points,
we
shall speak
of "
the
s<
menfc
AJB"
first
the ray
AB"
etc.
The
tant
corollary of
if
Theorem 18
theorem
the line
yields a very simple and imp< be taken as the line at infinity, namely
THEOREM
line the
I
24. The points of a Euclidean plane which are not on fall into two classes such that the segment joining two points
class does not
same
meet
I.
and
the
segment joining two poit

is
of different classes does meet
COROLLARY. If a
of a are either on
In like
I
is
or on the
any ray whose origin same side of L

yields
a point of
I,
all poii
manner Theorem 18
25.
THEOREM
a plane
TT
The points of a Euclidean
three-space
which are not
fall into two classes such that the segment joining two poit of the same class does not meet TT and the segment joining two poii of different classes does meet IT.
* shall refer to a line, plane, or n-space in the sense of Chap. I, Vol. I, a: projeetive line, plane, or n-space whenever there is possibility of confusion w:
We
Euclidean plane, according

of
I.
DEFINITION. The two classes of points determined by a line / in a to Theorem 24, are called the two sides
The two
classes of points determined
by a plane
TT
in a Euclidean
sides of
TT.
three-space, according to
Theorem
25, are called the
two
The two
sides of
TT
are characterized algebraically in
Theorem
23.
ordered pair of rays h, k having a common origin is called an angle and is denoted by 4/i/tf. If the rays are and A C, the angle may also be denoted by 4 BA C. If the rays are opposite, the angle
DEFINITION.
An
AB
is
called a straight angle

h,
The rays
if the rays coincide, it is called a zero angle. k are called the sides of hk, and their common origin the
;
vertex of <# hk.
EXERCISES
1.
The
.hk
angle
points of a Euclidean plane not on the sides or vertex of a nonzero fall into two classes such that the segment joining two points ol
hk is not a straight different classes contains one point of h or k. In case angle, one of these two classes consists of every point which is between a point
of h
2.
and a point
Generalize
of k.
Theorem 25
to n dimensions.
Euclidean space can be characterized completely by means of a set of assumptions stated in terms of order relations. Such a set of assumptions is given below.
29. Assumptions for a Euclidean space.
It is a simple exercise,
which we
shall leave to the reader, to verify
that these assumptions are in the last section.
all satisfied
by a Euclidean space
as defined
reverse process is also of considerable interest. This consists in deriving the elementary theorems of alignment and order from Assumptions I-VIII below, and (2) in defining ideal elements and
(1)
The
showing that these, together with the elements of the Euclidean space, form a projective space. For the details of (1) and an outline of (2) the
reader
the
may
consult the article by the writer, in the Transactions of
also a note
Society, Vol. V (1904), pp. 343-384, and Moore, in the same journal, Vol. XIII (1912), p. 74. On (2) one may consult the article by E. Bonola,Giornale diMatematiche,
American Mathematical
by E.
L.
Vol.
XXXVIII
(1900),
p.
105, and also that
by
F.
actions of the
Society, Vol.
I.
W. Owens, XI (1910),
Transp.
141.
Compare
also the Introduction to Vol.
called points
This set of assumptions refers to an undefined class of elements and an undefined relation among points indicated by
The assumptions
I.
are as follows
are distinct. If points A, B, C are in the order {ABC}, they II. If points A, B, C are in the order {ABC}, they are not in
the
order {BCA}.
DEFINITION. If
A and B are distinct A
points, the segment
AB consists
of all points X. in the order
said to be between
called the interval
{AXB}] all points of the segment AB are and B\ the segment together with A and B is
the line
AB]
AB consists of A and B and all points X
{ABX}, {AXB}, {XAB}; and the ray AB consists of B and all points X in one of the orders {AXB} and {ABX}. III. If points C and D(C = D) are on the line AB, then A is on the
in one of the orders
line
CD.
IV. If three distinct points A, B, and C do not lie on the same line, and are two points in the orders {BCD} and {CEA}, then a and
point
the
E D F exists
line.
in the order
{AFB} and
such that D, E,
and
F lie
on
same
V. If
that A, B,
and B are two distinct points, and C are in the order {ABC}.
there exists
a point
C such
VI. There exist three distinct points A, B,
C not
in
any of
the orders
{ABC}, {BCA}, {CAB}.

DEFINITION. If A, B,
are three noncollinear points, the set of all
points collinear with pairs of points on the intervals called the plane ABC.
AB, BC, CA
is
VII. If A, B,
not in the
are three noncollinear points, there exists a point
same plane with A, B, and C. VIII. Two planes which have one point in common have two points in common.
distinct
IX. If A is any point and a any line not containing A, there more than one line through A coplanar with a and not meeting a.
is
not
XVII. If
the line
is
there exists
an
infinitude of points, there exists a certain

[cr]
pair of points A,
C such
that if
is
any
infinite set
of segments of
AC, having
the property that each point of the interval

is
AC
a point of a segment a, then there with the same property.*
a finite
subset,
cr
1,
<r
2,
., <rn ,
* The proposition here stated about the interval AC is commonly known as the Heine-Borel theorem. The continuity assumption is more usually stated in the form Of the "Dfirifikind Out. Avioni " Hf. T?. T)Prlolrinrl St.oMrrlroi+ nnd I'l.i.a+t/vnoltvn
.
is
are sufficient to define a three-space which capable of being extended by means of ideal elements into a projective space satisfying A, E, S. This space will not, in general, sat-
Assumptions
I to
VIII
Assumption P. If the continuity assumption, XVII, be added, the corresponding protective space is real and hence properly protective. Assumption IX is the assumption, with regard to parallel lines. Assumption VIII limits the number of dimensions to three.
isfy
Sense in a Euclidean plane. Suppose that fc is the line at Euclidean plane. Every collineatiou transforming the Euclidean plane into itself effects a projectivity on fc which is either direct or opposite ( 18). Since the direct project! vities on L form a
30.
infinity of a
group, the planar collineations which effect these transformations on I<B also form a group.
DEFINITION.
collineation of a Euclidean plane
which
effects a
direct projectivity
on the
line at infinity of this plane is said to be a
direct collineation of the
Euclidean plane.
Any
other collineation of
the Euclidean plane is said to be opposite. Let A, JB, C be three noncollinear points the class of all ordered triads A'B' C1 such that the
;
collineation carrying A, B,
is
and G to
is
A', B' }
and
C' respectively is direct,
denoted by S(ABC}. Two ordered triads of noncollinear points in the same sense-class are said to have the same sense or to be in the same sense. Otherwise they are said to have
and
opposite senses or to be in opposite senses.
A'B'C'
Since the direct projectivities form a group, it follows that is in S(ABC), then 8 (ABC) S (A'B'C').
if
a triad
THEOREM
plane.
2 6. There are two
and only two
sense-classes in a Euclidean
If A, B, and C are noncollinear points, S(ABC)
S(BCA)
S(ACB).
Proof. Let A, B, C be three noncollinear points. If A', B', C' are any three noucollinear points such that the projectivity carrying A, B, C to A', C' respectively is direct, S(ABC) contains the triad A'B'C'.
Because the direct projectivities form a group,
S(AC) = S (A'B'C').
The
triads to
which
ABC
is carried
direct all
form a sense-class,
by collineations which are not because the product of two opposite
Thus there are two and only two sense-classes. Suppose we denote the lines BO, CA, AB by a, I, c respectively and let A', B', C' be the points of intersection of a, b, c respectively
collineations is direct.
with
a, b,
I*.
The
c
projectivity carrying
ABC
to
BOA
evidently
cai
and
to
b, c,
and a respectively, and thus

(
carries
A'B'C' to B'C
and thus
is direct
19).
Hence
S(ABC)=S(BCA).
The
projectrvity carrying
is
ABC to ACB
8 (ABC)
carries A'B'C' to A'C'B',
hence
not direct
and hence
*S (ACS).
D
are on opposite sides of a
=
THEOBEM 27. Two points C and AB if and only if
S(ABC)
8(ABD).
25
This theorem can be derived as a consequence of Ex. 1,
can also be derived from the following algebraic considerations. Let us choose a system of nonhomogeneous coordinates in su
way that the
singular line of the coordinate system
is
the sam
the singular line of the Euclidean plane.

tive collineations
The group
of all pr<
itself
transforming the Euclidean plane into
reduces
67, Vol. I) to
(15)
If
=0.
to the
we change
homogeneous
coordinates for
which x
and y
= xjx^
the line at infinity has the equation X Q

y*
<)
= 0,
and
equations (15) reduce to
(16)
_ =
/yi
>
4-
4i
On
the line at infinity this effects the transformation.

x[
a1
which
is direct if and only if A 0. Let the nonhomogeneous coordinates of three points A, B, a as)> (\> ^2)> ( c i> cz) respectively. The determinant ( i>
>
cr
^ \
6
i
is
multiplied by
whenever the points A, B, C
are subjected
t(
unangeu uy an otners.
Jtience
we nave
ordered triad of points (CL I} a ), (b^ b ), (c c ) 7ms a 2 lt g the same sense as an ordered triad a' z ], (b' v b' 2), (c(, c^) if and only if (d(, the determinants # 1 a
a, I
,
,
THEOREM
28.
An
\
Acm ê same
sign.
b.
1 1
and
c[
b>
c' z
Theorem 27 now follows
as a corollary of
Theorem
23,
27.
EXERCISES
2.
(^45(7) SÂ'B'C'), the vertex of 4- hk, A a point of h, C a point of k, and A', B', C' points analogously defined for 4-h'kf. Define positive and negative angles and develop a theory of the order relations of rays through a point.
Let 4- lik be said to have the same sense as 4- h'k' if

/i is
where
3.
let us
by
P!
Let p and or be two planes of a projective space which meet in a line l x denote the two Euclidean planes obtained by leaving out of p and aand o^ respectively; and let Sp be an arbitrary sense-class in p r A1J
;
ordered point triads of
of a fixed sense-class S,r in
from from
by p and
cr if
Sp are projected from a point not on p or o- into triads <r r Any other point P not on p or cr is separated and only if triads in the sense-class Sp are not projected
into triads of S^.
*31. Sense in Euclidean spaces. The definition given above of direct transformations in a Euclidean plane, based on the concept of
direct transformations
on the singular
is
line,
cannot be generalized to
three dimensions.
because the plane at infinity is projective and, as will be proved in the next section, does not admit of a disThis
tinction
between direct and opposite
projectivities.
is
Nevertheless, the
basis of the
algebraic criterion
definition
A>
does generalize and
made the
which
follows.
With
of
reference to a
nonhomogeneous coordinate system,
of
which
the singular (n
any
(18)
the equations 1) -space at infinity, l)-space is the (n * projective collineation of a Euclidean ?&-space take the form
where the determinant a^
is
different
from
zero.
The
resultant of
* The reader may, if he wishes, limit attention to the case n = 3. We have not actually developed the theory of coordinate systems in n dimensions, but as there is no essential difference in this theory between the three-dimensional case and the n-dimensional, we do not intend to write out the details.
two transformations
of this
form has a determinant which
is
the
two transformations. product of the determinants of the

coefficients
Hince the
appear nonhoniogeneously in (18), it is clear that a selfof all transformations (1.8) is defined conjugate subgroup of the group It follows by the same reasoning as used 0. by the condition a~
>
in
18 that this subgroup is independent of the choice of the frame of reference, so long as the singular (n l)-space coincides with the
singular (n
l)-space of the corresponding Euclidean %-space.
DEFINITION.
The group
of all transformations (18) for
which
the
determinant a
>0
is
called the
lt
Euclidean ?i-space let

points
;
A A
a,
group of direct collmeations. In a -4 B+1 be n+\ linearly independent

Z A[A'
the class of all ordered (ft+l) -ads*

, ,
A' n+1 such that the

,
collineation
tively
is
A n+l into A[, A'.it A^ +l respectransforming A I} A z direct is called a sense-class and is denoted by S(A A A n+1 ). a 1
29. There are two
sense-class
THEOREM
n-space.
The
and only two sense-classes in a Euclidean of an ordered n-ad is unaltered l>y even perodd permutations.
for the three-dimensional case is typical of
of four points
mutations
Proof.
and
altered
~by
The argument
&
6
c
the general case.

(a 1}
Let the coordinates

B ),
A, B,
C,
be
a z a & ),
,
(6^
2,
(c^
z,
8 ),
(d v d^,
d a ) respectively. The determinant
(19)
is
multiplied
to a
transformation (18).
by a g whenever the points are simultaneously Hence the algebraic sign of (19)
subjected
is left in-
variant by all direct collineations.
odd permutation of the rows of (19) would change the sign no such permutation can be effected by a direct collineation. The remaining statements in the theorem now follow directly from
Since an
of (19),
the theorem that any ordered tetrad of points can be transformed by a transformation of the form (18) into any other ordered tetrad.
*32. Sense in a protective space.

linear transformations
Let us consider the group
of all
(20)
for
^=i/y^ =
J
(=<),...,)
from
zero.
which the determinant
a.,
is
different
-,xn ) (X Q homogeneous coordinates, the equations (20) continue to represent the same tnmsl urination when all the a^'s are multiplied by the same constant p; and two sets of equations like (20) represent the same transformation only if the coefficients of one are
,
If
is
a set of
proportional to those of the other.
be multiplied by p, is Hence, tj multiplied by p \a 1 and negative and n is even, we may multiply each a {j by thus obtain an equivalent expression of the form (20) for which [a,.
If
.
each a
is
n+l
i}
if
is positive.
if
is
n+l k < has no real root. Hence, however, n is odd, p is odd, a transformation (20) for which a negative is not equivIf,
{J
alent to one for
which
a,,
is positive.
Hence the
condition
a
i}
>
determines a subset of the transformations (20) if and only if n is odd. This subset of transformations forms a group for the reason given in 18 for the case n = 1.
DEFINITION. If
(ty
n is
odd, the group of transformations (20) for
which
>
is
called the group of direct colliueations in w-space.

is
This definition of the group of direct collineations

of the choice of the
independent
frame
of reference, as follows
by an argument
18. precisely like that used to prove the corresponding proposition in In a space of three dimensions, let us inquire into what sets -of five
points the set (1, 0, 0, 0), (0,
1, 0, 0), (0, 0, 1, 0), (0, 0, 0, 1), (1, 1, 1, 1)
can be transformed by direct collineations. If the initial points are to be transformed respectively into the points whose coordinates are the
columns of the matrix
(21)
the collineation
must take the form
(22)
<= Poo i' < = Po Vo + a x

Pi
Si
where the
/o's
satisfy the equations

PoÔO
+/ i a ol +
)
(23)
W" + P ^ n + P
this
determinant
is
(aâ n a22 a s
(24)
where the expressions in parentheses are abbreviations for determinants formed, from the matrix (21) having these expressions as their main diagonals. The number (24) has the same sign as
(25)
(% VV* 88
is
) (
VW*) (WuWJ Vuaa J V*u V*)>

( fl (
which
entirely analogous to the expression found in
Theorem
16.
The
transformable into the points whose coordinates are the columns of (2 1) by a direct transformation if and only
initial set of
points
is
if
(25) is positive.
This result
may be
If a
stated in the
set
form
of a
theorem as follows
of five points whose homogeneous coordinates are the columns of the matrix (21) be such that the product of the four-rowed determinants obtained by omitting columns of this matrix
30.
is
THEOREM
positive, it
other set
can be transformed by a direct collineation into any of points having the same property, but not into a set for
the analogous product is zero or negative. COEOLLABY. Any even permutation but no odd permutation of the vertices of a complete five-point can be effected by a direct collineation.
which
DEFINITION. Let A, B, C, D,
coplanar.
E be five points no
four of
which
are
The
D,
class of all ordered
A,
C,
pentads obtainable from the pentad by direct collineations is called a sense-class and is
its
denoted by
S(ABCDE}.
corollary
Theorem 30 and
now
give at once the following
and only two sense-classes in a realprojective three-space. The sense-class of a set of five points is unaltered by even permutations and altered by odd permutations.
31. There are two
If
THEOREM
we should have had that
an analogous definition of sense-class had been made in the plane, all planar collineations are direct, and hence
that there is only one sense-class in the plane. This remark, together with Theorem 3 1, expresses in part what is meant by the
proposition
The
space
real protective plane is one-sided

two-sided.
and
the real protective three-
is
6T
Although we have grounded
this
discussion
upon propositions
regarding certain groups of collineations, the notion of sense is connected with a much more extensive group. shall return to this study, which will give a deeper insight into the notions of sense
We
and of one- and two-sidedness, in a
later chapter.
33. Intuitional description of the projective plane. may assist our intuitive conception* of the one-sidedness of the real projective plane by a further
We
consideration of the regions into which a plane is separated by a triangle. These

are represented in
able into
fig.
10.
Since any
tri-
angular region is protectively transformany other, it follows that any
triangular region
may
be represented
like Region I in fig. 16.
In fig. 18 the four
FIG. 18
regions are thus represented, together with a portion of the relations among them. The representation is more complete if the two segments labeled /? are
superposed in such a way that the end labeled A of one coincides with the A of the other. This is represented in fig. 19 and may be realized in a model by cutting out a rectangular strip of paper, giving it a half twist,
end labeled
and pasting together the two ends.
FIG.
1!)
to bring the two edges labeled 18 into coincidence. This, however, is not possible in a finite threedimensional figure without letting the surface cut itself .f The twisted strip as an example of a one-sided surface is due to Mobius.J
To complete the model

fig.
it
would be necessary
ft
in
It
(fig.
has only one boundary AftCfiA. An imaginary man OP on the surface a path which is the 19) could walk, without crossing the boundary, along
section, but
* It would not be difficult to of the propositions in this give a rigorous treatment it is thought better to postpone this to a later chapter.
image
position OQ.
till lie arrived at the antipodal small triangle RtiT were to In; moved with the man without being lifted from the surface or being allowed to time.h the man, it would be found, when the man arrived at the position OQ, that the triangle could be superposed upon itself, R coinciding with itself, bul S and T interchanged.
of a straight line in the projective plane,

If a
In other words, the
boundary of the triangular region containing

the triangular region
O would
coincide with itself with sense reversed.

It is not essential that
that the figure

triangle
ORST move
is
and the point
be small, but merely continuously so that the triangle /'ST remains a never on one
RST
of its sides.
The
possibility of
making this
transformation of the figure into ORTS is not affected by joining the two
/?-edges together,
ORST
because
none of the
paths need meet the boundary of the strip. Therefore a corresponding continuous
deformation can be
jective plane.
If
made
in
the proin the

enter
we think
of the figure
ORST
projective
plane,
the
four points
symmetrically.
Thus, since
S and T
can
be interchanged by continuously
the complete quadrangle,
moving
vertices
any two
can be interchanged by such a motion, and hence any permutation of the four vertices can be effected by such a motion.
This
is
that all projectivities

direct (
intimately associated with the fact in the plane are

32), as will be proved in a later where the notion of continuous
chapter,
deformation
of
complete
in a projective plane is given formulation.
quadrangle a precise
FIG. 20 The triangle RST may be replaced by any small circuit containing 0, and it still remains true that and the circuit may be continuously deformed till coincides with itself and the circuit coincides with itself reversed. For example, the circuit may be taken as a conic section, and the projective plane
imaged
as the plane of
I,
elementary geometry plus

3, 4, 5,
a line at infinity " (see the
28 above). The ellipse I (fig. 20) may be deformed into the parabola II, this into the hyperbola III, this into the parabola IV, and this into the V. The reader can ellipse easily verify that the sense indicated by the arrow on I goes continuously to that indicated on V. The figures may be regarded as the projections from a variable center
also
introduction to Vol.
and
This deformation of an
quadrangle
UHHT
ellipse and also the corresponding one of the depend on internal properties of the surface; i.e. they are
independent of the .situation of the surface in a three-dimensional space. They are sharply to he distinguished from the property expressed by saying that the man <>/' comes back to the position (.HI, for the latter is a of
property
the space in
that
if
comes
that
if
In fact, the closely related proposition, along a straight line in a project! ve plane till he back to the position ()Q, the triangle comes back to RTS, implies a tetrahedron (e.g. P^/i.S") be deformed into coincidence with itself KO
lies.*
which the surface
the
man OP walk
RST
that two vertices are interchanged, the, other two vertices will also be interAnd the last statement is a manifestation of the theorem ( 32) that although the projective plane is one-sided, the project! ve three-space is
changed.
two-sided.
sort of
jcctive
three-space
model of the promay be
obtained by generalizing the discussion of the plane given

above. Any one of the eight regions determined by a tetra-
hedron
lent to
is
projectively equiva-
any other. Hence we pass from fig. 17 to fig. 21, which represents in full only
the relations
among the
seg-
ments, triangular regions, and

tetrahedral regions having A l as an end, or vertex. Each of
the triangles having
Az A
,
A4
FIG. 21
as vertices is represented
by two
triangles in fig. 21. Tims, in order to represent the projective space completely we should have to bring each of the triangular regions A z A a A into coincidence with the one which
In. other words, fig. 21 would is symmp.trical with it with respect to A r represent a projective three-space completely if each point on the octahedral A were brought into coinciA surface formed by the triangular regions 2 S A^ dence with the opposite point.
EXERCISE
Show that the octahedron
the projective three-space
cides with its
* E.
is
in
fig.
21
may be
distorted into a cube so that
represented by a cube in which each point coinsymmetric point with respect to the center of the cube.
Sitzungsberichte
cler
Steinitz.
Berliner
Matheinatischen
Gesellschaft,
Vol.
VII
(1908), p. 35.
CHAPTER
III

34. The geometry corresponding to a given group of transformations. The theorems which we have hitherto considered, whether in general
projective geometry or in the particular

erties of figures
to collineations.
geometry of reals, state propwhich are unchanged when the figures are subjected For example, we have had no theorems about indi-
vidual triangles, because any two triangles are equivalent under the be distinguished from general projective group, and thus are not to
one another.
On
the other hand, there does not, in general, exist a
collineation carrying a given pair of coplanar triangles into another
given pair of coplanar triangles and thus we have the theorem of Desargues, and other theorems, stating projective properties of pairs
;
of triangles.
of figures,
We
and so have dealt hardly at
have thus considered only very general properties all with the familiar relations,
ments, which
such as perpendicularity, parallelism, congruence of angles and segmake up the hulk of elementary Euclidean geometry. These properties are not invariant under the general projective group, but only under certain subgroups. shall therefore approach their
We
study hy a consideration of the properties of these subgroups. There are, in general, at least two groups of transformations to con-
with a given geometrical relation (1) a group by which the relation may be defined, and (2) a group under which the relation is left invariant. These two groups may or may
sider in connection
:
means
of
not be the same.*
We
relation
have already had one example of a definition of a geometrical by means of a group of transformations. In 19 two collinear
same sense-class if they are conjugate under the group of direct projectivities on the line. The relation between pairs of triads which is thus defined is invariant
triads of points are defined as being in the
under the group of

*
all
projectivities
18).
The group
(1) -will
from the definition
of
always be a self-conjugate subgroup of (2), as follows directly a self-con iuerate subsrrcmp. See 39. below, where the role of
Programm,* a geometry. Obviously, all the theorems of the geometry corresponding to a given group continue to be theorems in the geometry corresponding to any subgroup of the given group; and
the more restricted the group, the more figures will be distinct relatively to it, and the more theorems will appear in the geometry.
The extreme case is the group corresponding to the geometry of which is too large to be of consequence. For our purposes we restrict attention to groups
collineations,t
identity, the
of projective
theorems
we
order to get a more exact classification of narrow the Kleinian definition by assigning to the
in
and
geometry corresponding to a given group only the theory of those properties which, while invariant under this group, are not invariant
under any other group of projective collineations containing it. This will render the question definite as to whether a given theorem belongs
to a given geometry.
in a plane
Perhaps the simplest example of a subgroup of the projective group is the set of all projective collineations which leave a line
of the plane invariant. The present chapter is concerned chiefly with the geometry belonging to this group. The chapter is based entirely on Assumptions A, E, P, Q In fact,
48 depend only on A, The class of theorems which depend on assumptions with E, H 28-30. regard to order relations has already been touched on in 35. Euclidean plane and the affine group. Let L be an arbitrary but
the theorems of
.
36, 38, 39, 40, 42, 45, 46,
fixed line of a projective plane
TT.
In accordance with the definition in
28 we shall refer to lx as the
line at infinity.
The
points of 1 M shall be
called ideall points or points at infinity, whereas the remaining points
and
lines of TT shall
is
ordinary points
a Euclidean plane.
be called ordinary points and lines. The set of all In the rest of this chapter the
unmodified, will refer to an ordinary point.
term
"
point,"
when
* Cf F. Klein, Vergleichende Betrachtungen liber neuere geometrische Forschungen, Erlangen 1872 also in Mathematische Annalen, Vol. XLIII (1893), p. 63. t From some points of view it would have been desirable to include also all
.
projective groups containing correlations. t There is some divergence in the literature with respect to the use of this and the word "improper." On the latter term see 85, Vol. I.
word
DEFINITION. Any projective colliueation transforming a Euclidean of all such collineations plane into itself is said to be affine the group is called the affine group, and the corresponding geometry the a/me
;
geometry.
THEOREM
1.
There
is
one
and only one
affine collineation
transform-
ing three vertices A,

tively
B,C
is
of a triangle
to three vertices A', B', C' respec-
of a triangle.
Since
18,
Proof.
transformed into
I.
itself,
this is a corollary of
Theorem
35, Vol.
With
which
respect to
is
any system
of
nonhomogeneous coordinates
of
the singular line, any affiue collineation in the form {67, Yol. I)
may
be written
a/âp +
'
where
A=
DEFINITION.
=0.
Two
ordinary lines not meeting
36. Parallel lines.
in an ordinary point are said to be parallel to each other, and the line is also said to be parallel pair of lines is said to be parallel.
to
itself.
Hence, in a Euclidean plane we have the following theorem as a consequence of the theorems in Chap. I, Vol. I
:
THEOREM
only one line
parallel
to
2.
;
In a Euclidean plane, two points determine one and two lines meet in a point or are parallel ; two lines
a third
is
point there
line are parallel to each other ; through a given one and only one line parallel to a given line I.
DEFINITION.
parallel to
CD
and
DEFINITION.
such that the side AB is simple quadrangle BC to is called a parallelogram. The lines and are called the diagonals of the
ABCD
DA AC
BD
simple quadrangle ABCD. In terms of parallelism, most projective theorems lead to a considerable number of special cases. Moreover, since the affine geometry
is
not self-dual, theorems
which
are dual in projective geometry
A few affine theorems which are obtainable by direct specialization are given in the
may have
essentially different affine special cases.
nf ovOTfTdoa
fnllmxrinrr lief
emr? a lawria-n -rmTv^Vâ-n in
fl-ia
noTrf. oo/^f.lATl
EXERCISES
the sides of two triangles are parallel by pairs, the lines joining corresponding vertices meet in a point or are parallel.
1. If
2. If in
two projective
is
flat pencils
three pairs of corresponding lines are
parallel,
then each line
parallel to its
3.
is
With
respect to any system of

'
homologous line. nouhomogeneous coordinates in which
the singular line, the equation of a line parallel to ax
by
is
ix
+
4.
%+c
0.
A homology (or an elation) 'orrns U into a line parallel to the

5. If
whose center and axis are ordinary transaxis.
is p + I, the number of points n a Euclidean plane is p the number of triangles in a Euclidean plane is a 2 (p l) (p+ l)/6, and the latter is also the number of projective collinea-
the
number
of points
z
,
on
a projective line
ions transforming a Euclidean plane into

37. Ellipse, hyperbola, parabola.
itself.
DEFINITION.
A conic meeting L
ellipse.
n two
distinct points is called a hyperbola, one meeting it in only

it
me
point a parabola, and one meeting
in
no point an
The
Ellipse
Parabola
Hyperbola
FIG. 22
pole of
is
called the center of the conic.
jenter is called a diameter.

)f
The tangents
intersection with
are called its
Any line through the hyperbola at its points asymptotes. A conic having an
to a
)rdinary point as center is called a central conic.
EXERCISES
1.
An
2. 3. 4. 5.
.cally
a hyperbola is a central conic, but a parabola The center of a parabola is its point of contact with /.
ellipse or
is not.
No two
The asymptotes
tangents to a parabola are parallel. of a hyperbola meet at its center.
Two
6. If
44, Vol. I) of a hyperbola are harmonconjugate diameters (cf. conjugate with respect to the asymptotes. a simple hexagon be inscribed in a conic in such a way that two of
7. If a parallelogram be inscribed in a conic, the tangents at a pair of opposite vertices are parallel. 8. If the vertices of a triangle are on a conic and two of the tangents at the vertices are parallel to the respectively opposite sides, the third tangent
is
parallel to the third side. 9. If a parallelogram be circumscribed to a conic, its diagonals
meet in
the center and are conjugate diameters. 10. If a parallelogram be inscribed in a conic, any pair of adjacent sides are meet at the center of the conic. parallel to conjugate diameters. Its diagonals
let
P and P' be two points which are conjugate with respect to a conic, the diameter parallel to PP', and let Q, and Q' be points of intersection with the conic of the diameter conjugate to p. The lines PQ and P'Qf meet
11.
Let
p be
on the conic.
is such that the sides OA and OB are conjiv 12. If a parallelogram gate diameters of a hyperbola and the diagonal OP is an asymptote, then the is parallel to the other asymptote. other diagonal
OAPB
AB
two lines OA and OB are conjugate diameters of a conic which, they and B, then any two parallel lines through A and B respectively meet in meet the conic in two points A' and B' such that OA' and OB' are conjugate
13. If
diameters.
14.
l<a
Any two Any two
parabolas are conjugate under a collineation transforming
into itself.*
15.
hyperbolas are conjugate under a collineation transforming
l,
into itself.*
16. Derive the equation of a parabola referred to a
nonhomogeneous
coor-
dinate system with a tangent and a diameter as axes. 17. Derive the equation of a hyperbola referred to a
nonhomogeneoua coor-
dinate system with the asymptotes as axes. 18. Derive the equation of an ellipse or a hyperbola referred to a nonhomogeneous coordinate system with a pair of conjugate diameters as axes.
38.
as
The group of translations. DEFINITION. Any elation having an axis is called a translation. If / is any ordinary line through
is
the center of a translation, the translation
said to be parallel to
I.
COROLLARY.
line
translation carries every proper line into a parallel
and leaves invariant every line of a certain system of parallel lines. THEOEEM 3, There is one and only one translation carrying a point
a point
J5.
A
la
to
Any translation carrying A to B must be an elation with and the point of intersection of the line AB with L as center. Hence the theorem follows from Theorem 9, Chap. Ill, Vol. I.
Proof.
as axis
lation to
if
THEOKEM 4. An ordered point pair AB can be carried by an ordered point pair A'B such that A' is not on the and only if ABB' A' is a parallelogram.
1
a,
trans-
line
AB,
Proof.
Let Lv and
v be the points at infinity
on the lines AA' and
AB respectively.
AM*
the
to
1
The
translation carrying
A to A'
must carry the
line
A M.X and leave line BL X invariant.

the
is
t point the intersec-
Hence which
tion of
is is
AMX with BLX
carried to B', which.
A'
the
intersection of
A'M^withBL*. Hence
the points A' and B' to
which
tively
and
respec-
<
are
carried
by
FIG. 23
a translation are such

that
ABB A'
1
is
a parallelogram.
to A', the
and only one same reasoning shows that whenSince there is one
ever
ABB' A'
is
and
to A'
5.
a parallelogram there exists a translation carrying and B' respectively.
THEOREM
An
ordered point pair AB is carried by a translation to an

ordered point pair
A'B', where A' is on
the line
AB,
if
and
only if
Q(L X AA',
L^B'B),
of AB.
Proof.
Lx
being
the point at infinity
Let
P be
ro
any point not on the line AB, and let M* and JV" respectively be the points of interLet Q be the point of intersection of section of PA and PA' with l x
.
J>T.
Kir t.Ka loaf. f.Tionram
f.lip
f.r ft
n si
ft fvi
nn
r>.flrrviTor
A to B
of
carries
P
t
to Q,
and hence carries A'

Q,
to the point of intersection
QNa
with AB.
Hence N^,
and 5'
are collinear,
and hence we
have Q(L*AA' LB'B).
THEOREM
6.
If A, B,
C
to
are any three points, the resultant of the
translations carrying
and
to
is the
to C.
Proof,
Let
Ax
B^,
Ca
be the points of intersection of the lines
BC, GA,
A,,,
,
AB
respectively
with
lw .
Suppose
first
Bm Cm are all distinct. The the line ABX into the line BBX
,
that the three points A to changes
and the
.
B to
changes the line BB, into under the resultant of these
CBn
Hence
the line
ABm
is
invariant
two
translations.
Consider
line
now any
other
let it
f ' t
through
Bx
and
meet A A M in A' and

also let B'
OmC
be the point of
of
intersection
BC (fig.
A
25).
We
A'C* with then have
that the translation carrying to carries A! to B'
(Theorem
count of
4),
and
on
ac-
Q(A.B&, A^C'C)
5) the translation
(Theorem
carrying
to
to
carries
B'
of the
f
.
Hence the resultant
two
;
translations carries A' to C'
and
thus leaves the line A'
m invariant
that
is,
it
leaves all the lines
Since it obviously leaves all points on lm a translation (Cor. 3, Theorem 9, Chap. Ill, Vol. I). If two of the three , Cw coincide, they all coincide, and points in this case the theorem is obvious.
through
Bx
invariant.
invariant, it
is
X,
By
definition, the identity is a translation.

set
Hence we have
COROLLAEY. The
THEOREM
Proof.
7.
of all translations form a group. The group of translations is commutative.
Given two translations T and T and let A be a 2 any point, =.4' and T a (A')=B'. If B' = A, T is the inverse of T and
rl
a,
,fnî,r,->
1>l -4-
line AA', let B (fig. 23) be the point of intersection of the line through A parallel to A' B' with the line through B parallel to AA', then ABB' A' is a parallelogram, and it is obvious that 1 (B)=B'
on the
and
and
(A)=JB.
r
Hence
B',
= B'. (A)
is
T/T^) ='.
But,
by the
also,
definition of A'
Hence, in this case

let
T and T
x
are
commutative.
In case B'
on the line AA',
and
(fig.
24) be
two
points
such that A'B'QP is a parallelogram, let B be the point of intersection of A A with the line through Q parallel to AP, and let .,,
1
JJ/.a.-ZVa,
Then, since
be the points at infinity of PQ, PA, and PA' respectively. T 2 (A ')=B', it is obvious that T a (P) = Q, and hence that
T (A)=.
9
Moreover, on account
a
of
2
Q(L X AB, L X B'A'), ^^(A) = A

(^)
implies that
also,
T
1
and
T ()='. Hence T T T are commutative.

1
= ^',
and
thus, in this case
&
THEOREM
is
8.
If
OX and OY are
parallel
is
two nonparallel lines

I?
and Tis any

that
translation, there is
a unique pair of translations

to
parallel to
Proof,
OX, T 2
OY, and
T 1T =
2
T T 2 sucJi
theorem
T.
is trivial.
In case
parallel to
OX or
Y the
If
is parallel
to neither of them,
letP=T(0)
Then
and
to
let
X
l
and
be the points in which the lines through

respectively meet allelogram, and if
P parallel
OX and Y respectively.
T
x
Y and OX OX^P Y is a parto
be the translation carrying

to
:,
and T 2 the
6,
Y^
T[
it
follows,
by Theorems 4 and
that
T T 2 =T.
a
On
the other hand,
if
any translation
parallel to
to
product T[Tg carries = allelogram, But P'
if
is any translation parallel to OX, and TJ V the OY, and T[(0) =X[ and TJ(0) = Y' a point P' such that OX[P'Y[ is a parand only if X{ = l and F/= Y^ Hence
determines
T and T 2
1
uniquely.
THEOREM.
in which
lm
9.
With
respect to
a nonhomogeneous coordinate system

to the x-axis
is the
singular line a translation parallel
has the equations
Proof.
The point
into
which
is (0, 0)
transformed by a given trans(a, 0).
lation parallel to the a-axis
may
be denoted by
By Theorem
the given form
(2).
allel to
Conversely, any transformation of the type (2) leaves all lines parthe #-axis invariant and transforms any other line into a line
parallel to itself.
Hence
it is
a translation parallel to the x-axis.
THEOREM
in which
the
lw
10.
is the
With respect
to a nonhomogeneous coordinate system singular line, any translation can be expressed in

.
form
()
Proof.
x'=x+a, y'-y + S.
By Theorem
8 any translation is the product of a translation by one parallel to the y-axis. Hence it is the
parallel to the a>axis
product of a transformation of the form
by a transformation
of
the form
x'
x,
EXERCISE
Investigate the subgroups of the group of translations.
39. Self-conjugate subgroups.
Congruence. DEFINITION.
to be
Any
2
sub-
group under
G G
of
if
a group
and
G is said only if 2T2" is

1
self-conjugate or invariant*
is
an operation of G' whenever
an
operation of
and T
of G'.
The geometric significance of this notion is as follows: Suppose that two figures f\ and are conjugate under G', and T is a transs formation of G' such that ^ o T(^ ).
>
If JP
and
F
=
are
changed into
1
F{ and
(^ )
'
F' z
by any transformation
F.
Hence t
2 of G, then T2" (F[) = F

,
* These terms have already been defined in 75, Vol. I.

t
These relations may be illustrated by the

to S. Lie).
accompanying diagram (probably due
Fio. 26
3]
CONGRUENCE
J
79
nd 2T2~
.gures
_/''/
(/<'/)
l^.
Therefore,
if
G'
is
self-conjugate under G, the
G'. Hence the property of einy conjugate lender the self-conjugate subgroup G' is a property in-variant by ihe group G. Thus the theory of figures conl/'t iigate under G' belongs to the geometry corresponding to G, pro-
and 1^ are conjugate under
ided that
f
is
not a self-conjugate subgroup of any other group
protective collineations.
THEOREM
ffine
11.
The group of
translations
is self -conjugate
under
the
group.
Proof.
*
Let
L'ansf urination.
be an arbitrary translation and 2 an arbitrary affine We have to show that 2T2" 1 is a translation. If
of
ln , la
he
any point
2(P)
is
also
on
ln
Therefore, since
1
.
T
of
saves all points of

lies
invariant, so does
2T2"
The system
;
through the center of T is a system of parallel lines 2 transarms this system of parallel lines into a system of parallel lines ud hence the latter system of parallel lines is invariant under
;
JT2-
1
.
Hence
(cf.
Cor. 3,
Theorem
9,
Chap. Ill, Vol.
I)
2T2-
is
translation.
COROLLARY
1.
The group of
the affine
translations
is
self-conjugate
it.
under
ny subgroup of
COROLLARY
here exists
2.
group which contains
For any
affine collineation 2,
and any translation

a,
T,
a translation T' such that
2T=T'2 and
translation
T"
uch that
Proof.
v/
T2 = 2T".
Let
2T2~ = T'
a
and 2~ a T2
= T". By
J
the theorem, T' and
are translations.
But
= T'
nply
DEFINITION.
and 2-
2T = T'2
Two
and
T2 = T" T2 = 2T"
respectively.
if
figures are said to be congruent
they are con-
igate under the group of translations.
This definition will presently be extended by giving other condi-
40. Congruence of parallel point pairs.
The
figure consisting c
two
to
distinct points A, J3
may
be looked at in two
ways with
respec
congruence.
We
and
consider either the two ordered* point pair
AB
to
and
BA
or the point pair
AB
the
second case
AB
BA mean
without regard to order. In th same thing and AB is congmerj
BA
On
the other hand, the ordered pair
because the identity belongs to the group of translation; AB is not conjugate to th
ordered pair
lation carrying
BA under the A to B does

).
group of translations, because the tram

not carry
to
(this is
under Assumj
tions A, E,
THEOREM
12.
If
ABDC is
the
AB
is
congruent
to
a parallelogram, the ordered point pat ordered point pair CD. If the conditio
where P^
to
Q(JÂC, P,DB]
point pair
Proof.
is satisfied is
is
an
ideal point, the ordere
AB
congruent
the ordered point
pair CD.
This
is
a corollary of Theorems
Let
4 and
5.
COROLLARY
I.
A and B
pair
le
any two
distinct points
the line
and
tl
harmonic conjugate of the point at infinity of

to
AB with respet
A and
B.
Then
the
AO
case
is
congruent
to the
pair OB.
DEFINITION.
point
of
The point the pair AB. In
in the last corollary is called the mil
B = A, A
is called
the mid-point
(.
the pair
AB.
COROLLARY
AB
and
AC
2. The line joining the mid-points of the pairs of vertict is parallel to the line BC. of a triangle
ABC
Proof.
Let
Ba and Ca
let
AC respectively, and and AC respectively.
be the points at infinity of the lines AB an B^ and Cl be the mid-points of the pairs A.
of "mid-point,"
Then, by the definition
Hence the
and
lines
B C are
BQ,
BC, and B*C* concur, which means

say A,
that BJ.
parallel.
DEFINITION.
The
line joining a vertex,
of a triangle
ABC
the mid-point of
BC is
The
called a
median
of the triangle.
THEOREM
13.
three
medians of a triangle meet in a point.
40, 4ij
CONGRUENCE
Let the triangle be
81
,
Proof.
ABC]
let
A a Bx
;
infinity of the sides
BC, CA, respectively points of intersection of the pairs of lines BB*
AB
C* be the points at and let A^ Bv C\ be the

,
and
(7(7,,
CCM and AA,
AA X
and
BBX respectively
sets
(
(fig.
27).
Then, by well-known theorems on
harmonic
BB^ and CC^
the medians of the triangle and these three lines concur.

31, Vol.
I),
ABC are AA
lt
EXERCISES
diagonals of a parallelogram bisect one another; that is, a parallelogram, the mid-points of the pairs A C and coincide.
1.
The
HABCD
AB
is
BD
2.
Let a and
be two parallel
lines.
The
mid-points of
all
the pairs
where
3.
is
on a and
f
on
b are
on a
line parallel to a
and
b.
If the sides
t
AB,BC, CA
of a triangle
ABC are respectively
parallel to
the sides A'E'.
is B'C , C'A' of a triangle A'B'C', and the ordered point pair congruent to the ordered point pair A'B', then the two triangles are congruent. 4. The mid-points of the pairs of opposite vertices of a complete quadri-
AB
lateral are collinear.

5.
Let us
call this line
line
through a diagonal point
the opposite side of the diagonal

sides of the
the diameter of the quadrilateral. of a complete quadrangle, parallel to triangle, is met by either pair of opposite
in a pair of points having
quadrangle which do not pass through
as mid-point.
41. Metric properties of conies.
contains a
number
of theorems on conies
The following list of exercises which involve the congru-
82
THE AFFINE GBOUP IN THE

EXERCISES
1.
PJjAJSJfi
LCHAF.HI
The mid-points
if
etc.
are collinear
the lines
of a system of pairs of points of a conic A A', BE', CC' A', BE', CC' are parallel. The line containing
t
the mid-points is a diameter conjugate to the diameter parallel to A A'. be two points of a parabola. If the line joining the midand 2. Let
point
then
3.
of the line AB meets the conic in 0, to the pole the mid-point of the pair CP. If a line ineets a hyperbola in a pair of points H-flv and its asymptotes
of the pair
AB
is
in a pair A : A Z , the gruent to the pair

4.
HA
Z
two pairs have the same mid-point. The pair HÂ^

Z
.
is con-
The point
of contact of a tangent to a hyperbola is the mid-point of
the pair in which the tangent meets the asymptotes. a variable point of a hyperbola, 5. Let A l and A z be each a fixed and and meet one of and let 1 and a be the points in which the lines l Z
XA
XA
the asymptotes. The point pairs are all congruent.

6.
X^X^
determined by different values of
The
The
centers of all conies inscribed in* a simple quadrilateral
A BCD
are
on the line joining the mid-points of the point pairs

7.
CA and BD.
centers of
quadrangle
A BCD
conies which pass through the vertices of a complete are on a conic C 2, which contains the six mid-points of the
all
pairs of vertices of the quadrangle, the three vertices of its diagonal triangle, and the double points (if existent) of the involution in which Z is met by
the pencil of conies through A, J3, C, D. From the protective point of view, 2 according to which / is any line whatever, C is called the nine-point (or the
of eleven-point) conic
the complete quadrangle
A BCD and
the line lx
.
Derive
the analogous theorems for the pencils of conies of Types II-V (of 47, Vol. I). 8. The five diameters f of the complete quadrilaterals formed by leaving out one line at a time from a five-line meet in a point A, which is the center
tangent to the five lines. A determined, according to the last exercise, by the six complete five-lines formed by leaving out one line at a time from a six-line 2 are on a conic C
of the conic
9.
The
six points
10. The seven conies C 2 determined, according to the last exercise, by the seven complete six-lines formed by leaving out one line at a time from a seven-line, all pass through three points.
42. Vectors.
all of
Any
which, are equivalent to
ordered pair of points determines a set of pairs it under the group of translations. In
the notion of a vector.

*
order to study the relations between such sets of pairs we introduce The term " vector " appears in the literature
A conic is said to
be inscribed in a given figure
if
the figure is circumscribed
to the conic (cf.
43, Vol. I).
" the term " ordered pair of points is to be understood to include the case of a single point counted twice.
DEFINITION.
A planar field
of vectors (or vector

called vectors,
objects, the individuals of

is
which are
field) is any set of such that (1) there
(2) there is
one vector for each ordered pair of points in a Euclidean plane, and only one vector for any two ordered pairs AB and A'B'
which
are equivalent under the group of translations. vector corresponding to a coincident pair of points is called a null vector or a zero vector, and denoted by the symbol 0.
For example, a properly chosen set of matrices would be a vector field according to this definition. So would also the set of all translations including the identity; also a set of classes of ordered point pairs such that two point pairs are in the same class if and only if equivalent under the group of translations.
However a vector
field
be defined,
applications, only those properties above are actually used.
it will be found that, in most which follow from the definition as stated
system.
precisely similar state of affairs exists in the definition of a The objects in the particular number system determined for
number
a given
space by the methods of Chap. VI, Vol. I, are points, but a number system in general is any set of objects in a proper one-to-one correspondence with this
set of points.
In the following discussion has been selected, and all stat<

the
vector corresponding to the point pair is a definite object, and we shall denote it as
AB
"the vector AB"
or,
in symbols,
Vect (AB).
Since
any point
of
a Eu-
clidean plane can be carried by a translation to any other point,
the set of all vectors

as the set of vectors
is
is
the same
OA, where
a fixed
and
nition gives a
meaning
a variable point. Consequently, the following defi" " to the operation of adding any two vectors.
DEFINITION.
If 0, A,
are points of a Euclidean plane, the vector
00
is
called the
sum
of the vectors
OA
and AC.
In symbols this
is
Vect (AC). The operation of Vect (OA) indicated by Vect (00) obtaining the sum of two vectors is called addition of vectors.
An
obvious corollary of this definition
is
that
0.
Vect (AB)
+ Vect (BA) =
Hence we
define
DEFINITION".
vector Vect (AB),
The vector Vect (BA) is called the negative and denoted by Vect (AB).
are vectors, (a
of the
THEOREM
that
is,
14.
&, c
if a,
The operation of addition of vectors + &) + c = a -j- (b + c).

^[1?, i?(7
is associative;
Proo/.
definition,
Let the three vectors be 0J,
respectively
then,
by
both (Vect (0-4)+ Vect (.4.8))+ Vect (BC) and Vect (OA) + {Vect (,!#) + Vect (<?)) are the same as Vect (06'). DEFINITION. Two vectors are said to be collinear if and only if they can be expressed as Vect (OA) and Vect (OB) respectively, where
A,
are collinear points.

15.
XO, THEOREM
the vector
vectors OA and OB is OACB is parallelogram. Proof. By Theorem. 4, the vector OB is the same as the vector AC. Hence, by definition, the sum of OA and OB is OC. THEOREM 16. The sum of two collinear vectors OA and OB is a
The sum of two noncollinear
OC, where
is
such that
a,
vector
00 such
AB.
that
QÂO,
J^BC), where
is the
point at infinity
of
the line
M be
is
Proof. Let L and
two points
such that
OBML
a parallelogram.
Hence Vect (OB) =

Veot(iJf).
Then,
by
definition,
must be such that

Vect (LM)
(AC), that
that
= Vect
is,
such
is
ACML
be
the
parallelogram. Let
LM
ideal
point of intersection of the lines
OL and BM,
(7.
and
let
be the ideal
point of intersection of the lines
AL and If
such that Vect (OA) scale (cf. 48, Vol.

of the line OA,
Proof.
+ Vect (OB) = Vect(OC),

I)
then, with respect

J
to
any
in which P^
is
and
A+B = C.
Theorem
1,
Cf. Cor. 1,
Chap. VI, Vol.
I.
THEOREM is, if a and

Proof.
respectively.
17.
b
The operation of adding
vectors is
commutative ; that
are vectors, a
+b=b
-f-
Let the vectors a and

If 0, A,
be Vect (OA) and Vect (OB)
are noncollinear, the result follows from
Theorem
15,
and
if
they are collinear, from Theorem 16.
43. Ratios of collinear vectors.
By analogy with the

of points in
case of addition
we should he
vectors
led to base a definition of multiplication of collinear
upon the multiplication
49, Vol.
I.
There
are,
however, a great many ways of defining the product of two vectors, which would not reduce to this sort of multiplication in the case of collinear vectors. Hence, in order to avoid possible confusion we shall
not introduce a definition of the multiplication of vectors at present, but only of what we shall call the ratio of two collinear vectors.
DEFINITION. The ratio of two collinear vectors OA and number which corresponds to A in the scale in which J is
OB
is
the
0, J^ is
and Pn
is
the point at infinity of the line OA.
--
It is
denoted by
Vect (04) -
or
Vect (OS)
It is to
bv y
OA
OB
here defined
Vol.
I,
be emphasized that the ratio of two collinear vectors as is a number. By comparison with the definition in 56, we have at once
18.
THEOREM
If A, B,
C,
Dx
are collinear points, D*> being ideal,
Theorem
13, Chap. VI, Vol. I

19.
now
gives
THEOREM
If A
A 2 A A4 are any four collinear ordinary points,

,
s>
A
A
THEOREM
sides
20.
*^^1"^8 R (A A A A = A A 4 AA If two triangles ABC and A'B'C'

\
-
"^2"'^8
4)
are such that the
AB, BO, CA
are parallel to A'B', B'C',
C A'
!
respectively,
AB
A'B'
~~
CA BO = ~~
B'C'
C'A'
'
Proof.
B' to
carries A' to A carries Suppose that the translation which and C" to Cr Then J^ is on the line ^4 and Cl on the line
AC, and the
line
5^
is
parallel to
<?
BC. Thus,
if
Ba
be the point at
infinity of the line
AB, and
the point at infinity of the line
AC,
V TU 1Q Hence, by Theorem 18,

TT
_ = _ = _,
same
A'B'
as
*
1
which
is,
by
definition, the
AB ~ __ CA
C'A
In
like
manner,
it
follows that
AB _ BC
A'B'~
Since
B'C''
we have not defined the product of two vectors, it is necessary to resort to a device in order to compute conveniently with them. This we do as follows: DEFINITION. With respect to an arbitrary vector OA, which is called
a unit vector, the ratio
QJJ
OA'
where OB is any vector collinear with OA, is called the magnitude
of
OB.
Observe that the magnitude of OB is the negative of the magnitude of BO. Since the magnitude of a vector is a number, there is no diffi-
with magnitudes. In the rest of we shall use the symbol AB to denote the magnitude of the vector AB. No confusion is introduced by this double use of the symbol, because the ratio of two vectors is precisely the same as the
culty about algebraic computations
this section
quotient of their magnitudes. DEFINITION. If T is any collineation not leaving 1 M invariant, the l lines F(L) and T~ (lM ) are called the vanishing lines of F. If II is
any projectivity transforming a line I to a line I' (which may coincide with 1), the ordinary points of I and V which are homologous with
points at infinity are
If II
is
(if
existent) called the vanishing points of II.

I
an involution transforming
into itself but not leaving the

is
point at infinity invariant, the vanishing point
called the center of
the involution.
THEOREM
21.
DEFINITION. If
and
0' are the vanishing points,
43]
VANISHING LINES
I' ,
87
I,
parallel* line
to tin
and
X is
a variable point of
which
1
X is
transformed, the product

the transformation.
OX
and X' the point of V O'X' is a constant, called
power of
Proof.
Let P and let X^ and x be the point at infinity of I and I' be two values of X, and X/ and the points to which they are 2
;
'
transformed by the given projectivity.

property of a cross
(7iO,
ratio,
2)
Then, by the fundamental
X^
= R (0', X/X = 5 (JP 0', X 'X/),

2 ')
2
and hence, by Theorem

Hence, by the definition
18,
of
2
= __I.
of vectors,
magnitude
O'X!,
OX
COROLLARY
COROLLARY
point
the
1.
= OA\
O'X/.
center
27t0
power of an involution having a

is
and
a conjugate pair
2.
AA^
OA
be
is
OA^.
Let II
a homology whose center

line,
is
an ordinary
F and
whose axis
1
an ordinary
and
let
D be any point of
= H(P), and FD meets the
D'
is the
vanishing line IT" ^). If P is a variable point, P' point in which the line through P' parallel to
vanishing line II (fc), then
Proof.
Let
Q and
Q' be the points in
which the
line
FP meets the
vanishing lines
n~
(L) and Ii(L) respectively.
By
the theorem,
from which we derive successively
PF+FQ ~ ^ FP' +P'Q'

FQ
P'Q'
PF _ FP' FQ~ P'Q''

FP'~
Skice II
is
FP = QF
P'Q''
a homology, the two vanishing lines are parallel. Hence
QF _ DF
P'Q'~ P'D
1
Hence
FP
DF
EXERCISES
1. If
a projectivity
ABCD~j^ A'B'C'D'
the line
AB corresponds to
is such that the point at infinity the point at infinity of the line A'B',
of
AB _ CD
2. If
A'B'
'
C'D'
line in the points A', B' t C'
three parallel lines
a, b, c
are
met by one
respectively
and by another
line in
A"B"C"
respectively, then
'
A'B'
_ A"B"
A"C"
A'C'
3. If
ABCD are any four collinear points, AB CD + AC DB + AD -EC = 0.

.
4. Six points
form a quadrangular set Q
(/1 2
# 2 C2 J 1 B 1 C'
,
1)
if
and only
if
$ (AÂ V B&-) B (B&, CV4J B (C-jCj, A^BJ = - 1.

5.
The condition
for a
quadrangular
set
may
also be written
A&' B&' C^
6. If
A 1 Bs BjQ <V 2 ^
meet two other tangents in
three tangents to a parabola

respectively,
Pv Pv P
and
Q v QV Q 8
then
~ p _
Conversely, if five lines are such that the points in which two of them meet the other three satisfy this condition, the conic to which the five lines are
tangent is a parabola. 7. Let be the center of a hyperbola, and A 1 and A z the points in which the asymptotes are met by an arbitrary tangent if another tangent meets the
;
asymptotes
OA v OA 2 in 5 X and B z respectively,
OAi ~ OB Z
'
OB l
-
OA 2
8. If
a fixed tangent p to a conic at a point
meets two variable conju-
PQ' is a constant. Let be the center gate diameters in Q and Q', then of the conic. If the diameter parallel to p meets the conic in S, then
PQ
PQ-PQ' =9.
Let
a conic.
X Z is constant. If is the center of the conic and B is respectively, O l X l 2 a point of intersection of the diameter through parallel to the fixed tangents,
be the points of contact of two fixed parallel tangents to If a variable tangent meets the two fixed tangents in l and X%
(?!
2
and
43, 44]
THEORY OF TRANSVERSALS
of
89
44.
Theorems
Menelaus, Ceva, and Carnot.

points A', B', C' of the sides BC, of a triangle are collinear if and only if
THEOREM 22 (MENELAUS). Three

CA, AB,
respectively,
A'B B'C C'A

1
FIG. 30
Proof.
Let the points at infinity

,
, ,
of
BC, CA, AB,

\
C' collinear,
A'B' be
A M BM Cn
the intersection
P x respectively, and let A" be of AP X with BC. Then, supposing A', B', p p x x
A
A
it
The converse argument
is
now
obvious.
THEOREM 23
(CEVA). The necessary
and
sufficient condition
,
concurrence of the lines joining the vertices A,
for the of a triangle to
Suppose
first
that
C"
is
an ordinary point.
Then, by the theorem
of
Menelaus,
(5)
A'B B'C C"A A'C B'A C"B
if
The point C" is harmonically conjugate to C' with respect to A and S and only if the lines AA', BB', CC' meet in a point. Thus,
(6)
. '
C'B
C"A = C"B
But
is
on multiplying
a necessary and sufficient condition that AA', BB', CC' concur. (5) by (6) we obtain (4).
In case C"
is
an ideal point, the line A'B'
is
parallel to
AB
and,
by Theorem 20,
(7)
A'B B'C A'C B'A

'
I.
The condition that C" be harmonically conjugate to C' with regard to and B now takes the form C'A
-
C'B
On
multiplying this into (7)
we
again obtain
(4).
THEOEEM 24 (CARNOT). Three pairs of points, respectively, on the sides BC, CA, AB, respectively, of a on the same conic if and only if
(8)
triangle are
A v B I} GI and A t follows directly

multiply
conditions
B C
Z,
respectively.
The formula
(8) in this case
from
the
J3
lt
Theorem 22 when we
together that
A^
2
C: and A z
B C
Z
,
be
re-
spectively collinear.
Now
A^, line
consider
any
proper conic through
lt
Bv B
meeting the
C[ C2
'
.
A B in
By the
V,
PKJ. 32
theorem
of
Desargues
(Theorem
Vol.
,
19, Chap.
I)
the
pairs
AB,
C: CZ and C[C[
involution.
are in
Hence
x C C[BA ^
t
and hence
> '
C[A
G[A
or
Hence the formula

by substituting
for
(8) is
C[,
for C^,
equivalent to the formula obtained from it Cz respectively. Hence the formula holds
any conic. The converse argument is now obvious. The last three theorems are the most important special cases of the "theory of transversals." A few further theorems of this class and some other propositions which can readily be derived from them are
stated in the exercises below.
Further theorems and references will
be found in the Encyclopadie der Math. Wiss. Ill

III
AB
5,
2,
and
1,
23.
EXERCISES
1.
The
six lines joining the vertices
A, B,
of a triangle to pairs of points

if
AÂ^B^Ry C^CZ on
and only
2.
if
the respectively opposite sides are tangents to a conic

is satisfied.
the relation (8) If the sides BC, CA,
AB
'
of a triangle are
tangent to a conic in
Av
Bv
C, respectively,
CA, AB, EC,

'
3. If a line .BC*
meets a conic in
^ and A v and two

C2
and
' '
and
C, respectively,
meet
it
in the pairs C\,

'
^B
parallel lines throug]

z
respectively,
A^B A 2 B B^C BzC AjC A Z C C^B CJB
==l
meet a conic in the pairs A v A 4. Let two lines a and 6 through a point aid Bv B z respectively. If 0, a, b are variable in such a way that a and remain respectively parallel to two fixed lines,
OA l OA Z OBi OBZ
is
a constant.
5. If the sides of a triangle meet a conic in three pairs of points, the threi pairs of lines joining the pairs of points to the opposite vertices of the triangli
are tangents to a
second conic. State the dual and converse of this theorem
6. If
lines
conic.
7.
two points are pined to the vertices of a triangle by six lines, thesi meet the sides in six points (other than the vertices) which are on
Dualize.
If a line
meets the sides
A A V AÂ V
,
A nA
respectively, of a simpL
polygon AÂ-Â^
A n in points J5 J5 A Q B AjBi
meets the lines
1}
Bn respectively,
A n Bn _
8. If a conic
pairs of points
5C
BCV
A A V AÂ V AnA Bn Cn respectively,

,
respectively, in th<
^ô AtC A
9. If a conic is
A
3
AB
Ci
A Bn A Q Cn
,
tangent to the lines

,
A A V AÂ Z
A nA
respectively,
ii
the points
Bv
Bn respectively,
00
AiBo
A 2 B,
A Bn
nn =
'
45. Point reflections.

axis
is
l
DEFINITION.
reflection.
A homology of period two whose

once
:
n is called
a.
point
From
The
this definition there follows at
THEOREM
25.
point
reflection is
fully determined
"by its center
center is the
mid-point of every pair of homologous points. Everi
two homologous lines are parallel. THEOREM 26. The product of two point reflections whose centert are distinct is a translation parallel to the line Joining their centers
Proof.
The product obviously leaves
fixed all points of
and
alec
and let P be any point not on the line C^. Also let P' be the transform of P by the point reflection with C as center, and let Q be the l
transform of P' by the point reflection with C as center. Since C is 2 1 the mid-point of the pair PP' and C of the pair P'Q, the line is 2 parallel to (Theorem 12, Cor. 2). Thus the product of the two
y
PQ
C^
point reflections leaves invariant is a translation.
all
Hues parallel to C^C^ and hence
COROLLAKY. The product of any even number of point
reflections is
translation.
THEOREM
one of which
Proof,
tion,
27.
is
Any
translation
is the product
of two point reflections
arbitrary.
Lefc
be any translation,
Cs =
it
T((71 ), and
(7
C2
X
of the reflections in
<7
Cl the center of any point reflecthe mid-point of the pair C^C& The product and <72 is a translation, by Theorem 26, and
.
since
carries
to
<7
g,
it is
the translation T,
by Theorem
3.
COROLLARY
is
1.
The product of any odd number of point
reflections
a point reflection.
Proof.
Theorem 26 the product by Theorem 27, is the product
Let the given point reflections be P a , P 2 , P 2B+r By P2B reduces to a translation, which, of two point reflections one of which is
,
P^
P 2n+1 Hence
.
there exists a point reflection
P
l
such that
p r2 r p
l
'
'
'
p ^Zn +
pp rjr 2n + p2n +
l jr
p *
a,
COROLLARY
is
2.
The product of a translation and The

set
point reflection
a point
reflection.
3.
COROLLARY
a,
of
all
point reflections and translations form,

is
group.
THEOREM
self-conjugate
28.
The group of point reflections and translations subgroup of the affine group.
Proof.
lation
It
and 2 an
has been proved, in Theorem 11, that if T is a trans1 affine collineation, 2T2" is a translation. Precisely
similar reasoning
shows that
if
is
a point reflection,
2T2"
is
point reflection.
COROLLARY. The group
of point
reflections
and
translations
is
THEOREM
29.
With
respect
to
any system of nonhomogeneous

line, the
coordi-
nates in which
I* is the
singular
.
equations of a point
reflection
have the form

(9)
x'=
+ a,
fr.
y^-y +
The
point reflection
Proof.
whose center
~~"
J(j
is
the origin
is of
the form
because this transformation evidently leaves (0, 0) and / pointwise invariant and is of period two. Since any other point reflection is the
resultant of this one
and a
translation,
it
must be
of the
form
(9).
EXERCISES
1.
An
ellipse or a
is
whose center
2.
hyperbola is transformed into the center of the conic.
itself
by a point
reflection
Let [C 2 ] be a system of conies conjugate under the group of translations a to a single conic. Under what circumstances is [C' ] invariant under the group
of translations
3.
and point reflections? Investigate the subgroups of the
group
of
translations and point
reflections.
4.
Any odd number of point
reflections
P p P2
7l
satisfy the condition,
5. Let T be the point reflection whose center is the pole of L, with respect to the n-point whose vertices are the centers of n point reflections
PI>
P2
'
'
>
P-
Then*
PjTPjTPgT
PBT =
1.
46. Extension of the definition of congruence. are figures are said to be congruent if
DEFINITION.
Two
they
conjugate under the group

39.
of translations
and point
is
reflections.
This definition
obviously in agreement with that given in

57, Chap.
It will be completed in
IV.
The main
it
significance of the
of
present extension of the definition
is
that
removes any necessity
distinguishing between ordered and nonordered point pairs in state-
ments about congruence.

* Of. pp.
ters of Pj,
. .
46, 84, Vol. I.

.,
Pn
The center of T is the " center of gravity " of the cenCf H.Wiener, Berichte der Gesellschaft der Wissenschaften zu
.
THEOREM
30.
Any
ordered point pair
AB
is
congruent
to
the
*dered point pair
BA.
Proof.
Let
Dint reflection
with
be the mid-point of the ordered point pair AB. as center interchanges A and B.
reflection
The
COROLLARY. If a point
transforms an ordered point pair
to A'li',
By Theorem 26 the given point reflection is the product of point reflection in the mid-point of AB and a translation. The Dint reflection in the mid-point of AB interchanges and B, and the anslation leaves all vectors unchanged.
Proof.
IB
47.
lx
The homothetic group. DEFINITION.

a dilation.
homology whose axis
is called
Dilations and translations are both called
imothetic transformations.
letic
Two figures conjugate under a homotransformation are said to be homothetic.

figures are also called, in conformity
"
Homothetic
itroduced
with definitions
The
le
similar and similarly placed." point reflections are evidently special cases of dilations.
later,
Since
product
of
is
two perspective
collineations
(28,
Vol. I) having a
all
Dminon axis
a perspective collineation, the set of

;
homothetic
ansformations form a group and by an argument like that used >r Theorem 11 this group is self-conjugate under the affine group.
[ence
we have
31.
is
THEOREM
roup which
The set of all homothetic transformations form a a self-conjugate subgroup of the affine group.
Further theorems on the homothetic group are stated in the

sercises below.
EXERCISES
ratios of parallel vectors are left invariant by the homothetic group. 2. If two point pairs arid are transformed by a dilation into A'B'
1.
The
AB
CD
.
ad C'D' respectively,
-^
=
A^B"
'
IjD'
lines joining corresponding vertices point or are parallel. 4. The equations of the homothetic group with respect to any nonhomoeneous coordinate system of which fc, is the singular line are
3.
If
two triangles are homothetic, the
leet in a
48. Equivalence of ordered point triads. Although the theory of congruence as based on the group of translations and point reflections does not yield metric relations between pairs of points unless they
are
on parallel lines, yet when applied to point triads it yields a complete theory of the equivalence (in area) of triangles.* In this section we shall give the definitions and the more important
sufficient conditions for equivalence,
first
gous to those in the

angles, however,
book
of Euclid's
using methods somewhat analoElements. Instead of triThis
we
shall
work
-with ordered triads of points.
permits the introduction of algebraic signs of areas, though, as we do not need to refer to the interior and exterior of a triangle, we shall
not actually employ the
word
" area."
The
triads of points
which
are
referred to are all triads of noncollinear points.
Our
triangle
definitions
ABC is
are equivalent in area
have their origin in. the intuitional notions: that any equivalent in area to the triangle BCA, that two triangles if one can be transformed into the other by a transla-
tion or point reflection,
and that two triangles which can be obtained by adding
equivalent triangles are equivalent.
ABC and ACD are two ordered point triads, and ABD is and B D 33), the point triad called the sum of ABC and ACD and is denoted by ABC + ACD or by
DEFINITION.
If
ft C,
and
are collinear,
(fig.
ACD + ABC.
DEFINITION. An ordered point triad t is said to be equivalent to an ordered point triad t' (in symbols, t if t can be carried to t' t') (1) and by a point reflection, or (2) if t and t' can be denoted by
ABC
building up the theory of areas without the aid of a full theory of congruence is due to E. B. Wilson, Annals of Mathematics, Vol. (2d series) (1903), His method is quite different from ours, being based on the observation p. 29. (cf 52, below) that an equiaffine collineation is expressible as a product of simple shears. Still another treatment of areas based on the group of translations and employing continuity considerations is outlined by Wilson and Lewis, "The Spacetime Manifold of Relativity," Proceedings of the American Academv of Arts and.
of
The idea
BCA
that
t[,
t' z
respectively, or (3)
t
if
there exists an ordered point triad
such
t
lt
and t^t', or
(4) if
there exist ordered point triads
such that t^t'i, t 2 t = t -\-t and tz and z l v An ordered t'=t[+t' point triad t is not said to be equivalent to an ordered point triad t'
unless
it
z,
follows,
by a
t
finite
number
(2), (3), (4),
that
of applications of the criteria (1),
t'.
Since any translation

(1)
is
a product of two point reflections, Criteria
and
(3) give
THEOREM 32. Two ordered point triads are equivalent if they are conjugate under the group of translations and point reflections. THEOREM 33. If A, B, and G are noncollinear points, ABC ABC,
ABC^BCA, ABC^
Proof.
CAB.
the definition
(3),
it
From
(2) of
follows that
ABC ^ BCA
(2),
and
BCA *=
Hence,
CAB. Hence, by
ABC *=
CAB. But, by
CAB
ABC.
tyÂBCÂBC.
the form of the definition
34.
35.
From the last two theorems and from we now have at once
THEOREM THEOREM
,
If t^ t z then t z ^ t f If A, B, C are any three noncollinear points and

the
the mid-point
of
pair
AB
then
AOCÔBC.
Proof.
Let
is
changed
be the point to which by the translation
shifting
to
0,
and
let
be the
nonSince
is
point
of
intersection
lines
of
the
parallel
BC
and
OC'.
COBC'
is
a parallelogram,
of
the
C'O.
mid-point
the pairs
CB and
Thus we have
and
But the point Tlms

and
AOC sc OBC' *= BC'O = BC'M + BMO OB C = OB M + OMC. reflection with M as center carries OMC
into
C'MB.
OMC *z C'MB ^ BC'M, OBM^BMO,

above,
and hence, by comparison with the equivalences and equations nrtn
THEOREM are 3= C
,
36.
where Two ordered point triads ABCl and Z the line AB. equivalent if the line C 1 (7a is parallel to
,
f
ABC
Proof.
Let
Cs
be
such
that
the mid-point of (7^, and let the line (72 <78 in 0, which meet the line
is
AB
is
an ordinary point because Cs is not on the line C^.

It follows ( 40) that mid-point of the pair
is
the
.
C2 C3
By Theorems 34 and
definition,
35,
CSOA. By Theorem 35,
CS BO 4CS BO Cn AO. and CjOA CS BA
FIG. 35
Hence
^ (7^
ABC
.
4- C,
Hence ABC,
COROLLARY. If a point
different line
is
on a line
OB
and a point
C' on
00, and the lines BC' and B' C are parallel, BOC^B'OC'.
hypothesis,
Proof.
By
BOC=OC'-{-BC'C
and C'B'O = C'B'B + C'B 0.
But C'B'B^C'CB^BC'C,
by Theorems 36 and 34, and C 'BO BOC', by
Theorem
34.
Hence
BOG
FIG. 36
Ô'B'O^B'OC'
P PQ such
Proof.
THEOREM 37. If A, B, and C are any three noncollinear points, and and Q are any two distinct points, there exists a line r parallel to
that if
is
any point of
r,
ABC^PQE.
not
Let
be the translation such that
--B'
and T(<7)
<7'.
If
'
is
T (A) = P, and let on the line PQ, let R?

C
1
be the intersection
(fig.
37) of the line through

parallel to QB'.
If
parallel to
line
PS
with the line through

-R'
is
on the
PQ,
let.
be the point of intersection with
PC'
of the line
through
parallel, so that
In both cases the lines which intersect in E' are by hypothesis nonR' is always an ordinary point. By Theorem 32.
In case
follows,
'
ABC^PB'C'.
is
not on
PQ,
it
by
'
Theorem
is
36, thatP^'C"^ PB'R'^PQR'. In case
on PQ,
it
follows,
by the
36,
corollary of Theorem that PB'C'^PQR'.
By
r
Theorem 36
the
line
through R' parallel to PQ is such that

for
every point
R
JFia.
^P QR.
EXERCISES
1.
87
and AB'C' are equivalent if the points ordered point triads f B, C, B', C' are collinear and Voct (BC) = Vect (B'C ). 2. Let be the point of intersection of the asymptotes I and m of a
be the intersections with Z and m respectively. hyperbola, and let L and of a variable tangent to the hyperbola. Then the ordered point triads are all equivalent.
49.
Two
ABC
OLM
Measure
of ordered point triads.
The theorems
of the last sec-
tion state sufficient conditions for the equivalence of ordered point

triads.
In order to
c,
obtain necessary conditions,
Q -----------
we
shall in-
troduce
the
notion
of measure, analogous
to the
magnitude of
Let
a vector.
DEFINITION.
0,
P,
be three non-
collinear points.
The
FIG. 38
measure of an ordered
point triad
tive to the
ABC relaordered triad

:
OPQ
as a unit is a
number m(ABC)
deter-
mined
as follows
If
the line
BC is
not parallel to OP, let
and C1
to
OP, meet the
line
to
through A, parallel
tude of the vector
OQ, and let A^ be the point in which the line OP, meets the line BC. Let AA l denote the magnii
AA
relative to the unit
OP
magnitude
of the vector
B^C^
*
relative to the unit
( 43), and Bfi^ the OQ. The measure of
the ordered triad
ABC is
^
If
^^
and
is
denoted by
m (ABC}.
the line
BC is
parallel to
OP, CA
is
not parallel to OP, aiid the measure of C is defined to be (BCA). If this definition be allowed to apply to any ordered point triad whatever (instead of only to noncollinear triads, cf. 48), we have
AB
m(ABC] =
THEOEEM
Proof.
whenever the points A, B,

38.
IfABCÂ'B'C
1 ,
then
C are collinear. m(ABC] = m(A'B'C'}.
Let us examine the four criteria in the definition of equiva48.
lence in
(1)
In case
the vectors
corollary),
ABC is carried to A'B'C' by a point reflection, each oi AA and B C is transformed into its negative (Theorem 30,
l
l
and hence the product of their magnitudes is unchanged, (2) According to the second criterion, ABC^BCA. Suppose, first that neither BC nor CA is parallel to OP, and let A B C have the lt lt l meaning given them in the definition above. Then
m(ABC)=AA -B C
1
be the point in which the line 2 (fig. 38) through B, parallel tc OP, meets the line CA, and let A 2 be the point in which OQ is met by the parallel to OP through A. Then if and represent the
Let
BB
CÂ 2
magnitudes
of
the
corresponding vectors
relative
to
OP
and OQ
By Theorem
But since the
2 0,
lines
44 = 4 B BC BB B B
Z
CC AÂ 2 BB
lt
,
are parallel,
it
follows from
43
that
BC
Hence
or
*
44 ^
m(ABC]
=AA ^
l
V=
1
Cl
The
factor \
is
lacking in this expression, because
we
are taking a triangi
In case
BC is
parallel to
OP, the
last clause of
the definition states
[n case
CA
is
parallel to
OP,
AB
and
B C are
not parallel to OP, and
lence the argument above shows that
But,
by
definition,
ELence
m(CAB)=m(ABC). m(BCA) = m(CAB). m(ABC)=m(BCA).
(3)
\~tv we
f
TO
(4)
= t and t 2 t then Corresponding to the fact that if t l z s have that, since m(t) is a uniquely defined number, = TO (7 2 ), and TO 2 ) = TO ( g), then TO (tj = TO (tf 8). (^)
,
(z!
Let B,
C,
be three
col-
inear points
and A any point not

In case
m
et
the line
BC (fig. 39).
;he line
B C is not parallel to OP,

OP,
l
be the point in which the
ine through A, parallel to
neets
;he
BC, and
let
B G D
I}
I}
be
points in
which the
lines
through B, C,
parallel to
respectively,
OP, meet OQ. Then
m(ABiy)
In case the line JSC

;o
is parallel
OP, let S be the point in which BC meets OQ, and A 2 be the
which the line through 4,paralleltoOP,meetsOQ. Then

)oint in
m(ABD)=m(BDA)=:
TO (BCA)
Chus, in every case,
if t
:
+ m(CDA) = m + = m (^) + m (t
t
t
2
TO
s,
2)
(t 3 )-
)f
Comparing the results proved in these four cases with the definition t equivalence, we have at once that whenever t m(t) = TO (t ).
,
102
THE AFFINE GROUP

39.
IN
THE PLANE
and
[CHAP,
ill
THEOREM
not on the line
If B, BC,
C,
and
D are
collinear points,
the point
A is
m(ABC) BC m(ABD)~BI>'
C is not parallel to OP, let A lt Proof. In case the line the meaning given them in the definition of measure, and let
lt
C have
l l
be the
39).
to OP, meets point in which the line through D, parallel
OQ (fig.
m(ABD)
But,
by
43,
In case
BC is
parallel to
OP,
let
through A, parallel to OP, meets OQ, and meets OQ. Then

in
be the point in which the line S the point in which -BC
(ABC)
m (ABD)
COROLLAEY
1.
BC SA ~ = BC ~~ = m(BCA] = ~
-
'
m (BDA)
BD SA
BD
IfB,
C,
D, E are points no two of which are
collinear
with a point A,
AC ,AC,
COEOLLABY
2.
If B,
C,
D are points no
is the
with a point A, and if
P^
two of which are collinear point at infinity of the line CD (the
latter not being parallel to AJB),
B AC AC>
'
THEOREM
Proof.
40. If
m(ABC} =
m(A'B'C'}=i-
0,
then ABCÂ'B'C'.
By Theorem 37 there exists a point C" on the line A'C' such that AB C Â'B'C", Hence A'B' C' Â'B' C", and by the last
theorem, C'
= C".
of
In consequence
the last
two theorems the unit point
triad
may
to
to
The unit triad AB. Then if C is the

Proof.
l
OPQ may
be chosen so that
OP
is parallel
OP,
point in which the line through C, parallel meets OQ, and J$ 1 the point in which meets OQ,
AB
m(ABC)=AB-B
If C[ is the point in
which the line through
C',
parallel to
OP, meets OQ,
By Theorem
line
38,
m(ABC} = m(ABC'},
AB.
and hence C^=

is
C[.
Hence the
CC'
is
parallel to
42.
THEOREM
the line
If ABCÂB'C', and B'

the line
AC, then
BC'
is
parallel
to the line
if
on the line AB, and C' on B' C.
Proof.
By
the corollary of Theorem 36,

to B'C, then
C"
is
a point of
AC' such
that
BC"
is parallel
ABCÂB'C".
41 the only points C such that ABCÂB'C are on the line through C", parallel to AB'. Hence C" = C".
By Theorem
although the sufficient conditions for equivalence given the discussion of proved on the basis of Assumptions A, E, the ratios of vectors, and hence all the necessary conditions for equivalence,
in
It is notable that
48 are
all
involve Assumption
in their proofs.
This
is essential,*
as
we can show by
proving that Assumption P is a logical consequence of these theorems, together with the previous theorems on equivalence. As was pointed out in 3, Assumption
P
I.
is
a logical consequence of the
Vol.
When
one of the lines of the configuration

:
theorem of Pappus, Theorem 21, 36, is taken as /, this theorem
assumes the form
C
If a simple hexagon
AB'CA'BC'
that
line
is
such
another
is
A, B, C are on one and A', B', C' on line, and if AB' parallel to A'B and
is
J3C' parallel toB'C, then
CA'
parallel
to
C'A.
In case the lines

containing
ABC
and
A'B'C', respectively, are parallel, this can be proved from the Desargues so that we are interested only in the theorem on perspective triangles
;
The
r61e of
the theory of
Assumption P (or rather of the equivalent theorem of Pappus) in areas was first determined in a definite way by D. Hilbert, Grundlagen
projective
geometry
if it
be desired to base projective geometry

54).
upon
elemen-
tary Euclidean geometry (cf. Ex. 3,
The notion
points,
i.e.
of
measure can be extended to any ordered set
of
11
(cf.
14, Vol. I) to
any simple n-point.
The
details of this
problems below.
Vol. II,p.
It is to
An outline is furnished by the The principal references are to A. F. Mobius, Der 1, 17, 18, 165; Werke, Vol. I, pp. 23, 39, 200; barycentrische Calcul,
discussion are left to the reader.
48 5. See also the Encyclopadie der Math. Wiss., Ill AB
9,
1 2.
be borne in mind in using these references that our hypotheses are narrower than those used by the previous "writers.
EXERCISES
1.
For any three points A, B,
C,
2.
For any four points 0, A, B,
C,
m(ABC) = m(OAB~) + m(OBC) + m(OCA).

3.
For any n points
Av Az
An
the number
is
the same for
all
choices of the point 0.
We
it
define it to
be the measure
of
the simple n-point
AÂ 2
An
and denote
by
m(A l A z
An
~).
6.
Derive a formula for
m(A A Z
1
'
A n)
analogous to the definition of
m(ABC)
1.
in terms of vectors collinear with two arbitrary vectors and OQ. Prove the converse propositions to those stated in the exercises in 48.
OP
8. If
ABCDaud.A'B'C'D'
are two parallelograms

'
whose sides are
respec-
tively parallel,
m(ABCD} = AB_ BC_ m (A'B'C'D') A'B' B'C'
'
9. The variable parallelogram two of whose sides are the asymptotes of a hyperbola and one vertex of which is on the hyperbola has a constant measure.
10. If a variable pair of conjugate diameters meets a conic in point pairs A A', BB', the parallelogram whose sides are the tangents at A, A', B, B' has a constant measure. The parallelogram ABA'B' also has a constant measure.
50.
The
equiaffine group.
are,
l
point triads t and t a then t[ *= t' t[ and t! 2 v

,
THEOREM 43. If two equivalent ordered transformed, by an affine collineation into
Proof. It is necessary merely to verify that the relation used in each of Hie criteria (1), in the definition of equivalence , (4) 48) ( is unaffected by an affine collineation. For Criterion (1) this reduces
to
Theorem
and
28.
For Criteria
(2), (3), (4) it is
a consequence of the
fact that an affine collineation transforms ordered triads into ordered
triads
collinear points into collinear points.

44.
THEOREM
If an
affine collineation
it
triad into an equivalent point triad, triad into an equivalent point triad.
transforms one ordered point transforms every ordered point
Proof. It follows from Theorem 43 that if ABC is transformed by a given collineation into an equivalent ordered point triad A'S' C', then every point triad equivalent to A B C is transformed into a point
triad equivalent to A!B'C'
Theorem 37 any ordered

point triad
ADC,
and thus into one equivalent to ABC. By point triad whatever is equivalent to some where is on the line AB. Hence the present
theorem
will be
proved
if
we can show
that
ADC is
transformed into
an equivalent point triad. Denote the point to which

tion
is
transformed by the given collinea-
by
D'.
By Theorem
39,
m(ADC) _AD_ ~~
m (AB C)
AB
<m(A'D'C'}
m (A'B' C')
= A'J> ~
_ '
'!;'
By
43,
where
P x is
the point at infinity of the line AB. But since the given
collineation is affine, R, is transformed to the point at infinity
P^
of
the line A!B', and
Since
m(ABC] = m(A''C'),
it
follows that
ADCÂ'D'C*.
DEFINITION.
Any
affine collineation
which transforms an ordered
subgroup of the
Proof.
affine
group.
By
the last theorem an equiaffine colliueation transforms
triad. Hence, by every ordered point triad into an equivalent point Condition (3) in the definition of equivalence, the product of two
is equiaffine. equiaffine collineations
By Theorem
affine.
43,
2T2'
is equi-
affine
whenever
is
equiaffine
and 2
1
THEOREM
let
46. Let A, B, A',
be points such that
B and A
B';
a
B'.
be
line
is
on
! ,
but not on B,
and
let a' be
a line on A' but not
on
There
one and only one equiaffine collineation transforming
to A',
to
and a
to a'.
a.
Proof.
there
is
Let C be any point distinct from A on a point C' on the line a' such that
By Theorem
37,
ABCÂ'B'C'.
By Theorem 1 there is one and only one affine transformation carrying A, B, C to A', B', C respectively, and by definition this transformation
1
is
equiaffine.
By Theorem
41, C'
is
the only point on a' such that

is
1
ABOÂ'B'C'.
Hence (Theorem
44) there
only one equiaffine
transformation carrying A, B, a into A', B', a respectively.
EXERCISE
Any
affine collineation leaves invariant the ratio of
the measures
of
any
two point
triads.
*51. Algebraic formula for measure.
Barycentric coordinates.
Consider a nonhomogeneous coordinate system in which L is the singular line. Let the unit of measure for ordered triads be OPQ,
where
= (0,
c
2)
;
C=
(c x ,
)j
(
Let A (0, 1). 0), (1, 0), Q (a lt aj, B (^ &,), the line through A, parallel to OP, consists of the points where X is arbitrary, and the line has the equation
P=
BC
64, Vol. I),
=
In case the line
{
0.
BO is not parallel to OP, and therefore b

definition of
3s
2,
the point
l
which appears in the where X satisfies X
measure
49)
is (
+ X,
#,j)>
= o.
01J
JbUJti
107
leuce
AA =
5
1
-1
C
l
C
2
1
(0, & 2 )
'he points
and
G^
of the definition of
measure are
7,
and
(0, c
),
espectively, so that
n r
-~
lence
(10)
m(ABC) =
c
12
C
i 2
"hat the.
same result holds good in

A, B,
case
BC
is
parallel to
OP
is
eadily verified.
Now
(11)
f
if
are transformed to A', B',
C
ft
respectively
by a
ransformation
/lf
A=
the affine group,
(12)
m(A'B'C'}
561
c
i
lence
we have
47. -^ transformation (11) o
affine
THEOREM
f a?i^
group
is
equiaffine
ow^
*/*
" *ift
(b : , b 2 ),C
'
= 1.
c
g)
Let
A = (a
a z ),B =
= (G
I}
be the vertices of any triangle,
nd
P = (x,
y)
any
point.
}Jx9 =x, x2 /x
y,
In the homogeneous coordinates for which a t> a2 ), etc. these points may he written
A(\,
lence by the result established in iase, the numbers proportional to

1
27
for the three-dimensional
x y
I 1
0=
x
c
i
y
C
2
nay be regarded as homogeneous coordinates of vhich ABC is the triangle of reference.
in a system for
an equiaffine colliueation. In like manner, [Lj,^ {LI}
is also equiaffi
Hence the product {L z ty
{L^}
is equiaffine.
THEOREM
reflections.
49.
An
equiaffine collineation is
a product of two
Proof.
Let
be any equiaffine collineation.
If
there be any pc
which
,
is
not on an invariant line of F, let
be such a point.
A A A
z
,
be defined by the conditions
By
the hypothesis on
the points
equiaffine
A A A
Q>
I}
are uoncolliuear,
and
the hypothesis that
is
AA4>-A^M-Â
,
Hence, by Theorem 41, the line A A 3 is parallel to AÂ Z or else A of the pair A. be the mid-point of the pair AÂ Z and Let z i
Let L^ be the point at infinity of the line A A Z> L z of the line AÂ^ i is parallel it follows t to of the line A,A,. Since A.A. * * o l 18
AA
A,
L A A ZL^=A,A.^K, and hence, by the definition of mid-point, that M and L are collinear. Since A A and the point at infinity of the
Q
z
,
lt
z,
A AS
are transformed
by
to
^
/
^4
3,
and the point
at infinity of
Let ^ be the line

to the
A^M^ and
mid-point of
M^M By
.
the line joining the mid-point of 2 the above,
and
Hence
But
since
{Z,y
{
,
r(J ^ 1 Jf1 ) = ^ 1 ^ 2 lT2
it
follows,
by Theorem
1,
that
a contradiction with Theorem 46

concurrent.
if
the invariant lines were not
Also
Let A^ be a point which is not invariant, and let A = F (^4 ). 2 let B be another point which is not invariant and not on
1
the line AÂ# and let

sarily
If
F (B^ = B 2
The
lines
AA
V
and
BB
1
neces-
meet in
is
0.
points of
ln
ordinary, then since any line through it is invariant, all are invariant, and hence A B is parallel to A B Since
1
F is
equiaffine,
Hence,
"by
parallelogram.
Theorem 42, A^B Z and A Z B^ are parallel, and A^B^B^ is a is the mid-point of Hence AÂ Z and B^B^ and F is
a point reflection. Let a he the line
AÂ 2
and A the point

of a,
at infinity
and
let &
be the line
B^
B
l}
and
the point at
&. The product {Ah}- {Ba} transforms A { ,
infinity of
into
tively,
A 2> B2 ,0 lesseeand hence is F.

is
'
Mj
FI<J _
4g
If
an ideal point,
into 0,
1.
joining the mid-points of

0,
A B
I}
A B
2
,
let be the line A^B^ and let m be the line AÂ Z and B^B Z Then {Om} {01} transforms respectively, and hence, by Theorem 46, is F.
I
.
COROLLARY
2
An
equiaffine collineation
all choices
such that A,
F (A) and
axis
I*
(A) are collinear
for
of A
is either
is
a point
reflection or
a translation or an elation whose center

is
at infinity
and whose
an ordinary
Proof.
line.
it
In the argument above
was proved that
if
the point
is
ordinary,
If
F is
I
and
a point reflection; and that if is ideal, are parallel, F is evidently a translation

it is
F = {Om}
;
{01}.
are not parallel,
an elation with
as center
and if m and I and the line joining
to the point
Im
as axis.
112

An
2.
[CHAP,
DEFINITION.
is
elation
is
whose center
is
at infinity
and whose
axis
an ordinary line
called a simple shear,
COROLLARY
with
I
If F
is
{/}
{L^,
then
for
every
^ne
concurrent
and
which
not a double line
ofT
there exist points
and
M and
a line m, such that
r = {Mm}
There also exist a point
{LI}.
M' and a
F
-
line
m' such that

f
r
= {Ll} {M m }.
If
be
taken as variable,
Proof.
choice of
follows
The
first
conclusion follows from the arbitrariness in the
in the proof of the theorem above.

first,
The second conclusion
from the
combined with the
fact that
The
projectivities follow
theorem for
A A
Q,
z,
from the constructions given in the proof
of
r,
etc.
COROLLARY
which
lines
I
is
3. If T {L {L^, then for every point L of L not a double point of T, there exists a point of l m and two and L such that and concurrent with L 1
2,
T = {Mm}
There also exist a point
{Ll}.
and a
{Ll}
line
m' such that

1
T=
THEOREM
of
is either
{M'm }.
which are products
this
50.
The
set
of
all affine collineations
line reflections
form a
group.
Every transformation of
group
an
equiaffine transformation or the product of

reflection.
an
equiaffine
transformation by a line
Proof.
By Theorems 48 and 49
of line reflections is equiaffine

reflections.
the product of an even number and reduces to a product of two line Henue the product of an odd number of line reflections
r\f fl-iraa
rAflnnoe fn a nrnrl n r>f.
lina -raflanfiryna
Tl-ia
of nf QTV-,/->V^CI n\~,mit\
EXERCISES
Let the points at infinity of l v ?2 1 respectively in Theorem 49, Cor. be denoted by /,;, 14, 77. Jf tho points L L{, L L'^ are v v distinct, the
1.
,
2,
pairs
ZjZj, L 2 L 2 LL' are in involution. 2. In case Z is on and Z 2 is z

,
not on l v {L 2 12 }
{LJJ = T is
a collineation
and of period two on the line and lr If I be any line, except ?2> joining L i { P the point in which I meets L, and L the harmonic through the point 2 conjugate of L { with respect to P and T(P),
(cf.
of
Type //
40, Vol. I), parabolic on to the point of intersection of l
If
M
3.
T=
(
be the harmonic conjugate of

2 }.
with respect to
if
and
if
{/VH^
Z2
on
The product {LJ2 } {Z,^}

l
and
is a point reflection as center point reflection with

j
and only
0,
1
is
on
is
the product of any

2
two line reflections {Z^} and {Z 2 /2 and Z 2 on lr

4.
}
for
which
is
on
on
0,
Z on
:
1 , 2
is
The product {Z 2 /2 {Z,^} is a translation if and only if L : = Z 2 and ^ The ideal point Zj is the center of the translation. If T is parallel to /2
.
any
translation, To, its center,
point of the pair

respectively,
5.
{
PPV
P^ any ordinary point, P = T (P a ), P2 the midand p l and ^; 2 two parallel lines through P1 and P2
-
T = r, ^ 2
Z 1
T.,^}.
{Zj,^} is a simple shear
,
if L ^ L and ^ = ls or l 2 an ordinary point, but not in any other case. 6. Let 2 be a simple shear whose axis is I and whose center is Z. Let Px be any point of lx P = ^(P^, and P2 the harmonic conjugate of L with respect to P and P2 Then 2 = {P 2 {P^}- If Pi be any line meeting I in an ordinary point, p = 2(/> 1 ), and p z the harmonic conjugate of I with respect to p and jo v
The product {Z 2 /2 }
and
if -î
= Z2
intersects
in
7.
Let
PP P P P
1
be a simple pentagon.
1
Let
C C2 C
1}
,
,
C7
4,
(75
be the
mid-points of the pairs

.PP-L is parallel to
3
lines C^C^, 2 1 Discuss the degenerate cases. 8. Every equiaffine transformation is either the identity or a point reflection or an elation whose center is at infinity a translation or a simple shear) (i.e. or expressible as a product of two elations whose centers are at infinity.
,
PP V P P P P P^v P*P respectively. P P4 and PP is parallel to P P the three

2, Z
3>
If the line
C2 C
5,
PC*8 are concurrent or
parallel.
9.
Prove Cors. 2 and 8 of Theorem 49 directly, without using the theory
of equivalence.
10.
necessary and sufficient condition that a planar collineation be the
product of two harmonic homologies is that it transform ordered point triads into equivalent point triads relative to a fixed line of the collineation regarded
by 11. Let us denote an involution whose double points are L and {LM}. If ^-{L^M^} and I2 = {L 2 Z } are two distinct involutions on the same line, then for every point L s of this line, L 3 not being a double point and involution {L i } such that of I1 I2 there exists a unique point s
if
we
denote {L B
4t
S}
by J8 and {i 4 MJ by J4
VzA = 7
The
4>
and
72 J1
= Vr
are all pairs of the same involution, 4 4 pairs V LZ Z L S 3 have a point in common, in which case all unless the pairs L^M^ and L Z Z in common. this have four pairs point
L^M
M M M
,
12.
The
2
form a group. The last two exercises connect with the following algebraic considerations. An involution in a net of rationality is always of the form ( 54, Vol. I)
{i 1
1
M }-{I M
projectivities
a}
on a
line
which are expressible in the form
ax
where
a, 6, c,
d are rational.
The
cz
8
double points are the roots of
2 ax
-b=
0,
and both
will be rational if
is
rational in
o2
+ be =
k2
Now
of
any which may
projectivity is
the product of two involutions, a double point of one be chosen arbitrarily. The projectivity may therefore be written
,
o,
ff
c'
-ax
ex
ax
ex
ft
v +b
,
_ ~ (aa' + Vc) x + (a'fr

(ac'
a'c)
ab')_
',
(5c'
'
aa')
and so has the determinant

aa'bc'
+ aV 2 + &6'cc' + Vcaa' (aa'W aWc' = a2 (a' 2 + 6V) + be (b'c' + a' 2) = /W,
a' 2 6c
aa'&'c)
double k' 2 = o' 2 + b'c'. Hence (1) the product of two involutions whose points have ratioaal coordinates is a projectivity whose determinant is a perfect square; and (2) if the determinant of a projectivity is a perfect square, and
where
one of two involutions of which it is a product has rational double points, then the other has rational double points. Hence there is a subgroup of the
tions with rational double points.
group of collineations of a linear net of rationality generated by the involuThis is the group of transformations whose
The
line
L now
has the equation x
0,
and the equations
(1) of the
affine
group become
(13)
2
A=
==
^2^0
'
az
=0.
^2^1
On
the line
ln
this effects the transformation
According to
54, Vol.
&
is
2
Thus
=
1
& I, this is an involution if and only if a 1 = g a necessary condition that (13) represent a line
.
reflection.
The ordinary double points of (13) are given a a Z> g equations, in which we have put l
by the following
(a-l)
(1*)
If (13) is to
+ Mr + .
-0.
be a line reflection,
(14)
Hence the two equations

which, requires
(15)
it must have a line of fixed points. must represent a single ordinary line,
= a-l
c.
The
1. first of these conditions is equivalent to A Since the coefficients of x and y in (14) cannot all vanish, the conditions (15) are also sufficient that (14) represent a single ordinary
line.
Hence THEOREM
(16)
51.
transformation of
x'
l
the,
form
lf
y
is
= ax + \y + c = a x -ay + c,,
z
line reflection if
and only
1
*;
if
A=
lc.
=
-(a
0.
From
that
this
it
follows that a product of two line reflections is such
A
is
and a product of three line reflections is such that = 1. Bj Theorems 47 and 49 any transformation for which A = 1 a product of two line reflections. Any transformation T for which
1,
A = 1,
A=
tion
when
multiplied
by a
line reflection
yields a transforma-
for
which
follows
T = 2A.
(cf.
A = 1, i.e. an equiaffine transformation. From TA = 2 Hence T is a product of three line reflections. Thus
TJie
we have
Theorem 47)
52.
THEOREM
group of
affine
transformations which are prod-
ucts of line reflections has the equations
= ctjX 4- 5 y
t
-f-
Cj,
a,
&
= 1.
EXERCISES
1.
The
set of all affine transformations
which are products of equiaffine

is
of the affine group.
transformations by dilations form a group which Its equations are
a self-conjugate subgroup
= a^x +
o^y
bz y
y'=az x +
where k
2.
is
+ +
cv
c2 ,
The
set of all affine
and
dilations
any number in the geometric number system. transformations which are products of line reflections form a group which is self-con jugate under the affine group. Its
equations are
x'=ax-4-bv4-c
where k
is
any number in the geometric number system.
give below a list of the principal subgroups of the affine group which we have considered in 30 of Chap. II These are all this chapter and in self-conjugate
54. Subgroups of the affine group.
We
subgroups. We also include the groups which will be considered in the next chapter in connection, with the Euclidean geometry. The groups are all described by means of the conditions which
must be imposed on the coefficients of the equations of the affine group to reduce it to each of the other groups. In some spaces, i.e.
when the
variables
and
coefficients are in certain
number
systems,
these groups are not all distinct.
However, they are
all distinct in
where
A= ^
Z
0.
The
(2)
principal subgroups connected with the affine geometry are
A>0;
(
the transformations satisfying this condition are direct

(3)
30).
A = F,
Jc
where
(4)
is
in the geometric
number system
53, Ex.
1).
A=
Jc
F,
(
where
(5)
is
in the geometric
number system
53, Ex. 2).
A = l;
2
these are products of two or of three line reflections (Theorem 52).

(6)
A = l,
(
the equiaffine group

\
/
51).
~"
2
1
~~~
'
2'
the homothetic group

/Q\
\
'
(47).
7i
ft
C\
rt
'
7i
rt
1
1
'
2J
the group of translations and point reflections
45).
the group of translations ( 38). The principal groups connected with the Euclidean geometry are
the Euclidean group
(55
&1
and
62).
Its
transformations are called
similarity transformations.
(11)
al
+&
2 2
0,
a 151
+a &
2
=0,
A>0,
the direct similarity transformations.
where
Jc
is
in the geometric
number
system,
where k
is
in the geometric
number
system.
(14)
a-i
+
-f-
2 2
=1
ai
a2
(
=^
62).
the group of
(15)
displacements aud symmetries

af
a aa
= 1,
aj.
=&
>
a2
=~&
i
the group of displacements.
A/fine
The relations among

these groups may be indicated by the follow-
Euclidea
ing diagram, in which we have included only

those groups which are
distinct in case of the
real geometry.
A dotted
Homothctic
Displacements
line indicates that the
lower of the two groups joined is a subgroup of

the upper, and a solid line that it is a self-
conjugate subgroup.
The fundamental importance of the group of translations is indicated by the fact that it is a self-conjugate
subgroup
of
Translations
each of the other groups.
FIG. 44
EXERCISES
Supposing the number of points oil a line to be p + of transformations in each, of the groups listed above ?
1.
1,
what is the numl:
Supposing the geometric number system to be (a) the ordinary real, (6) the ordinary complex number system, how many parameters are there the equations for each of the groups listed above?
2.
3. Prove that the plane affine geometry as a separate science could be bas on the following assumptions with regard to undefined elements, called poir and undefined classes of points, called lines
:
and only one line. IT. For any line I and any point P, not on I, there is one and only one containing P and not containing any point of L III. Every line contains at least two points. IV. There exist at least three noncollinear points. V. The special case of the Pappus theorem given in the fine print in
I.
Two
points are contained in one
1;
CHAPTER IV
55. Geometries of the
Euclidean type.
We
come now
to
the
extension of the definition of congruence which was promised in 39 and 46. This requires the consideration of groups which are not self-conjugate under the affine group. Not being self-conjugate,
these groups are not determined uniquely by the affine group, and heuce our definitions will contain a further arbitrary element.
DEFINITION. Let
be an arbitrary but fixed involution on
Z OT
This
involution shall be called the absolute or orthogonal involution. The group of all protective collineations leaving I invariant shall be called a parabolic * metric group. The transformations of the group shall be called similarity transformations. Two figures conjugate under
the group shall be said to be similar. The geometry corresponding to the group shall be called the parabolic metric geometry.
The absolute involution

rest of the discussion,
is supposed to be fixed throughout the but of course there are as many parabolic metric
groups as there are choices of I. We nevertheless speak of the parabolic metric group in order to emphasize the fact that we are fixing attention on one group.
the
In case the plane in which we are working is a real plane and absolute involution is without double points, the parabolic
is
metric geometry
that
the Euclidean geometry.
It is for this reason
we
refer to the parabolic metric geometries
as geometries of
the Euclidean type.
The
of increasing
investigations in the following sections are arranged in order specialization. First we consider a perfectly general
Then we I, in a projective plane satisfying A, E, P, O consider a particular type of involution in an ordered plane, and finally limit the plane to be the real plane.
involution,
.
* The reason for the term "parabolic"

later chapter,
in this
connection
is
explained in a
where the
elliptic
and hyperbolic metric groups are defined.

119
WlllUli
Lilt)
aUBUllll/t/
JLUVUIUHUJJ.
iiao
u.uu.uj.0
jju-u-iuo.
JLJJ.UO
the theorems on the general type of involution (where the possible existence of double points is taken into account) come to have a
new
application.
56. Orthogonal lines.
DEFINITION.
Two
lines are said to

if
be orthogI* in
onal or perpendicular to each other if and only conjugate points of the absolute involution.
they meet
The following consequences
of this definition are obvious

lines
THEOREM
are the
through any point, 0, pairs of an involution. Through any point there is one and
1.
Tlie
pairs of perpendicular
but one line perpendicular to
a given
line.
A line perpendicular to
Two
lines
one
of two parallel lines

dicular to the
is
perpendicular
to the other.
perpen-
same
line are parallel.

I
DEFINITION.
points,
In case the absolute involution

JT
2,
has two double
Jx and
.jf
they are
through
or
J2
is
the circular points. Any line called an isotropic line or a minimal line.
called
Any isotropic
The
circle (cf.
line has the property of being perpendicular to itself.
circular points are so called because all ordinary points of

.
any
60) are on a conic through I1 and /2 of the conic section referred to in the following
The ordinary
points
lemma
will later be
proved to be on a circle. DEFINITION. homology of period two whose center L is on L, and whose axis I meets L in the point conjugate to the center with regard to the absolute involution, is called an orthogonal line reflection, and
denoted by {LI}. Since the center of a homology is not a point of the axis, the center cannot be a double point of the orthogonal involution, nor can the axis pass through such a point. An orthogonal line reflection is of
is
course a special case of a line reflection as defined in
52.
LEMMA. Let
point of the , as center, passing through l} and having the pairs having of the absolute involution as pairs of conjugate points.
two points not collinear with either double absolute involution. There is one and only one conic, C 2
v
and
le
Proof.
Let
P
2
and the point
at infinity,
be the harmonic conjugate of l with respect to ., of the line OP r Any conic containing
P
1
and having
Let
as center
must contain
,
2,
be a variable point of l m and Y the conjugate of in the absolute involution. Any of the triangles must be self-polar to is the point of any conic satisfying the required conditions. But if
by the definition
of center.
OXY
intersection of the lines

of
PX and P Y,
:
ti
and
the point of intersection
J^X and
1
Y,
POPP.IPQPX, 2 ^
1
and
are
hence
the
points PI
and
harmonically conjugate with respect to
X and
P
Q.
Hence
must
be on any conic with 1 through P

regard to which is the pole of Q
X
Y.
Hence
must be
satis-
on any conic
fying the hypotheses of the lemma.
PIG. 45
the points P, together with Since P^X] -^P 2 [T], tute a unique conic (41, Vol. I) and this conic, by satisfies the condition required by the lemma.
;
P
:
and
2,
consti-
its
construction,
COROLLARY. In case the absohite involution has double points the

conic
C 2 passes
2.
through them.
THEOREM
An
is
orthogonal line reflection leaves the absolute in-
volution invariant.
Proof.
its
2
If
center, let
(7
(cf.
be any point on
the axis of an orthogonal line reflection and L I and not on I. The l any point
as center, and has the Lemma), which contains^, has absolute involution as an involution of conjugate points, must have L and I as pole and polar. Hence, by the definition of pole and polar
conic
44, Vol. I)
(7
/
is
transformed into
having
and
1
as center
and
tfhp.
axis.
itself by the harmonic homology Hence the absolute involution is

linfi
transform en
inf.n ihsfilf
hv
nrth ocron al
reflection
THEOREM
to the axes.
3.
axes are parallel
The product of two orthogonal line reflections whose is a translation parallel to any line perpendicular
Proof.
meet in a point L'
Let the given line reflections be {Lfy and {A/2}- Their axes of l m , and L, and L must be conjugate to L' with 2
respect to the absolute involution.

fore leaves all points
Hence
any
L ~ L f The product therel
on
fc
invariant and also

line
all lines
through
L^.
Hence
it is
a translation parallel to
4.
through
L^.
THEOREM
translation, T, whose center is not a double point of
the absolute involution, is
a product of two orthogonal

of the translation.
line reflections^
is
{Ll}, {Ll^, where
L is
the center
If
trary ordinary point and I may be chosen as OJJ and
the mid-point of the

1
2
pair
an arbiand T (0),
as
PL'
1
.
where L'
is the
with respect to the absolute involution. and /2 as the line joining T (0) to L
double point of the absolute involution is
Or
l^
may
be chosen
conjugate of L as PL'
translation whose center is a
a product of four orthogonal
line reflections.
Proof.
If
\=OL'
and
1
2
=
inva-
the reflection {Ll^ leaves

riant
and {Ll^ carries

to
to
T(0).
Hence the translation {Ll^

carries
{Ll^
3,
Chap. Ill,
If
is
T(0), and, by Theorem identical with T.
p IGto
46
Q = T (0),
invariant.
l^PL'
and \
= QL'
where
carries
the reflection
~ T. Hence, as before, {Ll^ {Ll^ translation whose center is a double point of the absolute involube expressed as a product of two translation with arbitrary L as centers (Theorem 8, Chap. Ill), and hence is expressible
If the axes of
{Ll^
Q and
leaves
tion can
points of
as a product of four orthogonal line reflections.

DEFINITION".
two orthogonal
line reflections intersect
in an ordinary point, 0, the product is called a rotation about 0, and
the point
is
called its center.
THEOREM
5.
A rotation
which
is the
product of two orthogonal
line
Proof.
the point of intersection of ^ and

j&j
is
Let the two line reflections be {LJ,^ and l Since ^ and f
{L^
1 2
and
let
be
are orthogonal,
on
l a
and
on
lx
^.
The product {Z 2 / 2}
Moreover,
{jyj
it is
it is
therefore leaves
and eveiy point
of
invariant.
of period
two on the
axis of either of the line reflections.
Hence
a homology of period
i.e. a point reflection. perpendicular to a line ra, the point of intersection of the two lines is called the foot of the perpendicular /.
two with
as center
and
as axis,
DEFINITION.
If a line
is
A
A
line
is
if
and
said to be the perpendicular bisector of a pair of points it is and its foot is the perpendicular to the line
AB
mid-point of the \yd.\v AB. DEFINITION. A simple quadrangle ABCD is said to be a rectangle if and only if the lines and CD are perpendicular to and B C.
AB
AD
EXERCISES
1.
parallelogram
A B CD
is
a rectangle
if
and only
if
the lines
AB
and
AD are perpendicular.
2.
The The
The
angle
3.
ABC
perpendicular bisectors of the point pairs meet in a point.
AB, BC, CA
of a tri-
perpendiculars from the vertices of a triangle to the opposite sides

point.
lines
meet in a
4.
through the vertices
of a triangle parallel to the transforms

line reflection are concurrent.
of the opposite sides
by a fixed orthogonal
57. Displacements
and symmetries. Congruence. DEFINITION. The
product of an even number of orthogonal line reflections is called a displacement. The product of an odd number of orthogonal line
reflections is called a
symmetry.
THEOREM
subgroup of
6.
The
the
of all displacements parabolic metric group.

set
form a
self-conjugate
(cf.
Proof. That the displacements form a group is evident because 26, Vol. I): (1) the identity is a displacement, being the prod-
uct of any orthogonal line reflection by itself; (2) the inverse of a product of orthogonal line reflections is the product of the same set
of line reflections
taken in the reverse order;
(3)
the product of an
of
even number
of orthogonal line reflections

is,
by an even number
orthogonal line reflections
by
definition, a displacement.
The group
T
of displacements is contained in the parabolic metric
If {Ll} is
an
orth.ogon.al line reflection,

I'
2 a similarity
2"1
transformation,
and L'
1
= 2 (L),
= 2 (1),
I'
then
{.LZ}
is
a harmonic
homology
with L as center and

of
/
as axis.
But
-
since
are paired in the absolute involution, so are L'
and the point at infinity and the point
at infinity of V.
reflection.
Hence 2
{Ll}
^~ l
{L'l'} is
an orthogonal
1
line
If
similar
and A 2 are any two line reflections argument shows that

line reflections
n orthogonal
reflections
of
2A A 2~ = 2A 2~ 2 A 2 -1 Â^ An 2" is a product of A n are orthogonal line whenever A

1
.
I(
and
is is
in the parabolic metric group.
Hence the group
displacements
a self-conjugate subgroup of the parabolic metric
group.
COROLLARY
a
1.
The
set
of all displacements and symmetries

the parabolic metric group.
form
self-conjugate
subgroup of
DEFINITION.
Two
figures
such that one can be transformed into
the other by a displacement are said to be congruent. Two figures such that one can be transformed into the other by a symmetry are
said to be symmetric.
COROLLARY
a,
2.
If a figure
is
F
F
is to
congruent
to
a figure
z,
and Fz
2,
to
figure,
s,
then F^
congruent
F
F
is
is
COROLLARY 3. If a figure l is symmetric with a figure symmetric with a figure 3 , then l is congruent to COROLLARY 4. If a figure l is symmetric with a figure congruent to a figure S> then F^ is symmetric with
and F2
Z,
and Fz
Since translations
and point
reflections leave the absolute invo-
lution invariant, the definition of congruence given in this section includes the definitions in 39 and 46 as special cases. Theorem 6
to the
shows that the theory of congruence and symmetry in general belongs geometry of the parabolic metric group. It must be remem-
bered, however, that the theory of congruence of point pairs on parallel

lines belongs to the affine group.
of
In other words, the part

is
of the theory
of
congruence developed in Chap. Ill the absolute involution.
independent of the choice
In case the absolute involution has double points, the theory of congruence of point pairs on the minimal lines ( 56) is different from that Oil Other liTlftS. As Will armftnr in f.ViP Fnlln-Hrinrr oant.irvno flifl
57]
MINIMAL LINES
is
125
theory on any line which
not minimal is essentially the same as that developed in Chap. Ill on the basis afforded by the group of translations and point reflections. On a minimal line, however, the
set of points [P]
such that
OP is
congruent to
OP consists of all points

the double
.
on this
line except the point 0.
For
let
7 and /2 denote
X
points of the absolute involution, Jx being the one on the line OP g Let Q be a point of the line OI distinct from and from J2 , and let 2
be any point
of
OIl
distinct
from
and from
Tj.
If
be the
orthogonal line reflection whose center is the point at infinity of the line P Q
and whose axis passes through
0,
and
be the orthogonal line

center
is
reflection
whose
of the line
QP
through
0,
($)
= P.
and whose axis passes we have A l (P) = Q and
Hence the
rotation
AA
transforms
P to
P.
Combining
transformations
oftheformA^
with
tions it
translais
clear
that
we have
7. Any pair of points on a minimal pair of points on the same line.
THEOEEM
any
other
line is congruent to
For example, if a mid-point of a pair were defined to be a point G such that A C is congruent to CJ3, we should have that whenever the line AB is minimal, the point C may be any point on this
line different
AB
from
A and B.
The theorems on mid-points in Chap. Ill

cases.
would
that
in general
have exceptional
It is to avoid this difficulty

of
we have adopted
III.
the definition
mid-point given in
40,
Chap.
similar
remark
applies to the definition of ratio of
collinear point pairs in
43, Chap. III. parallelogram ABCD whose sides do not pass through double points of the absolute involution and in which the
DEFINITION.
EXERCISES
Prove that the group of displacements and symmetries could be defined as the group of all collineations leaving invariant the set of all conies obtain1.
able by translations from, a fixed central conic. 2. The parabolic metric group consists of all protective collineations trans-
forming the group of displacements into

3.
itself.
Two
point pairs on nonminimal lines are symmetric
if
and only
if they
are congruent.
The perpendicular bisector of a point pair AB contains all points P such that AP is congruent to BP. 5. The simple quadrangle AB CD is a rhombus if and only if the lines A C and
4.
if the point pair AC congruent to the point pair BD. 1. Specialize the quadrangle-quadrilateral configuration ( 18, Vol. I) to the case where the vertices of the quadrangle are the vertices of a square.
BD are the perpendicular bisectors of the point 6. A parallelogram ABCD is a rectangle if
pairs
BD and A C respectively.
and only
is
58. Pairs of
orthogonal line reflections.
THEOKEM
is
X
8.
If A x
2,
Aa
are three orthogonal line reflections whose axes pass through
a point
line
reflection
(ordinary or ideal), the product whose axis
AAA
8
2
an orthogonal
passes
through
0.
Proof.
In
case
the
three axes are parallel,
the product
translation,
AA
8
is
and
so
by
in the form
Theorem 4 is expressible A4AX where

,
is
an orthogonal line
reflection
whose
to
axis
is
parallel
axes.
the
other
Hence
In case two of the

axes are not parallel, the
third axis
PIG. 48
must
pass
be any point not collinear with through their common point 0. Let and a circular point. Let C z be the conic, existent and unique
according to the
lemma
of 8 56,
which passes through P, has O&s
points.
If
Ql
be any point of
<7
let
\(Q^=Q A
Z,
(Qz )=Q a
this construction the line Q^z is parallel to Q4 QS Q5 Q6 where in case Q = QJt the line Q Qj is taken to mean 2 the tangent to <7 at Q Hence, by Pascal's theorem (Chap. V, Vol. I) or one of its degenerate cases, it follows that Q Z Q is parallel to Q6 Qr Hence
According to
and
QQ
2
to
and
Since
(AÂ
Q l is an
arbitrary point of
2
(7
(AÂ^l.
The transformation
AAA
3 2
is
1
not the identity, because
it
cannot leave
invariant a point, different from 0, of the axis of A 1 unless A 2 = A 8 , and in the latter case the product is equal to A x Since A 3 A 2 A 1 leaves invariant the line Q^Qt (or the tangent at Q I} if Q 1== QJ, it leaves invariant the point at infinity of this line and also the line through
.
perpendicular to
it.
As
AgA^ is
1?
of period two, it follows that it is
an orthogonal
line reflection.
1.
COROLLARY
reflections
If
A A
2,
and
are
any
three orthogonal line
orthogonal
whose axes meet in a point or are parallel, there exists an line reflection A 4 such that A2 A 1 = A 3 A4 and an orthogonal
,
line re/lection
&
such that
Proof.
By
the
Hence
AA =AA theorem, A exists such that A AA = A AA =A A

.
jr
In like manner,
exists such that
AAA = A
8
Hence
COROLLARY
reflections
line,
A A = A A,.
a
1
2. The product of any odd number of orthogonal line whose axes meet in a point or are parallel is an orthogonal
reflection,
Proof.
By
the theorem, whenever
n^
3,
the product of n orthog-
onal line reflections

Of
whose axes
is
are concurrent reduces to a product
_ 2.
Thus,
if
odd, the
number
of line reflections
can be
tiic uj.uu.uuu
i/u
uwu.
3.
JLJ.J.U.O
we
jjo,vc
COROLLARY
reflections is
product of any even number of orthogonal a rotation in case their axes meet in a point, and
2Vie
line
is
translation in case the axes are parallel.
COROLLARY
COROLLARY
4.
5.
An
The
orthogonal line reflection

set
is
not a displacement.
of all rotations having a
common
center
is
a,
comimttative group.
Proof.
reflections
rotation is defined as a product of
two orthogonal
line
whose axes meet in an ordinary point. So, by definition, the identity is a rotation, and the inverse of a rotation A.A is the rotation A A- The product of two rotations is a rotation by Cor. 3.
Hence the rotations having a given point as center form a group. To show that any two of these rotations are commutative amounts
to
showing that
(i)
AAAA= AAAA
line reflections
whenever the A's are orthogonal By the theorem we have
whose azes concur.
and hence
(2)
But
since
A A A = \\\,
8
AAMi^VAA
which combined with
(2) gives (1).
THEOREM
9.
Any
displacement leaving a point
invariant
is
rotation about 0.
Proof.
The given displacement
is
of orthogonal line reflections,
AM
Ar
a product of an even number, n, Let Af be the line reflection

A,..
whose
axis
t
is
the line through

is
parallel to the axis of
Then
the
product
T = AA'
a translation (Theorem 3) and
Thus
where each T^
if is
A, = TA'. An ..A = TA...TA,

a translation.
{
But by
Cor. 2,
Theorem
is
11,
is
any
affine collineation,
T 2 = 2T/,
Chap.
Ill,
where T/
a translation or
the identity.
Hence
But
T,'
since
7l
TI leaves
invariant,
and A^ invariant, the product A{ leave and hence, by Theorem 3, Chap. Ill, is
the identity.
Hence
VAÂ:-..^
,
where
Aj,
0.
A^
are orthogonal line reflections
through
59.
By
Cor. 3,
Theorem
8,
A^
A/
is
whose axes pass a rotation about 0.

Let
be
The group
of displacements.
THEOREM
10.
an
arbi-
trary point.
is
displacement can be expressed in the form PT, where a rotation about and T a translation.
Any
Proof. By precisely the argument used in the given displacement can be expressed in the form
last
theorem the
A' ---A'T' ^-2n M J-2n

L
-
T' xV
reflection
where
i '(-i
= l,
tt
2 n)
is
an orthogonal line
,
whose
axis
passes through 0, and T/ (i
= 1,
2 n)
is
The product T Cor. 3, Theorem

leaves
8,
T{ is, by Theorem 6, A! A{ is a rotation or a 2n
a translation or the identity. Chap. Ill, a translation. By

translation.
Since
it
invariant, it is a rotation.
1.
COROLLARY
Any
T'P', where T' is a translation
displacement can also be expressed in the form and P' a rotation with as center.
COROLLARY 2. Any symmetry is a product of a line reflection whose axis contains an arbitrary point and a translation. THEOREM 11. Any displacement, except a translation having a
double point of the absolute involution as center,
is
a product of two
orthogonal line
reflections.
be an arbitrary point. By the last theorem the given Proof. Let displacement reduces to PT, where T is a translation and P a rotation about 0. If the center, L, of T is not a double point of the absolute
involution, by
Theorem
4,
where ^ and lz meet Z in the conjugate of L relative to the absolute involution and where lz passes through 0. By Cor. 1, Theorem 8, there exists an orthogonal line reflection {Mm} such that
Hence
PT = {Mm}
[Ll
-
{L12 }
= {Mm}
{LI}.
If
is
not the identity,
it is
clear that
and ^ cannot be
is
parallel,
and hence
In case
PT is T is a
a rotation.
translation
it
whose
center
a double point of the

8,
absolute involution,
can be expressed (Theorem
Chap. Ill) as
product of two translations T I} g of the absolute involution. Hence,

rotation,
whose centers are not double points

if
we
not the identity, PT 2 is a In case P is the identity, have the exceptional case noted in the theorem.
is
and thus
PT T
2
is
also a rotation.
COROLLARY.
displacement
is either
a rotation or a translation.
The following two theorems have the same relation to the parabolic metric group and the group of displacements, respectively, that
the fundamental theorem of projective geometry (Assumption P) has
to the projective
group on a
line.
transformation of the parabolic metric group leaving invariant two ordinary points not collinear with a double
12.
THEOREM
point of the absolute involution

or the identity. Proof.
is either
an orthogonal
line reflection
conic through
Denote the given fixed points by and P, and let C 2 be the P having as center and the absolute involution as an
Since
in
involution of conjugate points.
Cz
is
uniquely determined by
these conditions
(cf.
the
lemma
56), it is left invariant
by
the
given transformation F. Now T leaves 0, P, and the point at infinity of the line OP invariant. Hence the line OP is point-wise invariant,
and every
2
<7
is also
C2
perpendicular to it is transformed into itself. Since and each of the lines perpendicular to OP meets in at most two points, F is either the identity or of period two.
line
I
invariant
If of period
two,
13.
it is
evidently an orthogonal line reflection.
and a> displacement leaving invariant a point line I containing but not containing a double point of the absolute involution is either the identity or a point re/lection with as center.
Proof.
2
THEOREM
Let
P be
any ordinary point
of
distinct
from
0,
and
let
C" be the conic
through
P having
<7
as center
and the absolute
invo-
lution as an involution of conjugate points.

invariant, being a product of
displacement leaving
axes meet in 0,
must leave
it
two orthogonal line reflections whose invariant. Hence it either leaves P
invariant or transforms
meets
C2
into the other point in which the line OP In the first case the transformation must, by Theorem 12
and Cor.
4,
Theorem
8,
given displacement, which
reduce to the identity. In the second case the we shall denote by A, multiplied by the
whose axis is the line through orthogonal line reflection dicular to OP, leaves invariant. Hence, by Theorem 12,
perpen-
where A'
is
a line reflection having
OP as A = A' A.
axis or the identity.
Hence
Since
the axes of
cannot be a line reflection, A' cannot be the identity. Since A and A' are perpendicular, A is a point reflection.
EXERCISES
1.
A displacement
which
carries a point
to a point
B and
has a point
(ordinary or not) as center

an.
is, if
the line
is

2.
whose axis
OA is not minimal, the product of OA followed by one whose axis is
the line joining
DA
to the mid-point of the p&ixAB. the perpendicular bisectors of the point pairs AB, JBC, CD, of a simple quadrangle meet in a point, the fourth perpendicular bisector
If three of
passes through this point. *3. Any affine transformation

line reflection
which
leaves a central conic invariant is a
whose center and axis are pole and polar with regard to the conic or a product of two such line reflections. *4. In case the absolute involution is without double points, the group of
displacements can be defined as the group of transformations
common
to the
parabolic metric group and the equiaffme group. Thus two ordered point triads are congruent if they are both equivalent and similar. Develop the theory of congruence on this basis, and show what difficulties arise in case
the absolute involution has double points.

of all points [P] such a fixed point, are all congruent to a fixed point pair OP , provided that the line OP does not contain a is called the double point of the absolute involution. The point center of the circle.
60. Circles.
DEFINITION.
A circle is the set

is
that the point pairs
OP, where
Since the displacements form a group, it is clear that may be any of the points P. It has already been proved ( 57) that if the line OP contained an invariant point of the absolute involution, the
one
set [P]
would consist
14.
of all ordinary points, except 0, of the line
OP
THEOREM
points.
A circle consists
of the ordinary points of a conic
sec-
tion having the pairs
of the absolute involution as pairs of conjugate The center of the circle is the pole of l x with respect to the circle.
Proof.
circle.
Let
be the center of the
circle
and
any point of
the
The circle consists of all points invariant. If one ments which leave
each of these displacements
(Cor. 1,
is
obtainable from
P
Q
by
displace-
of the line reflections
of which
a product be taken to have
OP
as axis
Theorem
0.
8), it
follows that the circle consists of the points
through
the
by orthogonal line reflections whose axes pass But the system of points so obtained is identical by construction with the ordinary points of the conic referred to in
obtainable from
lemma
of
56.
COROLLARY. In case
tlie
absolute involution has
no double points,
every circle is a conic section.
In
case the circular points exist, they

conic section.
and
the points
of any
circle
form a
conic, with regard pairs of the absolute involution are pairs of conjugate points, form a circle. 15.
to
THEOREM
which
The ordinary points of any proper
the
conic C" with regard to which the pairs of the absolute Proof. involution are conjugate points cannot be a parabola, since all points
s of L are conjugate to the point of contact of a parabola. Hence as center. Let be any point of C\ By has an ordinary point definition there is one and only one circle through which has as
a conic through having O as center and the pairs of the absolute involution as pairs of conjugate 56 there is only one such conic. Hence points. By the lemma of
a center.
By Theorem
14, this circle
is
the circle through
P with
as center contains the ordinary points of
C\
THEOREM 1 6. Three noncollinear points, no two of which are on a minimal line, are contained in one and only one circle.
Proof.
Let the three points be
P P v
at infinity of the line
PP^
Q
and
and z Let be the point , the perpendicular bisector of the

.
P^. The polar of L x with regard to any circle through and must, by Theorem 14, pass through the mid-point of the conjugate of L a in the absolute involution. Hence the polar of
point pair
and
P^
Z. with regard to any

manner, the polar
circle
through
P
G 2
and
must be
I.
In like
of the point at infinity
Jfw
of the line JJJ*
with
regard to any circle containing P^ and bisector of JJJ* Since the pobts ,
P must be the perpendicular P P and v P are not collinear,

2 I
L^M, = L
by
with regard to any circle through P and z Since, 0) lf one and only one circle through with as and P^. center, there cannot be more than one circle through , i}
.
definition, there is
P P P P
Since the product of the orthogonal line reflection with OP as axis
by that with
0-ZjJ
as axis
transforms the point pair

into the point pair
OP
l}
P with
the circle through as center con-
tains P^. Alike
argument shows that it contains J^.
Hence there is one circle

containing
P P
,
l}
and
z.
Observe that we do
.
not prove at this stage that a circle has a point on every line through its center. This could not be done without further hypotheses on the nature of the plane than we are making at present.
49
EXERCISES
locus of the points of intersection of the lines through a point .4 with the perpendicular lines through a point B, not on a minimal line through
1.
The
A,
is
circle
whose center
is is
2.
A tangent to
Any two
If the
a circle
the mid-point of the pair AB. perpendicular to the diameter through the point
of contact.
3.
4.
conjugate diameters of a circle are orthogonal. tangents at two points A and B of a circle meet in a point 0,
the pairs
5.
OA
I
and
If
is
through
6.
and
OB are congruent. the perpendicular bisector of a point pair AB, then the circles B meet I in pairs of an involution whose center ( 43) is the
of all circles having a
mid-point of
pairs of
7.
AB. The system
common
center
meet any line in the
an involution.
A parallelogram
which circumscribes a
inscribed in a circle
is
circle
must be a rhombus.
8.
A parallelogram
two
circles
is
a rectangle.
9. If
have two points in common, the pair of tangents at one

sides
common
10.
of an
point
The
feet of the perpendiculars
symmetric to the pair of tangents at the other. from any point of a circle to the
arp
r>.nllirpn.r-
iriHP.Tiliprl trria.ncrlA
136
EUCLLDEAJSI
rLAJN.tt WJSUJM.JiiT.Ki
LCHAP.IV
If the coordinate
conjugate points
(3)
and (0, 0, 1) are system be chosen so that (0, 1, 0) in this involution, the bilinear equation reduces to
ax^-i- cxzx2
is paired (0, 1, 1)
0.
Here the point
with the point
(0, c,
a).
In
case the involution contains
two
pairs of points
be chosen as (0, 1, cally conjugate, one pair may In that case (3) reduces to the other pair as (0, 1, 1) and (0, 1, 1).
which are harmoni0) and (0, 0, 1; and

'
W
For the rest of
I~T~
2~~
j
this section
we assume
that the absolute involution
contains two pairs of points which are harmonically conjugate with exist in every plane satisrespect to each other. Such involutions
fying Assumption
Q,
since
any two
this
distinct collinear pairs of points
determine an involution. Hence
assumption
is
no restriction on
the nature of the plane in which we are working. It is, moreover, easy to replace the formulas which we shall obtain from (4) by the
more general but more cumbersome formulas based on

The equations
of the transformation required to
(3).
change (3) into
(4) are
Hence it is clear that in the complex geometry ( 5) every involution may be reduced to the form (4), and in the real geometry only those involutions can be reduced to this form which are such that a/c > 0. The involutions of the
latter type are direct
(18).
group are
,
The equations
of the affine
5
(5)
and
ojj,
if
the involution (4)
is
to be transformed into itself, all paira
xz and xlS xz which

also satisfy
satisfy
must
^+^,=
(a^+ 5^) (a^+ \x + (0,^+ & A (a&+ l^ =
2)
)
0,
which
is
2
the same as
2
Z>
(a,
+ a|) x& + (a& + a^ (x^ + x^ + (& +
2)
xf t -
0.
are the necessary and sufficient conditions that (5) leave (4) invariant. Combining these two equations, we obtain
X+
or
a}
- ZX - aî =
2
Thus we
infer
a2 =
5
1
and a t = T
Hence
group are
THEOREM
(6)
19.
The equations of
the parabolic metric
where
= 1.
If-
Any
conic section has an equation of the form
(66,
Vol. I)
which determines on the

elements
satisfy
2
,
line
XQ
an involution whose double
2.0
_a
Comparing with
(4),
we have
~~
11
that a circle must satisfy the condition
22
>
a iz ~
i.e.
If this circle is to
have
(1, 0, 0) as center,
as pole of #
= 0,
the
equation
(7)
must
also satisfy the condition
Thus, returning to nonhomogeneous coordinates, the equation of a

circle
(8)
with the
origin, as center
must be
7c.
of the
form*
xz -+-y 2
59, the transformations of the parabolic metric According to group leaving such a circle invariant are all displacements or sym-
metries, and, moreover, all displacements and symmetries leaving the origin invariant leave this circle invariant. Substituting (6) in (8), we
see that a displacement or
of
symmetry
,
leaving the origin invariant
is
the form
* This argument does not prove that every equation of
this
form represents a
Since any displacement or symmetry is expressible as the resultant of one leaving the origin invariant and a translation (Theorem 10,
Cor. 1),
we have
20.
THEOREM
The equations of
the
group of displacements and
symmetries are
(9)
^'
where a?
Z
-j-
{3
1
I,
and
=< = 1.
(9) effects
By
54, Vol.
if
a transformation of the form

e
an
involu-
tion on lx
and only if
1.
By Theorem 10,
Cor. 2,
any symmetry
the
1.
leaving the origin invariant is a line reflection.
Hence
THEOREM
type
The displacements are the transformations of which e = 1 and the symmetries those for which e = (9) for
21.
EXERCISES
1.
The equation
of
circle containing the point (oâ) an(*
having the
point (a L lj) as center is
2.
Two
if
lines
ax
only
aa'
W=
ly
and
a'x
b'y
c'
are orthogonal
if
and
0.
3. Tn case the absolute involution has double points, the equiaffine transformations of the parabolic metric group are of the form (9), where a2 + jp =
and
1.
63.
Introduction of order relations.
Let us
now assume
that the
plane which
of
15.
larly of
satisfies
an ordered plane in the sense We may therefore apply the results of Chap. II, particu28-30. Let us also assume that the absolute involution
are considering
is
we
the condition referred to in
62, that there exist
two
pairs
conjugate with regard to the absolute involution which separate each other harmonically. By Theorem 9, Chap. II, and ita corollaries, it follows that any two pairs of the absolute involution
of points
separate each other,

points.* This result
and that the absolute involution has no double

:
may conveniently be put in the following form THEOREM 22. Two pairs of perpendicular lines intersecting in the same point separate each other. No line is perpendicular to itself.
*
The geometry arising from the hyperbolic case has been studied by Wilson and
The
ments
restrictions
which we have
just introduced enable us to state
the fundamental theorem (Theorem 13) about the group of displacein the following
more
precise
form
THEOREM
the identity.
23.
The only displacement leaving a ray invariant
is
be the origin and Proof. Let any point of the ray. Since any collineation preserves order relations, A is transformed into itself. Since the
line
AB is invariant, the displacement is a point reflection or the identity 1 3). But a point reflection would change B into a point of the ray opposite to the ray AB, and thus not leave the ray AB invariant.
(Theorem
With the aid of this theorem we can complete the set of fundamental theorems on congruent triangles, the first two of which were
given in
61.
THEOREM
to the
24.
Two
triangles
point pairs AB,
AC
and
the angle
point pairs A'B' , A'C'

Since the angle*
ABC and A'B' C' are congruent if the & CAB are congruent respectively and the angle & C'A'B'.
X
Proof.
& CAB is congruent to the angle 4- C'A'B',

A
carrying
there exists a displacement
A to A'
and the rays
AC
is
and
con-
AB
to
A'C' and A'B' respectively. Since the point pair

is also
AB
gruent to A'B', there

to B',
and since
AC is
a displacement A 2 carrying A to A' and B congruent to A'C', there is a displacement A 8
carrying A to A' and C to C'. By Theorem 23, Hence the displacement a carries the triangle
ABC to A'B'C'.
\
if
and
\= A
EXERCISES
and A'B'C' are congruent triangles ongraent to the point pair A'B' and the angles 4.
1.
Two
ABC
CAB
the point pair AB is and 4- CBA are con-
gruent respectively to the angles 4. C'A'B' and 4- C'B'A'. and A'B'C' are congruent in such a 2. If two triangles to B', the angles 4. ABC, 4-BCA, 4corresponds to A' and
ABC
B
way
that
CAB
are con-
gruent to the angles

3. If
A'B'C', 4-B'C'A', 4- C'A'B' respectively. two triangles AB C and A'B'C' are symmetric in such a way that A corresponds to A' and B to B', the angles 4. ABC, 4-BCA, 4-CAB are congruent to the angles 4- C'B'A', 4-A'C'B', 4-B'A'C' respectively. 4. Let A, B, C be three collinear points and P x the point at infinity of the
line joining
&
them
is
between
A and C if
and only
if
5.
An
orthogonal line reflection interchanges the two sides of
its axis.
140
[CHAP. IV
64. The real plane. Let us finally assume that we are dealing with the geometry of reals. In consequence, we have the theorem 4) that any one-dimensional projectivity which alters sense (i.e. for (
which
A < 0)
has two double elements.
This
may
be put into the
following form as a theorem of the affine geometry.
THEOREM
line
I,
25.
If A^ and
A2
ellipse
meeting the line in two points.
are any two points of an ellipse, any in a point between A and A meets ihb i Z
,
Proof.* Let us denote the given ellipse
by
Let
z
,
and
let
be a
it.
variable
point
on
and
l
be the
I
points
**
in
which
is
met by A,1 A and A 2 nA respectively, and let Q t

and Q 2 be the points in which l a is met by AÂ
and
Also
point
AA
Z
respectively.
let
be
the
in
which A^L Z
meets L.
tion,
By construcdefi-
and by the
\
PIG. 51
nition of a conic,
A,
The projeetivity [$Jx[02] is direct, because, by the remark at the beginning of this section, if the projectivity altered sense it would have two double points, and these, by the definition of the projec-
would be points of intersection of ln with 2 hypothesis that JS is an ellipse. Let C and Ca be the points of intersection of A
tivity,
E
t
z
,
contrary to the
Az
I.
with
and
respectively.
Also let
LM
be the point at infinity of
Then, by the
hypothesis that
is
between
and A
THE REAL PLANE

But, by construction,
A CA
1
C^
Hence, by Theorem
6,
Chap.
II,
But the
points C.,
L x Q2
,
are carried to (7,

(10).
L x Q8
,
respectively,
by
the projectivity [QJ-^[QJ, indicated in

site.
Hence
the projectivity
is
[#J 7\[# ] From

8
is opposite.
Since
[Q^[QJ
it
is direct,
[QJj;[QJ
oppo-
this, since
Q 1 and Q a
are carried
AI as center
to
and
respectively,
by a perspectivity with follows (Theorem 6, Chap. II)
that the projectivity

is opposite.
By the remark at the beginning of the section this promust therefore have two double points, and by the definition of the projectivity these double points must be points of intersection of I with E\
jectivity
"by
COROLLARY 1. The points in which I meets the ellipse are separated A I and A z relative to the order relations on the ellipse.
Proof.
Let Z^ and Z>2
(fig.
51) be the two points in

,
which
meets
the
A, A I} A Z etc. have the meanings given them in is the proof of the theorem. Then since the projectivity -^- [Z/J
ellipse,
and
let
[J
opposite,
S(D 1 D2 L 1 ) =jt S(D 12)2 L2 ).
Hence
the lines
AD
and
AD
separate the lines
AA
and
AA
2,
which,
according
to the definition in
20, implies that the pair of points

ellipse.
I
DJ)^
separates the pair
AÂ 2 on the
COROLLARY
sides
2.
The points in which
meets the ellipse are on opposite
of
the line
AÂ f
and A^D^ Hence, if A' denote the point in which a meets J\>2 D { and X>2 separate A' and C. Now A is not between Z> and X>2 because if it were, the line a would meet x is between the ellipse in two points instead of only in one. Hence a and
:
2 1
r
,
Proof. Let a be the tangent at A separate the lines A 1
A r By
the
first
corollary the lines
X>x
and
Z>
2,
and hence D^ and Z>2 are on opposite sides

26.
of
I.
THEOREM
A rotation which
transforms a given
circle into itself
be any other Let the given triad of points be A, B, C, let circle, and let A*, B K <? be the points at infinity of the lines OA, OB, OC respectively let 0',A',B', C', A' X ,B' X} CL be the points to which 0, A, B, C, A x Bx (?,, respectively, are carried by the given
Proof.
point of the
rotation
let
A",
B'J,,
C" be the points
at infinity of the lines
OA' ,
OB',
OC' respectively.
The given rotation effects on la a transformation which is the prodr uct of two hyperbolic involutions. Hence S(A x B x Ca } S(ALSa,CL). As in the proof of Theorem 25, the projectivity A^BLCl-Â'^B^C'l
is
direct because otherwise it

circle
would have double points and these

a>
would be common to the

and, therefore,
S(A.S a C) =
and L. Hence S(A' B' x C' x )=S(A'lB'lC'l} Projecting from 0, we have, S(A%2%
(,').
by the definition of sense on a conic
20), that
S(ABC) = S(A'B'C'}.
Theorem 26, which is here proved only for a real space, can be proved for any ordered space by the methods of the next chapter. This theorem states one of the most intuitionally immediate properties of a rotation. In fact,
most
of
the older discussions of the notions of sense describe sense, without
further explanation, as "sense of rotation."
EXERCISES
1.
If 21
A OB
is
any
angle,
PR
to
on a given side of the line
PQ any ray, there is one and only one ray PQ such that 4 A OB is congruent or symmetric
and
is
4 QPR.
*2. Prove that
Theorem
25
not true in a space satisfying Assumptions
A, E, H, Q. 65. Intersectional properties of circles.
are
any two
distinct points, then

is
origin there
THEOREM 27. If A and B on any ray having a point as one and only one point P such that the pair AB is
pair OP.
be the point to which
congruent
Proof.
to the
Let
J5
B is
carried
by the translation
as center contains
which
all
carries
A to 0.
The
circle
is
through B l with
points
Q such
to
which
is
l
congruent to AB. Let B2 be the point transformed by a point reflection with as center.
that
OQ
and B z any line I through (and distinct two points, according to Theorem 25. But by Theorem 23 neither of the rays on I which have as origin
Then since
0-Sj)
is
between
from
must meet the
circle in
as rays contains just one point of the circle. Hence each, ray with such that AB is congruent to OP. origin, contains a single point
Combining
this
theorem with Theorem 23, we have

is
THEOREM
given ray
to
28.
There
one
and only one displacement carrying a
a given ray.
way
result characterizes the group of displacements in the same that the proposition, that there is a unique projectivity of a one-dimensional form carrying any ordered triad of elements to any
This
ordered triad characterizes the one-dimensional protective group.
THEOREM 29. If two circles are such that the line joining their centers meets them in two point pairs which separate each other, the circles' haw two points in common, one on each side of the line joining
the centers.
Proof. Let the two circles be C? and C2 and let them meet the and 2 QZ respectively. Let A line joining the centers in the pairs be the center ( 43) of the involution F in which : Q^ and 2 Q2 are pairs, and let a be the
,
P^
perpendicular to the line at A.

^
separate
P
z
and
the
ordered
triads
P^Q^ and
same
Q^Q
are in the
sense.
inter-
The
involution
changes these two triads and hence transforms any

triad into a triad in the
FIG. 52
between P and Q r Hence, by Theorem 25, the 1 and by the second line meets the circle C* in two points A^ and A 2 corollary of this theorem, A l and A 2 are on opposite sides of the line P^v and Q l are the ends The lines AJt( and A^Q V are orthogonal since P :
same
sense.
Hence A
is
of the diameter of a circle through A^. The line AÂ is orthogonal to the line through l parallel to P^Q V Hence the involution F is per-
spective with the involution of pairs of orthogonal lines through Hence v z and Qz are on a circle whose center is on the line
A^
A P
By Theorem
in
16 this
circle
must be
<72 .
common.
similar arerument shows that
Hence C\ and C* have A a is on C? and C2.
o ^_ J Euclidean geometry. Hence the geometry of the parabolic metric group in a real plane is the JSuclidean geometry.
lates
by considering a set of postufrom which the theorems of Euclidean geometry are deducible and proving that these postulates are theorems of the parabolic metric geometry. It then follows that all the theorems of Euclidean
geometry are true in the parabolic metric geometry.
This result can also be established
As a set of assumptions for Euclidean geometry of three dimensions we may choose the ordinal assumptions I-IX which are stated in 29,
together with the assumptions of
congruence (X-XVI) stated below.
For our immediate purpose, however, a set of assumptions for Euclidean, plane geometry is needed. To obtain such a set we merely replace
VII and VIII by the following

VII. All points are in the
same plane.
Euclidean plane geometry
is
Thus our
set of postulates for
I-VI,
vn, ix-xvi.
Assumptions
to
X-XVI make use of a new undefined relation between
ordered point pairs
CD"
It
when
this
which is indicated by saying "AB is congruent must be verified that the new assumptions are valid relation is identified with the relation of congruence
B, then on
defined above.
X. If
one
any ray whose origin

that
is
and only one point

is
D such
to
AB
is
congruent
a point to CD.
there is
Proof. This
the same as
Theorem
27.
is
XI. If AB
is
is
congruent
CD and CD
congruent toEF, then
AS
congruent
to
EF.
is
Proof. This
a consequence of the fact that the displacements
form a group.
XII. If AB
is
congruent
to
A'B',
and
BC is
to
{ABC} and
Proof.
carries
{A! B'C'}, then
AC is
B
1
congruent
congruent A'C'.
to
B'C' and
By Theorem
and
28, there is
to
A! and
respectively.
a unique displacement which This displacement carries
order relations. Moreover, the point C so obtained is such that BC is congruent toB'C' and A C to A'C'- and, by Theorem 27, there is only
one point C' in the order {A'B'C'} such that
BC is
congruent toB'C'.
AB is congruent to BA. Proof. AB is transformed into BA

XIII.
center
is
by the point
reflection
whose
the mid-point of AB.
XIV. If A, B, G are three noncollinear points and D is a point in {BCD}, and if A'B'C' are three noncollinear points and D' is a point in the order {B C'D'} such that the point pairs AB, BC, CA, BD are respectively congruent to A'B B'C', C'A', B'D', then AD
the order
1
is
congruent
to
A'D'.
Proof. Since AB is coiigraent to A'B', there exists a displacement A which Let A (C}=C^, carries AB to A'B'.
A(D)=D
Also
let <72
and
Z>
be the
are transpoints to which Cl and l formed by the orthogonal line reflection
having A'B' as
axis.
According to
57, the pair

;
BC is con;
gruent to B'Cl and to
B'CZ CA to CÂ' B'D2 and and CyA'; '; BD to B'D l and AD to A'D, and A'Di It follows that C' must coincide with Cl or C2 for
.
JL
PIG. 53
otherwise there would be two
circles,
one with A' as center and the
other with B' as center, containing the three points

If C'
G Cz
I}
C'.
Cj,
it
follows,
by Theorem
23, that
D'
=D
and hence that
AD is
congruent to A'D'. If
C"= C2
are
it follows,
similarly, that
D'=D Z
and hence that

DEFINITION.
of points
AD
If
is
congruent to A'D'.
and JF
two points
of a plane a,
then the set

a
circle.
\X] of a such that
OX is
congruent to
OX
is called
them XV. If the line joining the centers of two coplanar in pairs of points, J^Q l and 2 Q 2 respectively, such that {P^Q^ and each side of {P\Q\Q<i$> Mle circles have two points in common, one on
circles meets
the line joining the centers.
Proof.
This
is
the same as Theorem 29.
JB
three points in the order are points in the order {ABB^, {AB^B^, congruent to each of the point pairs BB^ B^BZ
-
XVI. If A, B, C are
{ABC} and
such thai
,
13
S,
AB
JB l
J3
2,
is
then there are not
more than a finite member of

Proof.
the points B^,
B2
between
A and
the
C. is
Let
Bn be
the point at infinity of the line
AB. Then
,
the harmonic conjugate of
with respect
to
B and Bm B
;
is
har-
monic conjugate of B with respect to B l and BM and so on. Thus A, JB, form a harmonic sequence of which J? M is the limit-point. B I} J52
,
Since
C has
a finite coordinate, the result follows from
8,
Chap. I
provides
in
The
set of
assumptions
I-XVI
is
not categorical.
It
merely for the existence of such irrational points as are needed

constructions involving circles and lines (see
77, below).
It
can
be be
made
categorical
by adding Assumption XVII,
29.
It
must
noted, however, that
when XVII is
added,
X-XVI
become redundant
in the sense that it is possible to introduce ideal elements and then bring in the congruence relations by means of the definitions in this
and the preceding chapters. In order to convince himself that the assumptions given above are a sufficient basis for the theorems of Euclid, the reader should carry
out the deduction from these assumptions of some of the fundamental theorems in Euclid's Elements. An outline of this process will be
found in the monograph on the subject from which the assumptions
have been quoted.* In making a rigorous deduction
of the
theorems
of
elementary
geometry, either from the assumptions above or from the general projective basis, it is necessary to derive a number of theorems which
mentioned in Euclid or in most elementary texts. These are mainly theorems on order and continuity. They involve such matters as the subdivision of the plane into regions by means of curves, the
are not
areas of curvilinear figures, etc., all of
applications of
geometry to
analysis,
which are fundamental in the and vice versa. In so far as
these theorems relate to circles, they have been partially treated in 64-65 and will be further discussed in the next chapter. The
methods used for the more general theorems on order and continuity, however, are less closely related to the elementary part of projective geometry and will therefore be postponed to a later chapter.
* Foundations of Geometry,
by Oswald Veblen,
in
Monographs on Modern
67. Distance.
In
43 we have defined the magnitude
of a vector
OB
as
its
ratio to a unit vector
OA
collinear
with
it
but in the
affine
lutely unrelated.
!
geometry the magnitudes of noncollinear vectors are absoIn the parabolic metric geometry we introduce the additional requirement that any two unit vectors OA and A' shall be such that the point pair OA is congruent to the
point pair O'A', Thus, if a given unit vector
OA is
fixed
and
(7
is
the circle through
any other unit vector must be expressible in the form Vect (OP), where P is a point of the circle. This gives two choices for the unit vector of any system of collinear vectors, and each of the two possible unit vectors is the negative of the other.
as center,
with
Therefore, while it is possible under our convention to compare the absolute values of the magnitudes of noncollinear vectors, there is
no relation at
classes on
all
the fact that there
between their algebraic signs. This corresponds to is no unique relation between particular sense
must,
of noncollinear vectors appear they state theorems of the Euclidean geometry, be such that their meaning is unchanged when the unit vector on any line is reif
Formulas
two nonparallel lines. in which the magnitudes
placed by its negative. This condition Exs. 2 and 4, 71.
is
satisfied, for
example, in
affine
The
ratio of
two
collinear vectors is invariant
under the
group; the magnitude of a vector is invariant under the group of translations; but the absolute value of the magnitude, of a vector,
according to our last convention, displacements. The last invariant
of
is
may
invariant under the group of be defined directly in terms
point pairs as follows
DEFINITION.
shall
Let AS be an arbitrary pair
of distinct points
which
be referred to as the unit of distance. If P and Q are any two points, let C be a point of the ray AB such that the pair A is
congruent to the pair PQ.
The
ratio
AC_
AB
is
called the distance
from
line,
to Q,
and denoted by Dist (PQ}.
If
1>
is
any point and
any
the distance from
to the foot of the
"*'""'
tit
P= Q.
corresponding theorems on the magnitudes of vectors there follows the theorem that if {^41? C}, then
the
From
Dist (AJS)
Dist (BC)
= Dist (AC).
may
be defined
}
Other properties of the distance-function are stated in the exercises.
The
follows
notion of the length (or circumference) of a circle Let P v P2 , n be n points in the order {P-jP z
: ,
a;
Pn
on a
circle.
and
It
all
let
j+
Dist
(Pf z ) +
circle
2
,
Dist (Pn PJ.
can easily be proved that for a given
C
,
2
,
the numbers
,
p obtained from
do not
possible ordered sets of points exceed a certain number.
Pv P
Pn
for all values of n,
DEFINITION. The number c, which is the smallest number larger than all values of p, is called the length or circumference of the circle C 2 The proof of the existence of the number c will be omitted for the reasons
.
explained below. The existence of c having been established, it follows without difficulty that if c and c' are the lengths of two circles with centers O and
(y, respectively,
and passing through points P and P', respectively,

c__ Dist (OP) c'~ Vist(O'P')'
stant
Choosing the point pair O'P' as the unit of distance and denoting the conc' by 2 TT, this gives the formula
(11)
c
TT
Dist (OP).
The theory
of the lengths of curves in general
could be developed at the
present stage without
essential difficulty. This subject, however, is yery different (in respect to method, at are least) from, the other matters which
any
we
considering, and therefore will be passed over with the remark that, starting with the theory of distance here developed, all the results of this branch of geometry may be obtained as applications of the integral calculus. Even the
theory of the length of circles which we have summarized in the paragraph:: above involves the ideas, if not the methods, of the calculus.
EXERCISES
1.
Two
If
point pairs
AB and CD
are congruent
if
and only
if
Dist
(AB) =
Dist (CD).
2.
A, B,
C are
noncollinear points, Dist
triangles to B', spends to A',
3.
Two
ABC and^'JS'C' are

and
similar in such a
if
(AB) + Di?t (BC) >Dist (A C). way that A correDistQBC)

Dist (B'C')
C to C'
if
and only
DistQO) _
Dist (A'B')
Digt (4 C) Dist (A' C')
__ ~
'
4. Relative to a coordinate
system in which the axes are at right angles,

is
~)
the distance between two points (x v yj, (xz , yz
VC*! -*)' +
(ft
-fc)*,
the positive determination of the radical being taken. The distance from a is the numerical value of point (x^y^) to a line ax + ly + c =
ax l
&ft
2
+
2
I)
Va +
68. Area.
of
The area
of a triangle, as distinguished
from the measure

:
an ordered point
DEFINITION.
triad,
may
be defined as follows
Relative to a unit triad
OPQ
49) such that the
lines
OP and OQ are orthogonal and the point pairs OP and congruent to the unit of distance, the positive number
m(ABC}\
called the area of the triangle C, and denoted by a (., As was brought out in Chap. Ill, the theory of r~~ -----ir belongs properly to the affine geometry. Bu'"
for the area of a triangle in
OQ
are
is
AS
terms
of base a"
involves the ideas of distance and perpendicularity and lience belongs to the parabolic metric geometry. It should be noticed that this
of the triangle which is regarded as the base does not pass through a double point of the absolute involution. This condition is satisfied under the hypotheses of 63, 64,
formula assumes that the side
but
of
whereas the definitions is not always satisfied in a complex plane equivalence and measure as given in Chap. Ill are entirely free of
;
such
restrictions.
of areas in general depends
The theory
on considerations of order and
continuity which
we have not yet
developed,
and which,
branch
of
like the
theory of
lengths of curves, belongs essentially to another
geometry than that
with which we are concerned in this chapter. shall, however, outline the definition of the area of an ellipse from the point of view of elementary geometry, because the derivation of the area of an ellipse from that of the
circle affords rather
We
an interesting application of one
of the
theorems about
-
the affine group.
be any finite number of points in the order [PiP2 2 ,'->,P n on an ellipse z with a point as center, and let
Let
Pv P
Pn
A=
can easily be proved that there exists a finite number, a("2), which is the smallest number which, is greater than all values of A formed according to
It
2 DEFINITION. The number a(E ) is called the area of the 2 In case E 2 is a circle, C it is easy to prove that
,
ellipse.
where
IT is
the constant denned above and r
Now suppose E'2 is an ellipse with two perpendicular conjugate diameters 2 OA and OB which meet E 2 in A and B respectively, and let C be the circle as center, and let C be the point in which the ray OB through A with Ob meets C 2 The homology T with OA as axis and the point at infinity of the as center, which transforms B to C, is an affine transformation carrying 2 z This homology transforms the triangle OAB to circle C' ellipse E to the the triangle OA C; and the
.
.
areas
of
these
triangles
satisfy the relation.
a(OAC} = T>isk(OC) ^ Dist (OB) a (OAB)

It follows,
by
50, that the
homology transforms any triis angle into one whose area

as large. By the definition of the area of an
k times
ellipse, therefore,
a(C
a
(
)
2
_ ~
Dist (0(7) Dist (OB)
'
Denoting Dist (OA) by a and

Dist (OB) by
&,
this gives
Fia.54
EXERCISES
is equal to value of the measure of a point triad A to Disb(AB) -Dist(CC'), where C' is the foot of the perpendicular from C the line AB. is a 2. If abed is a simple quadrilateral whose vertices are on a conic and variable point of the conic, -r,fr> \ TV j.^r> \ Pist (Pa) Dist (Pc)
1.
The numerical
BC
Dist (Pb)- Dist (Pd)

is
a constant
3.
(cf.
Ex.
2,
51).
a,
a projective collineation carries a variable point 6 to M', a', V respectively, the number
If
M and two
fixed lines
Dist(Jlfa)
Dist(MV)
W68.69J
4. Let If
ANGULAR MEASUKB
F be the center
of a
151
P is
homology T and
the vanishing- line,
a variable point and
Q = F(P),
where
5.
Jfc
is
a constant.
area of
an ellipse is ira/2, where a is the area of any inscribed parallelogram whose diagonals are conjugate diameters. 6. Among all simple quadrilaterals circumscribed to an ellipse, the ones whose sides are tangent at the ends* of conjugate diameters have the
least area.
7. Among all simple quadrilaterals inscribed in an ellipse, the ones whose vertices are the ends of conjugate diameters have the greatest area. 8. Of all ellipses inscribed in a parallelogram, the one which has the lines
The
joining the mid-points of opposite sides as a pair of conjugate diameters has
the greatest area.

9. Of all ellipses circumscribed to a parallelogram, the smallest having the diagonals as conjugate diameters.
is
the one
69.
The measure
of angles.
The
unit of distance
may be
chosen
arbitrarily, because
any point
pair can be transformed under the par-
abolic metric group into any other point pair. It is otherwise with angles or line pairs, because, for example, an orthogonal line pair can-
not be transformed into a nonorthogonal

of
pair.
measurement
for angles obtained
by choosing
Therefore the systems different units are,
in general, essentially different.
We shall give an outline of the generally adopted system of measurement, basing it upon properties invariant. of the group of rotations leaving a point z be an arbitrary point different from 0, and C the circle Let
as center. Let 1 (fig. 55) be the point different from through P with E in which the line meets C2 and letPj and P be the points in which s
the perpendicular to
P
Q
at
meets
2 (7
.
By
Cor. 1,
circle.
points are in the order
{P^P^PP^ 2 meets C in two points which r Any line through segment P P,P and P are separated by P v and hence meets a- in a unique point. Q
Let P, be the point in which the line through
on the
Theorem 25, these Let a denote the
perpendicular to be the set such
PP, *
meets
is
o-.
And,
in general, let [P^],

2*
= 1,
2,
that P.
the noint in which the line through
Deraendicular to
RP,
The
pair
line
OP
25,
obviously meets the line

is
PP, and
the mid-point
between
PP in the mid-point of the P and 7J. Hence, by Cor. 1,
Theorem
lation
we
re-
have the order
{P'P^Jp,
moment,
where P' denotes,

for
the
the point not on <r in which the

line
OP
meets the
Since
is
circle.
between .P and P',

the same corollary
Since Pis on the
segment a, we have
either
{PPPP}
The
these
FIG. 55
of
or
{P.PPP}.
second
alternatives, ever,
how-
trary
when combined with {P'PPP}, would imply {P'^PPP}, conto {^PP^P'}. Hence {PPPP} is impossible, and we must
In
like
,
have {PjPPPj.
general, that
manner
it is
proved that {PPPP}- and, in

>J
pp
(a
pp
^'^-tJ_''
Let II denote the rotation
leaves
fixed
Ti'-
point reflection in this case) which
and transforms
P to P v
OP 4
and
let
IF" denote the rotation
transforming
to
P v
The
rotation
H\
being the product of the
orthogonal line reflection with
as axis followed
OP
as axis, carries the point pair to the point pair OP^. OP^
by that with Hence*
In like manner
it
follows that
.&"
Now
all
rotations are direct (Theorem 26).
Hence S(PP,PJ
Combining these relations with {^J?^},we have the order relation {P^PJKPP^, and in general, hy a like argument,
S(PjPfy
S(P^ffl.
Hence we have {PP mP m ,P^, whenever

*> 2*
fvv\
<
2
"
<
fvyt
<
1, as
can easily
be seen on reducing the two fractions to a common denominator. 2 n Since H = 1, it follows that whenever m/2 is expressible in the
form 2 k
(12)
+ a,
k being an integer,
IF* +a
Let
TT
= na
and
=P P a zk+a
DEFINITION.
be the constant denned in
67, (11).
The
number a
to
lie
rr,
where a
= m/2 n
is
called the measure of
any angle
congruent to
equal
An angle whose measure is #TT is also said a 2 a right angles. The measure of an angle is indeterminate according to this defi.
AP OP
to
nition.
In
fact,
according to (12), whenever the measure of an angle
is /3, it is
/3, where Jc is any positive or negative integer. This indetermination can be removed by requiring that the measure
also 2 &TT -f
/3
chosen for aiiy angle shall always
satisfy a condition of
the form
^ /3 <
7T,
Or~7T</3^7T.
Since the rays
OP m
2"
do not include
all
rays with
as center, the
definition just given does not determine the measures of all angles.
tinuity considerations, the details of

(steps
:
The required extension may be made by means which we shall
of
elementary con-
omit.
The
essential
order be any point in required are (1) to prove that if the^ there exists a positive integral value of n such that {P^PP.P^ {P^PPP^, * * 2^
(2)
,
hence to prove that if P be any point on the circle not of the form P a ] and [JP], such m the points of the form P m fall into two classes, [P
that
{PP a PPp}, and there

;
is
no
of the circle
(3)
having required that
point, except P, on every segment P^PP^ 2 * +a! a 2, to define II
< < <
is
an integer,
positive, negative, or zero,
and x
is
the
number
sucn tnat a
ing
<.
x <.
P to P
n
(4) to
p ior an a s ana p s; as uue ruuai/iuu. auuuu t/ show that if x is a rational number m/n, (IF)" =
of
;
"
;
(5) to define
measure
a = m/2 removed (6) angles differs from the sum.

defined as below.
angle as above, but with the restriction that to prove that the measure of the sum of two
of the measures
by 2
for,
the
sum
being
DEFINITION. If a, 6, c are any three rays having a common origin, but not necessarily distinct, any angle 4.& l c l congruent to 4-ac is said to be the sum of any two angles 4. aj) z and 4. & 3 c 3 such that 4 a 2 & 2 is
& c a,b and congruent to 8 g a 2 & 2 -f- 4 & 2 e 2 denoted by
.
4 4
is
congruent to
4 &c.
The sum
4a
1 1
is
to
For some purposes it is desirable to have a conception of angle according which any two numbers are the measures of distinct angles. This may be
:
obtained as follows
ray associated with an integer, positive, negative, or zero, is called a numbered ray. An ordered pair of numbered rays having the same hk in the earlier the measure of an angle origin is called a numbered angle. If sense is a, where o ^ a < 2 TT, the measure of a numbered angle in which h is
DEFINITION.
associated with m,
and k with
n, is
2(n
Defining the
the
m)7r
-j-
a.
sum
of
two numbered angles in an obvious way, it is clear that two numbered angles has a measure which is the sum of their
sum
of
measures.
The trigonometric functions can now be defined, following the elementary textbooks, as the ratios of certain distances multiplied 1 according to appropriate conventions. This we shall take for by
granted in the future as having been carried out. 70. The complex plane. Instead of the assumption in
64,
we
could assume that the Euclidean plane is obtained by leaving out one
line
from the ccTaplex projective plane (A, E, J, or A, E, H, C, E, I). All the results of Chap. Ill and of the present chapter up to 63 are applicable to this case. The rest of the theory, however, is essentially from that of the real plane, because the absolute involution two double points and because a line does not satisfy
different
necessarily has
the one-dimensional order relations.

principal rOle
a large class
Thus the minimal
lines play
and must be regarded as exceptional in the statement of of theorems; and another large class of theorems of
order relations)
disappears
elementary geometry (those involving

entirely.
J.OO
For the
of reals,
so, of
present, therefore,
but shall
(
make
S',
shall confine attention to the geometry use, whenever we find it convenient to do
we
the fact
6)
in a
in a
complex space,
unique
line
is
that a real space S may be regarded as immersed in such a way that every line I of S is contained
I' of S'. As a direct consequence it follows that any a subset of the points of a unique conic of S'. For 2 any five points of C , regarded as points of S', determine a unique conic of S' which, by construction ( 41, Vol. I), contains all points
conic
of
of C and is uniquely determined by any five of its points. Similar reasoning will show that any plane TT of S is contained in a unique plane IT' of S' and like remarks may be made with regard to any
;
one-, two-, or three-dimensional form.
A like situation arises with respect to transformations. A projective

transformation II of a form in S
is fully
fundamental theorem * of elements of S. set

S',
of projective geometry,
determined, according to the by its effect on a finite

is
Since the fundamental theorem

II'
also valid in
there
effect
unique projective transformation on this set of elements as II.

is
which has the same
Specializing these remarks

TT
somewhat we have
Euclidean plane
a subset of the points of a certain Euclidean plane TT' of 'S'. The line at infinity 1 9 associated with TT is a subset of the line at
of
is
infinity IL associated
with
I'
TT'.
l' n
The absolute involution

in
on
deter-
mines an involution
the pairs of I are paired. The involution I' has two imaginary double points, the circular points TT is a ( 56), which shall be denoted by 7X and J2 Since a circle in
all
.
on
which
conic having I as an involution of conjugate points, every circle in TT is a subset of the points on a conic in TT' which passes through II and J2
The problem
of
the intersection of a line and a
circle, or
indeed
of
a line and any ellipse, can now be discussed completely. In the proot z was seen of Theorem 25 the intersection of a line I and an ellipse
to
depend on finding the double points of a certain projectivity [LJ [LJ on /. Any three points L{, L[', L[", and their correspondents
L' L' a projectivity on the complex line I' containing 2> L", 2 ", determine this proI, and, by the fundamental theorem of projective geometry,
jectivity is identical
with [JDJ-^-fXJ so far
as real points are concerned.
The double points
of this projectivity are
common
to the
complex
line containing
are ieal
I and the complex conic containing JE*. These points the hypothesis of Theorem 25 is satisfied; they are real and coincident if I is tangent to E* otherwise they are imaginary.
if
A similar
lem
definitions
discussion will be
made
in the next section of the prob-
of the intersection of
two circles, but first let us make certain and conventions which will simplify our terminology.
6, any point of S' is said to be According to the definitions in complex, and a complex point is real or imaginary according as it In the case of lines, however, we have is contained in S or not.
three things to distinguish
a line of the space
S, a line of S'
which
contains a line of S as a subset,
and a
line of S'
which contains
In current usage a line of the last sort is called imaginary, a line of either of the first two sorts is called real, and a line of either of the last two sorts is called complex. The current
terminology therefore permits a confusion between a real line as a locus in S and a real line as a particular kind of a complex line.
this
cases, however, no misunderstanding need be caused by ambiguity of language, and we shall in future usually employ the same notation for the real line I of S and the line V of S' which
no such subset.
In most
contains
all
I.
The same remarks apply
to conic sections and, indeed, to
one-dimensional forms.
DEFINITION.
of S' is said to
Any
element
(point, line, or plane) or set of
elements
of
be complex.
Any
element or set of elements
is
said to be real.
respectively, of
A line
S
is
or plane of S' which contains a line or plane, said to be a real line or real plane of S'. one-
dimensional form of
of S'
S',
a subset of whose elements are real elements

S, is
and contain
all
the elements of a one-dimensional form of
called a real one-dimensional
form of
S'.
An
element or one-dimen-
sional
form
of S'
which
is
not a real element or real one-dimensional
form
of S' is said to
be imaginary.
transforms each real element of
S' into a real
DEFINITION.
said to be real
projective transformation of a real form of S' is
if it
element of
S'.
Strictly speaking, these definitions distinguish between the " senses of the word " real by phrases such as " real line of S'."
two
But
as
in practice
a
we
shall drop the " of S'."

all of
The one-dimensional forms
thus far defined are

f"f/Yfi
the
4~
first
or second degrees, but the defini-
rtot-i
i^r^~/i
r\
r\
'nr\ I*T* /AI 1
QeidrtT^4-"r*l
are really the double points of
I'
will be referred to in future as the
double points of the absolute involution I. In like manner, any line I 2 and circle C which have no real points in common will be said to
have in common the two points common to the complex 2 complex conic which contain I and C respectively.
line
and the
The
if
utility of these conventions will be
understood by the reader
he will write out in
full the discussion of pencils of circles in the
following section, putting in explicitly, in notation and language, the distinction between elements of S and S'.
It is also convenient in
many
(#2 ,
cases to extend the formulas for

to
distance, area, etc. given in

for example, in case (x v ,
(
67-69
y^ and
y^
are
imaginary elements. Thus, imaginary points such that
x i~
2)
+ (î- y
a a)
is
a positive real number,

(a^, y^)
V(^- a^ + ^-y/
to (xz , yz ). Extensions
will be referred to as the distance from
of terminology of this self-evident sort will be
made when
needed,
without further explanation.

71. Pencils of circles.
Consider two circles
C and
2 (72
in a real
,
Euclidean plane.
= ,
Cz let case Cj circle in a pair of real points

respectively.
let
Let their centers be denoted by Cl and Cz and in & denote the line C^. By Theorem 25, & meets each
which we
The two
pairs
may
&
denote the involution on
shall denote by J^Q : and P Z Q2 be entirely distinct, in which case transforming each pair into itself or
;
they
point in common, in which case the line through this point perpendicular to & is a common tangent of the two circles. The two pairs cannot coincide, because the circles would then coincide.
may have one
Thus four cases may be distinguished

(1)
(2)
(3) (4)
The circles have the same center. The circles have a common tangent and point The involution T is direct. The involution F is opposite.
of contact.
A circle is, by
lution.
60, a real conic which, according to the terminology
of the last section, contains the double points of the absolute invo*
Let us denote these points (the circular points) by

of
JT
and
I,
and apply the results
47, Vol.
I,
on pencils of
conies.
The
lines 01^
respectively.
and 0/2 are then tangent to both circles at 1^ and /2 47, Vol. I, it is evident that Hence, by reference to
the two circles belong to a pencil of circles of Type IV. In the second case C? and Cl have in common the points 1^ and /2 as well as a common tangent and point of contact. Hence they belong
to a pencil of
Type
II
which contains
all circles
touching* the given

the pairs
line at the given point.
In the third
case, since
the involution
is direct,
P^
and I$Q2 separate each other. Hence, by Theorem 29, the circles have two real points, A^ and A 2> in common. Hence they belong to a pencil of Type I consisting of all conies through A^ A 2 TI} and J2 This may
,
.
also be seen as follows
Since the involution F has no double points ( 21), it has a center 43) which we shall call 0. Let a be the line perpendicular to & at 0. Then by the argument used in the proof of Theorem 29, is between and Q r Hence a meets Of in two real points A 1 and A 2 (fig. 52). l The pencil of conies through A^ A 2 , Ilt J2 meets & in the pairs of an
(
involution
are
among which
and the
of
b.
J^Q l and
at
is
point
infinity
Hence C%
meets
(7
2 2
a conic of
the pencil, and hence a in

case,
and
In this
therefore,
the two circles belong to

a pencil of
Type
I.
In the fourth case the

involution
cannot have
a double point at infinity, because then the other
to
double point would have be the mid-point of

l
FIG. 56
%Q
and also
of
PQ
2
2,
and
thus C* and C\ would have a

the center
common
center.
Hence
in this case also
of the involution
its
is
an ordinary
point.
Let a denote
* A conic and one of

contact.
Two
tangent lines are said to touch each other at the point of conies touchiru? a line at the same point are said to touch each other.
the perpendicular to b at 0, and
let
and A^ be the points
in
which
a meets
C*.
These points are imaginary;
for otherwise, since they
are interchanged by the orthogonal line reflection with b as axis, would be between them, and hence, by Cor. 1, Theorem 25, would
be between 1\ and Q v contrary to the hypothesis that F is opposite. 2 Precisely as in the third case it follows that A l and A 2 are also on (72 Hence in this case also Cf and <722 belong to a pencil of Type I.
.
In
eacli case the facts established
make
it
clear that the
two
circles
could not both be members of more than one pencil of conies. any two circles fall under one of the four cases, we have
Since
THEOREM
30.
DEFINITION.
Any
circle contains the real points
of a
certain conic in the complex plane.
Two
conies determined by circles
are contained in a unique pencil of conies, which is of Type I, II, or IV. The set of circles which the conies of such a pencil have in common with
the real plane is called
a pencil
of circles.
If
the pencil
of Type IV,
the pencil
;
of
circles is the set
of
is
all circles
of conies is having a fixed
point as center
circles is the set
if the pencil of conies

circles
of Type II, the pencil of

line at
of all
tangent
to
a given
of
a given point ;
of
all
else
if the pencil of conies is of Type

circles
I, the pencil
circles is the set
the set
having a given pair of distinct real points in common, or of all circles with centers on a given line and meeting this
line
in the pairs of
an involution with two ordinary
double points.
DEFINITION.
The
line
a joining the centers
of
two nonconcentric
of centers of the two circles or of the pencil of circles which contains them. If the circles have a common tangent and point of contact, this tangent is called the radical axis of the two
circles is called the line circles or of
the pencil of circles
if
not, the line perpendicular to

circles of the pencil
at the center of the involution in
which the
meet
is
called the radical axis.
The double
points of this involution are
called the limiting points of the pencil of circles.
Any
circle of
the
pencil is said to
be about either one, or both, The discussion above has established
of the limiting points.
points
THEOREM 31. The radical axis of two common to them which are not on
imaginary points
circles
passes through all
the line at infinity.
The
the cirlimiting points of the pencil which they determine are real if cles meet and in if they meet in
only two real points.
imaginary
Jll
U VjU JJL>JCJ.a..l.>
JT -LUi.-Pl Oil
THEOREM
of
circles
circles
32.
The circular points,

I,
the limiting points of
a pencil
of Type
and
the two points not at infinity in which the
of the pencil intersect are the pairs of opposite vertices of a
complete quadrilateral.
The
sides
of the diagonal triangle of
this
quadrilateral are
pencil.
la
the radical axis,
and
the line
of centers of
the
Proof.
Let
and
z (fig.
common
to the circles of the pencil,
57*) be the points other than Jx and Ja and let B l and B 2 be the points
of intersection of
the pairs of lines IÂ.^ fÂ z and IÂ Z ,
1^ respectively.
Whether A, and 1
An
are
real or imaginary, the line
AÂ Z = a, which is the radical axis, is real. Hence its

point at infinity A n
is
real;
and hence the

the polar of
to
line
B^BZ
Aa
with regard
the pencil,
any
circle of
is real.
Since the line 6

is
=B B
:
the polar of A^,
it
conall
tains
the
centers
<
of
I., 1
conies through A, 1 A.,

J
L. Z
of
FIG. 57
Hence
and
Z>
is
the
line
centers of the pencil of circles
through
and
The points
JE?
being diagonal points of the complete quadrangle A 1A ZIÎZ are evidently the double points of the involution in which the
z
pencil of circles
meets
&,
and hence are the limiting points of
the pencil.
points A^,
Taking Theorems 31 and 32 together, we see that any pair of real A 2 determines a pair of imaginary points B v B2 such that;
either pair is the pair of limiting points of the pencil of circles through the other pair ; that, conversely, any pair of imaginary points B^ ZT which are common to two circles, determines two real points A
lt
B Az
which
,
are in the above relation to
B^ B
and that the three pairs
AÂ Z B^Bzt Z^ are pairs of opposite vertices of a complete quadrilateral. The relation between the two pencils of circles, the one
symmetrical. It can be described in purely real terms by means the following theorems and definition:
of
THEOREM 33. DEFINITION. If two circles have a point in common such that the tangents to the two circles at this point are orthogonal, the two circles have another siwh point in common. Two circles so
related are said
to le
orthogonal
to
each other.
orthogonal line reflection whose axis is the line of centers transforms each circle into itself and transforms the given point of intersection into another point of intersection. Since orthogonal
Proof.
lines are transformed to orthogonal lines, the tangents at the second
An
point are also orthogonal.
THEOREM
circle
34.
If a
line
through the center of a
circle
Cz
meets the
in a pair of points
PQ
1
a pair of points P^Q^ the

harmonically.
monically,
Proof.
let t
and meets any orthogonal circle K z in each other pairs P^Q^ and P Z Q Z separate
1
Conversely, if J^Q 1
circle
and
P Q%
Z
any
Let
through P^ and
Q2
is
separate each other har2 orthogonal to C .

circles,
T be
one of the points common to the two

circle
and
be the tangent to the tangent to t at T meets

in the pairs the line P^P Z of an involution F, and
TP Q
Z
at T.
The pencil
of circles
hence the
of the
if
first
statement
theorem will follow

can prove that
^*
we
and Q 1 are the double

points of this involution.
The
to
t
line
at
perpendicular and the line
FIG. 58
Z at P^ perpendicular to P^P are tangents to the circle
and MT are such that the pairs MP l meet in a point as center is tangent with congruent. Hence the circle through T to t at T and to JJJ at J> Hence ^ is a double point of T. A similar
(Ex. 4,
TP^
at
T and P
l
respectively,
and hence
60)
argument shows that Ql is also a double point. To prove the converse proposition we observe that there
one circle through
is
only
P and T and orthogonal to

2
(7
One such
circle,
by
the argument above, passes through, the point

ically separated
from
2
.
P
2
by
JJ
and
Q2 which is harmonQ r Hence the circle P Q T is

,
orthogonal to
As
a corollary
1.
we have
The
set
COROLLARY
is the
of all
circles
orthogonal
to
a pencil of Type
pencil of circles through the limiting points of the first pencil.
Another form in which this result
COROLLARY
the
to
2.
Let
C2
be
circle,
may be stated A any point not

t
is
the following
its center,
and
Az
point on the line joining A to the center of C which is conjugate 2 Then all circles through A I and with regard to the conic C l
2
.
orthogonal
to
C 2 meet
in
DEFINITION.
circle if
Two
if
and only
with
its
points are said to be inverse with respect to a they are conjugate with regard to the circle and
collinear
its center.
The transformation by which

is
corresponds to
~by
inverse
every point called an inversion or a transformation
reciprocal radii.
Thus the center
of the circle is inverse to every real point at infinity.
We
shall return to the
study of inversions in a
later chapter.
EXERCISES
1.
In case the limiting points of a pencil of
circles are real,
the radical axis
perpendicular bisector. is any point of the plane of a circle, and a variable line through O 2. If meets the circle in two points X, Y, the product OX OYis constant, and equal
is their
to
(07
3.
in case there is a line

called the
OT
OX OF is
power
of
with respect
tangent to the circle at T. to the circle.
The product
The power
of
respect to all circles of the pencil is a constant, for all points of the radical axis.
4. If
is
any point of the radical axis of a pencil of circles with and this constant is the same
the center of a circle,
any point
it,
of the circle,
and A^ and
any two points inverse with respect to
OA
5.
OA Z = (OC)
it.
2
.
Through two points not inverse

circle
relative to a given circle, there
is
one
and but one

6.
orthogonal to
a center of similitude of two circles is meant the center of a dilation ( 47) or translation which transforms one of the circles into the other. If the circles are concentric, they have one center of similitude; if are not con-
By
centric,
they have two.
The
they centers of similitude harmonically separate the
sircles is called
the
interior,
similitude. similitude.
7.
The common tangents

whose centers similitude which are the
circles
and the other of two
is
called the exterior, center of
circles
meet in the centers
of
Three
centers of
are not collinear determine by pairs six vertices of a complete quadrilateral having
its
the centers of the circles as vertices of

case of n circles.
8.
diagonal triangle.
Generalize to the
If a circle
meets two circles C% and C| in four points
at
which the pairs
tangents are congruent or symmetric, the four points are collinear by pairs with the centers of similitude of Cj2 and C|. Prove the converse proposition.
of
/2. Measure of line pairs. The circular points / 72 figure in a a very important formula for the measure of a pair of lines.* With the and two lines i i which of these two points, pass through exception lt z them, all the points and lines to which we shall refer in this section
:
are real.
The
center
and the point
at infinity of the axis of an orthogonal line

.
reflection are
harmonically conjugate with regard to It and 72
Hence
any orthogonal line
reflection, regarded as a transformation of the

,
complex, space, interchanges

Jj
7X and J2 and any displacement leaves and 72 separately invariant. Moreover, there exists a displacement transforming any (real) point of Zto any other (real) point of L. Hence a necessary and sufficient condition that a pair of points P, P' of l n be
/
transformable by a displacement to a pair Q, Q' of

(13)
is
Now any pair of lines meeting L in P and P' can be transformed P' by a translation into any other pair of lines meeting it in P and and any pair of lines meeting L in Q and Q' can be transformed by a translation into any other pair of lines meeting it in Q and Q'.
3
Hence the necessary and sufficient condition that a pair of lines of lines meeting it meeting L in P and P' be congruent to a pair
in
Q and
Q' is (13).
This suggests as a possible definition of the measure of a pair of

nonparallel lines
l^
are the lines joining the point of intersection of
It
where
and
1 2
and
1
i
a
to 7
and J2 respectively.
is rlne
would
satisfy the
requirement of
9,
*This formula
to
A. Oavlev. Cf Encvclopadie der Math. Wiss. Ill AB

.
JliU VJJj.UJJCj.tt.-LX
pairs,
being unaltered by displacements. however, we have

Dist (AS)
In the case
of
measure of point
+ Dist (BC) = Dist (A C)

by the
cross ratio
whenever {ABC}, and

given above.
this condition is not satisfied
We
have, in fact,
whenever
Z^
2,
are concurrent. This is easily verified

(
by
substituting
in the formula for cross ratio
56, Vol. I).
From
(15)
(14) it is obvious that if
we
define
2
m(^)=clogB(y2 ,v ),
the measure of line pairs will satisfy the condition
^w+^w^fty
whenever
l
l^,
z,
are concurrent.
Since the logarithm
is
a multiple-
valued function,
we must
specify
which value
is
chosen; and we
must
also determine the constant c conveniently.

of the
Making use
Z ro
same coordinate system as in
62,
any point on
may be denoted by (0, a, /3). In case a/ ft is real, (a/ftf > 0, and hence a and /3 may be multiplied by a factor of proportionality so that
(16)
a2 +)8 2 =
of this section
is
l.
Throughout the rest

to this condition.
we
shall suppose
a and
/3
subjected
This
equivalent to supposing that

/3
a
where
= cos (6 + 2 mr),
TT,
= sin (6 4- 2 nm),
=6S2
and n
of
is
an
integer, positive, negative, or zero.
The double points
the absolute involution satisfy the condition
and so may be written
^=(0,
where
i
1,
Z
t
<>)
and
1
2
Ia =(0,
1,
-i),
=V
1.
Now
if
and
meet
in (0, a lf J3J and (0, aa ,
respectively, it follows that
(58,
Vol. I)
to, a.
-4- fl.
S.\
-4-
i (a.
8.
a.B.\
The numbers a = <x^+ a 2 + /3 2 = 1. In fact, if a^

/3
/^ and
cos 6
l
/3
and
= a/ -a /S = cos
2
2
satisfy the condition
= sin 6,
where
=#
2,
then a
= cos
and
QZ
+ 2 mr.
Hence
Here
In
again,
=a
/3
and
@=2aft
a2 +
2
satisfy the condition
= l.
fact,
= cos 2 0.
Thus
(17)
B(^ ,v = + ^
2
2)
Hence
(18)
where 2
is real
and may be chosen

c in (15) as
so that
S2 <
TT.
Hence,
choosing the constant

(19)
2i
we have
2,
m (^ = =~ log 9 (y v = 0,
2) a)
where
that the
may
sum
be chosen so that
=6<
IT.
The formula
(19) is interesting in connection with the theorem of the angles of a triangle is equal to two right angles.
This proposition can easily be established without the consideration of imaginaries, on the basis of the definitions in the last section. From
our present point of view, however, it appears as follows: Let the three sides of a triangle be a, &, c, and let them meet the line at
infinity in A*,
Bx C, respectively. It is easily verifiable that R (A.B., J/ ) R (J5.C., IJ) 9 (CU, If) - 1,

,
-
from which
it
follows
by
(19) that
IT.
m (ab) + m (be) -f m (ca) =

Here we have a theorem on the
of a triangle.
line pairs rather
Indeed, (19)
is necessarily a
than on the angles formula for the measure

of
of a pair of lines
and not
of
an angle, because
the fact that two
opposite rays determine the
same point
also
at infinity.
The number m(ab) may

between
be defined as the smallest value

of
and
2-Tr,
inclusive, of the measures of the four angles

Z.
4^^ which
Following the
may be formed by a ray a^ of a, and a ray \ common usage, we shall say that two pairs
angles, etc.
of lines
which are conerruent make eaual
EXERCISES
1.
Jf
A and B are any two
points, the locus of a point
P such that the rays
PA
and
PB
make a constant
angle
is
circle.
2. If in two projective flat pencils three lines of one make equal angles with the corresponding three lines of the other, the angle between any two lines of the one is the same as the angle between the corresponding lines of the other. 3. If OA, OB, OC, OD are four lines of a flat pencil,
B (OA,OB;
V
= OC,OD) '
4AOC +
sin
set,
sinAAOD
4 BOC 4 BOD
In case the four lines form a harmonic
2cot4.AOB = cot4.40C + cot44OD.

4.
If
Av A2 As A4
,
,
are four points of a circle,
A A S A 2 A * ~ A A Z A A i + A A i' A 2 A 3>
'
where
Dist (A { Aj) according as S Aj represents Dist (A { Aj~) or S (OAÂ^) or not, being an arbitrary point of the circle and being a sense-class on the circle.
{
(OA Aj) = S (OA AJ)

{ {
5.
If a,
1),
c are the sides of a triangle

be, ca,
through the vertices

tangents of a conic
siix(a 1 b)
and a^, b^, c x c2 are pairs of lines ab respectively, the six lines a v a z , b v bz c v c 2 are
,
if
and only
if
sin.
(a 2 &)
(
2 c)
sin (j^c)
sin (62 c)
sin (!<?)
6.
sin
sin (â)
sin (bz a)
sin (c x a) sin (Cji)
sin (c 2 a)
_-
sin (c2 i)
The
points of a ray having (x, y) as origin

/
may
be represented in
the form
(z
Xa, y
0.
+
,
o\
is
A/3),
where a and
ft
are fixed
and A >
There
a one-to-one reciprocal corre-
spondence between the rays having (x, y) as origin and the ordered pairs of values of a and ft which satisfy the condition
a2
ft*
= 1.
A
is
When
7.
a and
ft
satisfy this condition, the numerical value of
the distance
between
(x, y)
Two
and (x + Aa, y + A/3). angles formed by the pairs of rays

*o
(a?
*>
A5,
% % + AJ8)
if
an <
*o (x
Vo
and
if
Ao', y^
A/?')
respectively are congruent
and only
oa'
8.
ftft'
= an" + ft
,
ft'.
for
homogeneous coordinates employed above, the formula the distance between (z xv a?2 and (y y y ) v z ) may be written
,
KeJative to the
2/2
73]
GENEBALIZATION BY PEOJECTION
Generalization
167
66
73.
by
projection.
The
relation established in
of
between Euclidean and projective geometry furnishes a source theorems in each. A theorem which has been proved for
new
projective
geometry can be specialized into a theorem of Euclidean geometry, or a theorem of Euclidean geometry may be generalized so as to furnish
a theorem
of projective geometry.
The two
processes, of generalization
in a
and
of specialization,
may
often
be combined
the principle of duality or with other general methods of projective geometry. Thus a theorem proved
happy way with
for Euclidean geometry can be generalized into a theorem of projective geometry and the dual of the general theorem specialized into a new theorem of Euclidean geometry. As an example, let us take the
theorem
A.
of Euclid: the vertices of a triangle to the opposite
The perpendiculars from
sides meet in
a point (the orthocenter).
The sides of the triangle meet the line at infinity in three and the three perpendiculars are lines from the vertices
Euclidean theorem
projective theorem:
B.
is
points,
to
the
conjugates of these three points in the absolute involution.

therefore
The
special
case
of
the
following
The
to
lines
respect
an arbitrary
joining the vertices of a triangle to the conjugates, with elliptic involution on a line I, of the points in
which
the opposite sides
meet
I,
are concurrent.
This
is
and the three
a portion of Theorem 27, Chap. IV, Vol. I, the orthocenter vertices of the triangle being the vertices of a complete
quadrangle. But though the Euclidean theorem is a special case, yet the general theorem for elliptic involutions in real geometry may easily
be proved by means of it. For, given any elliptic involution whatever and any triangle, the involution can be projected into the absolute
transformation which carried the involution into the absolute involution would carry the triangle into one whose sides are not all real.
Now
B'.
consider the plane dual of the projective theorem, B.

sides
The points of intersection of the
of a triangle with the
conjugates in
an arbitrary
involution at a point L, of the lines joining
the vertices to L, are collinear.

If the involution at
is
taken as the orthogonal involution
we
have the Euclidean theorem:

A'.
TJie three sides
of a triangle are met in three collinear points
ly the perpendiculars
to the
from a fixed point
to the lines
joining this point
opposite vertices.
The second of the two processes which we are here emphasizing, the discovery of Euclidean theorems by specializing projective ones, liantly illustrated in many of the textbooks on projective geometry.
mention the following
L. Cremona,
:
namely
is bril-
We may
Elements of Projective Geometry, Oxford, 1894. T. Beye, Geometric der Lage, Leipzig, 1907-1910. K. Sturm, Die Lehre von den Geometrischen Verwandtschaften, Leipzig, 1909.
K. Boger, Geometric der Lage, Leipzig, 1900. H. Grassman, Projective Geometric der Ebene, Leipzig, 1909. J. J. Milne, Cross-Ratio Geometry, Cambridge, 1911.
J. L. S.
The reader
Hatton, Principles of Projective Geometry, Cambridge, 1913. will find material for the illustration of the second process,
namely the discovery of projective theorems by generalizing metric ones, in Euclid's Elements, and even more in such books as the following J. Casey, A Sequel to the First Six Books of the Elements of Euclid,
:
Dublin, 1888.
C. Taylor,
J.
W. Kussell, Elementary
class of
Ancient and Modern Geometry of Conies, Cambridge, 1881. Treatise on Pure Geometry, Oxford, 1905.
theorems which are here in question will be dealt with to some extent in the following chapter, and the methods available will be extended in Chap. VI by the study of inversions. But on account of the magnitude of
subject many important theorems will be found relegated to the exercises and many others omitted entirely. In nearly every such case, however, a good treatment can be found in one or another of the books on projective geometry
The
the
referred to above.
The
geometries to which they belong

sary for their proof
current textbooks do not often classify theorems on the basis of the ( 34) and the assumptions which, are neces-
(17). Some
progress has been
made on such a
classifi-
73]
(mNEKALIZATION BY PBOJECTION
criticism
169
Another
on current books
in a rather shy and awkward previous to a logical treatment of the subject based on definite assumptions, the geometry of reals was regarded as having, somehow, a higher degree of
validity than the complex geometry.
that they employ imaginary points manner. This is doubtless due to the fact that,
is
The reader
will often find it easy to
abbreviate the proofs of theorems in the literature by a free use of imaginary elements (cf. 78).
EXERCISES
1.
Generalize project! vely the following theorems:
(a)
(&)
The medians of a triangle meet in a point. The perpendiculars at the mid-points of the
sides of a triangle
meet
in a point.
(c)
The
diagonals of a parallelogram bisect each, other.
lines joining the vertices A,B,C, respectively, of a triangle to the orthocenter, 0, meet the opposite sides. The and C*x contains the mid-points of the pairs AB, BC, CA circle through A v l
2.
Let A v B v Cj be the points in which the
and of the
circle of
pairs
OA, OB, OC. This

Cf.
circle is called
the nine point or Feuerbach
the triangle.
Ex.
7,
41.
or
of
3. hyperbola whose asymptotes are orthogonal is said to be equilateral rectangular. Every hyperbola passing through four points of intersection two equilateral hyperbolas is an equilateral hyperbola. 4.
All equilateral hyperbolas circumscribed to a triangle pass through
its
orthocenter.
5.
lie
The
centers of the equilateral hyperbolas circumscribed to a triangle

circle.
on the nine-point
CHAPTER V*
ORDINAL AND METRIC PROPERTIES OF CONICS
74. One-dimensional projectivities.
The general discussion

I,
of one-
dimensional
projectivifcies
in Chap. VIII, Vol.
has a great
many
points of contact with the ordinal
and metric theorems
of the last
three chapters.
transforms into
itself
For example, a rotation leaving a point invariant 2 as a center. The transforany circle C with
is
mation effected on the circle by the rotation

projectivity having the point
axis.
is
a one-dimensional
line at infinity as
as center
and the
The denning property

if
of the axis of the projectivity in this case of the circle be rotated into a pair
that
if
a pair of points
AB
A'B'
r be congruent to 2.A'OB'}, then the line AB to the line A'B, which is a well-known Euclidean theorem.
(i.e.
2.AOB
is parallel
The proposition that any rotation

of
is
tions corresponds to the proposition that
a product of two line reflecany projectivity is a product
two involutions. The point

with
2
reflection
with
as center is
effects
ative
all
the other rotations about

I)
and hence
all
commuton (72 an
involution
which (79, Vol.
belongs to
the projectivities effected
on C by the rotations of this group. This involution is harmonic z 78, Vol. I) to the involution effected on C by any orthogonal line (
reflection
whose
axis contains 0,
pencil.
and hence
all
all
the involutions of the

79, Vol.
I,
latter sort
form a
Thus
the theorems of
can
be specialized so as to yield theorems about the group of rotations
with
as center.
other applications of the theorems in Chap. VI, and Euclidean geometry (a few of them are indicated in the exercises below), but the main application which we are to
There are
I,
many
Vol.
to affine
consider at present
recall
is
to the theory of order relations.
Let us
first
some
of the ordinal
theorems which have already been estab-
oiueieu space
AS
wt/peruunc, paraoouc, or
eiwpiic
accoroing as
it
has two,
one, or
no double
points.
With
regard to involutions,
we have already
established the followsense,
ing propositions (21): If an involution preserves

separates every other pair.
each pair
If an involution
alters sense,
no pair
separates any other pair.

is elliptic
;
An
involution which does not alter sense
that
is
to say, the pairs of a hyperbolic involution do not
separate each
other.
The double points of a hyperbolic involution
separate every pair of the involution.
DEFINITION.
If
A, B, C,
are four distinct points of a conic, the
point
of intersection of the lines
AB
and
CD
is
called an interior
point in case the pairs
AB
and
CD
separate each other* and an
exterior point in case these pairs do not separate each other. The set of all interior points is called the interior or inside of the conic, and the set of all exterior points is called the exterior or outside of
the conic.
CD are conjugate in the involution with as these two pairs separate each other, this involution is such that any two of its pairs separate each preserves sense and which meet the conic meet it other. Hence any two lines through
The
pairs
AB
and
center.
Hence,
if
in pairs of points which separate each other. That is to say, the definition of an interior point is independent of the particular choice of
like argument applies in case is exterior. the points A, B, C, D. as center has double points, the lines In case the involution with
joining
to these points are tangent to the conic.
Hence the next

These results
to the last of the propositions about involutions stated above implies
that there are no tangents through an interior point. may be stated as follows
:
THEOREM
fliutually
exterior.
the
1.
The points coplanar with a conic fall into

classes
:
three
its
exclusive
the
is
conic
the
itself,
its
interior
and
Each
interior point
center
of an involution on
conic
which preserves
sense.
which alters
sense, and each exterior point of one All points of a tangent, except the point of
contact, are exterior points
of the conic.
# Cf .
20, particularly the footnote.
Now
rangle
let
be any interior point.
If 0' is
any point conjugate

fig.
to
with, regard to the conic, there exists (cf.
ABCD
whose
59) a complete quadvertices are points on the conic such that AB
and
CD
meet in
and
AD
and
CB
meet in
0'.
But by Theorem
7,
FIG. 69
does not separate BC. and Chap. II, if AB separates CD, then hence O'is an exterior point. Hence the polar line of any interior
point consists entirely of exterior points.
AD
Hence
interior point are exterior.
THEOREM
2.
All points conjugate
to
an
Suppose, further, that the tangent to the conic at B meets the line 00' in a point P and the line BD meets 00' in a point P' (fig. 59). Then P and P' are conjugate points with regard to the conic. Moreover,
ABCDjÔPO'P'.
and C do not separate B and D, does not separate the pair PP'. That is,
Since
it
follows that the pair 00'
THEOREM 3. On a line containing an interior point of a conic the pairs of conjugate points with regard to the conic do not separate
one another.
By
at
elementary propositions about poles and polar there follow once:

line
is
COROLLARY 1. The pole of a an exterior point.

COROLLARY
2.
which contains an interior point

in-
The polar of an exterior point contains some
In
of
78, Vol.
I, it
was established that any

is
projectivity
is
a product
two
involutions one of which
hyperbolic.
Since a hyperbolic
involution
it is
is opposite, it follows that if the given projectivity is direct, a product of two opposite involutions and if the given projectivity is opposite, it is a product of a direct and an opposite involution.
;
But
in the second case the direct involution

of
is,
made, a product
two opposite
involutions.
a,
by the argument Hence
just
THEOREM
involutions,
involutions.
4.
direct projectivity is
product of two opposite
and an
opposite projectivity is
is
a product of
three opposite
An
opposite projectivity
also expressible as a product
of a direct and an opposite In the case
involution.
of projectivities
is
on a conic, the axis
of the
product of
have, as
two involutions
the line joining their centers.
Hence we
consequences of this theorem,
COROLLAKY
COROLLARY
1.
Any
line in the
plane of a conic contains points
exterior to the conic.

2.
projectivity whose center is
an
interior point,
and
whose axis therefore consists
entirely of exterior points, is direct.
In the fourth exercise, below, we need the following

DEFINITION. The
passing through
at this point.
its
definition
line perpendicular to a tangent to a conic
and
point of contact
is
called the normal to the conic
EXERCISES
1.
Zoo
What
transformations of the Euclidean group effect projectivities on
to
which the absolute involution belongs?
How
from the remaining similarity transformations by their
are these distinguished relation to the cir-
cular points? What transformations of the Euclidean group are harmonic OB. L> to the absolute involution?
2.
Show that
the measure of a line pair as denned in
72
is
the logarithm
of the characteristic cross ratio of a certain projectivity on Z. Obtain an analogous formula for the measure of an angle in terms of the characteristic
cross ratio of a projectivity
3.
on a
circle.
noninvolutoric planar collineation which leaves invariant a conic and a line transforms the points of the line by a projectivity to which belongs
Any
4. If
P is
any
is
that
&RPQ,
a variable point pair such fixed point of a conic and meet in a fixed point on the a right angle, the lines
RQ
RQ
normal at P.
5. The lines joining homologous points in a noninvolutoric projectivity on a conic are the tangents of a second conic.
6. If
is
any fixed point
of a conic
and
RQ
a variable pair of points
A 1 and B v followed by another involution, and if T- 1 (J. 1 ) =A Â l and r(^4 1 ) = A z then A l and J3 t are harmonically conjugate with regard to A Q and A z whenever A jÂ z and B t = A whenever A = A z 8. If A l and B l are a pair of an involution I which is left invariant by a 1 = A Â l and T(A : ) = A 2 & A then A and projectivity T, and if T~ (A ) A z are harmonically conjugate with regard to A l and B r be A' 9. Let A and any pair of an involution I. If A & A', any projectivity II which transforms I into itself and leaves A invariant is either the involution, with A and A' as double points, or the identity.
points,
,
.
',
such that 4-RPQ, has constant measure, the lines RQ, are the tangents to a second conic. 7. If a projectivity P on a line is a product of an involution having double
10. Generalize 80, Vol. I, so as to apply to the group of translations and the equiaffine group, using the fact that the transformations in each of these groups are products of pairs of involutoric projectivities.
75. Interior
and exterior
of
conic.
THEOREM
5.
Any
two points of a conic are the ends of two linear
segments one consisting entirely of interior points and the other entirely of exterior points.
Proof. Let the given points be denoted by A and B, let G and be any two other points of the conic which separate A and S, and let cr and <r on the conic. By the represent the segments A CB and
ADB
definition of the order relations
on the
conic, the lines joining

of a
to
the points of
cr
meet the
line
AB in the points
segment
<?'
whose
ends are
points.
line
A and B, and these points satisfy the definition of interior In like manner the lines joining C to points of cr meet the
a segment
cr'
AB in
which
is
complementary to
is
cr'
and
consists
entirely of exterior points.
In a real plane the following theorem
we have just proved, but in order to have the result plane we give a proof which is entirely general.
a consequence of what for any ordered
THEOKEM
6.
Any
two interior points of a conic are the ends of a
INTERIOR OF A CONIC
175
roof (fig. 60). Let A and C be two interior points. Let A 1 be any it of the conic not on the line A C. The lines .^(7 and AÂ are not,
gent to the conic, since (Theorem 1) the involutions at A and C are h. elliptic. Let A Q and Bz respectively be the points, distinct from
in
which the
lines
AÂ and A^C meet
the conic whose ends are
and
J?
the conic. The two segments are projected by the lines 2
Dugh A^ into the two segments

s
of the line
A C which
have
and
their ends.
We
shall prove that the
segment &
of the line
Q
1
AC
Lch is the projection of the segment complementary to

s entirely of interior points.
A A BZ con-
Let
3
S be any point of
A
and
let
2
or.
The
A^B then meets the conic hi )int C which is separated from z

by
Let
BA
2
meet
it
conic in C^ let G^B meet
in
and
.t
A 2 C meet
1
*
it
in
Bv so
ABCAB,12 C, form a Pascal

1
cagon whose pairs of opposite es meet in A, B, C. Since A is

interior point,
.er
FIG. 60
we have
Since
the
l
{C A B.A J.
Z
,
B was chosen so that Cz and A
are separated
by
and A Q
we have [Bz CyA A$. From these there follows (Bz C^B^}, Hence insforming this by the involution at A we have { CyB^C^J. have {B^CyA^BjA^. Since the involution with center at C is Hence we have {Bz Cy A Q C^BÂ^. Hence iptic, we have {BZ BÂ Â 2}. and A^ separate A 2 and G and hence B is interior to the conic.
I}
THEOREM
1
7.
Any two
exterior points are ends of a segment consist-
entirely of exterior points.
If the line EÊZ f contact are exterior, points on it except the points of ice each of these points is the center of a hyperbolic involution the conic. In this case the theorem is obvious. If the line EÊ^
Proof.
Let the two exterior points be E^ and

all
tangent,
;ets
f.Vip
linp
the conic hi two points, the theorem reduces to Theorem 5. TV TT. rinps Tint, TYip.p.h t.hp, p.onic. and both the segments with
176
The theorems above
tions
:
CONIC SECTIONS
Any
[CHAP.V
are connected with the following algebraic considers involution, can be written in the form
v ' (1)
*-
= 5E+A ~
If we regard a, I, c as a set of homogeneous coordinates in a projective plane, then for every involution (1) there is one and only one point (a, b, c); and inversely for every point (a, b, c) there is a unique involution (1), provided
that the point does not satisfy the condition

(2)
a2
be
0.
By
18 the projectivities (1) for which

(3)
a2
+ +
6c
>
are opposite,
(4)
and those for which

a2
be
<
are direct.
the exterior
The equation (2) represents a conic section of which the points satisfying (3) are and those satisfying (4) are the interior. This may be proved as follows:
conic
is
The
given by the parametric representation
82, Vol. I)
c
is
x xz
:
1,
and any involution on the conic

parameter
x.
The
center of the involution
given by the transformation (1) of the is the point of intersection of the
The point (0, 0, 1) of the conic is of the parameter x and thus is transformed to the point given by the value a 2 ). The point ab, V*, b/a, namely, the point ( given by the value x =
lines containing pairs of the involution.
= GO and thus is transformed to the (0, 1, 0) of the conic is given by x 2 c2 The point of interpoint given by x = a/c, namely, the point (ae, a , ). 2 section of the lines joining (0, 0, 1) to ( a 2 ) and (0, 1, 0) to ab, b 2 2 a c is Hence is the center of the b, a, b, c) (ac, manifestly ( ) c).
(a,
involution (1), and therefore is interior to the conic
if
(4) is satisfied
involution direct, and exterior to the conic

tion opposite.
if (3) is satisfied
and the and the involu-
EXERCISES
Parabolic projectivities are direct. 2. Two of the three vertices of any self-polar triangle of a conic are
1.
exterior points.
3. 4.
5.
The The
center of a hyperbola is an exterior point. center of a circle is an interior point.

all
In a Euclidean plane all points interior to a circle and
points on it
ivhose diameter are
6. Tf a segment A l B l is contained in a segment A B the circle the ends Z 2 A and B l is composed of points interior to the circk ends of whose diameter are A a and B a A 7. ]n a Euclidean plane all points interior to an ellipse lie entirely on
,
.
side of
8.
any line consisting entirely of Any two pairs of conjugate diameters
exterior points. of an ellipse separate each other
9.
h respect
nterior
pairs of conjugate diameters of a hyperbola never separate each other. z If is the center of a conic the polar reciprocal of a conic C2 to K* will be an ellipse, parabola, or hyperbola according as O
to, on,
or exterior to
2
C2
10. Consider a conic

1
L
II.
The
]
[/
(1
C in a planar net of rationality satisfying Assumppoints of the net exterior to the conic fall into two classes two such that tangents to the conic can be drawn from any point
[Q E
no tangent can be drawn to the conic from any point F. On any line in ich one E is conjugate to an F with regard to C 2 every E is conjugate to F. On any line in which one E is conjugate to an E, every E is conjugate
[
,
an
I
or both do not
7? and every F to an F. The interior points fall into two classes [/] [J] .such that the pairs of conjugate lines on a point I either both meet meet C 2 whereas one member of any pair of conjugate lines
,
a point
./
meets
C 2 and
the other
member
,
does not meet
C2
11. Let the equation of a conic be/(z the coefficients of f(x , x v x 2 ) be
xv #2 )
and
let
the determinant
A=
x' 1}
00
U 01 "02
0,
Q point (x'
3)
is interior or exterior
according as
Q 'f(x'
x{, x' 2 ) is
ater or less
than
zero.
76. Double points of projectivities.
The preceding theorems hold
any ordered space. On specializing to a real space we have the iitional theorem that a projectivity which alters sense has two
able points
(
4).
In the case
of involutions this
result comis
led with the

posite
theorem that a hyperbolic involution
always
gives
8.
THEOREM
6
another,
another.
The pairs of an elliptic involution always separate and the pairs of a hyperbolic involution never separate
The
e
last half of this
theorem, combined with Theorem
3,
gives
condition for the intersection of a line with a conic, a condition
conic in the double points of this involution.
By
Cor. 2,
Theorem
3,
the polar of an exterior point
is
a line
through an interior point. The lines joining the exterior point to the points of intersection of its polar with the conic are tangents.
Hence
COROLLARY
to
1.
Through any
exterior point there pass two tangents
conic.
2.
COROLLARY
Two
involutions, one at least of which
is
elliptic,
have one and only one

Proof.
is
common
pair.
The center
of
an elliptic involution represented on a conic
an interior point. The line joining this point to the center of any other involution meets the conic in two points which are pairs of both involutions. Since any pair of an involution is collinear with
the center, the two points so constructed are the only pair
to the
common
:
two involutions.
this corollary
A special case of
COROLLARY
3.
may be
stated in the following form

there is one
In a given one-dimensional form
and
only one pair of elements which are conjugate with respect to a given elliptic, involution and harmonically separated by a given pair of
elements.
Since a hyperbolic involution is determined by its double points, it evident that any two hyperbolic involutions are equivalent under the group of all projectiviti.es of a one-dimensional form. The correis
sponding theorem for elliptic involutions is best seen by representing the involutions on a conic. The two centers / J are interior points, 2
x,
them meets the conic in two points C^, (72 which do not separate them (Theorem 5). Let O and be the double l 2 points (Theorem 8) of the involution in which Jx/2 and C^C% are pairs.
and the
line joining
An
evidently transform the one with

as center.
involution with either of the points O or O as center will z t JT as center into the one with J a
Hence
4.
COROLLARY
dimensional
Any
two elliptic involutions in the same real one-
form
are conjugate under the protective group of that
form.
EXERCISES
A1J involutions which are harmonic to (i.e. commutative with and distinct from) an elliptic involution are hyperbolic.
1.
2. If
two points A,
its
of a line separate each point
P(P-&A, Pj^B)
of
bhe line from
conjugate point in a given
elliptic involution,
and
are
conjugate
3.
in this involution.
homologous
hyperbolic projectivity is opposite or direct according as a pair of points does or does not separate the double points.
by
4. Elliptic projectivities are direct. 5. The center of an ellipse is an interior point. 6. The involution determined on the line at infinity of a Euclidean plane an ellipse is elliptic, by a hyperbola, hyperbolic.
7.
Any two
ellipses are conjugate
8.
An
involution in a
is
flat pencil is either

is
gate lines conjugate

9.
orthogonal or there
under the arnne group.* such that every pair of conjuone and only one orthogonal pair of
lines.
conic having two pairs of perpendicular conjugate diameters is
circle.
10. If A l and A z are the real limiting points of a pencil of circles, each and is on the opposite side of the radicircle of the pencil either contains : or contains and is on the opposite side of the radical cal axis from z z
axis
from A v Of two
circles of a pencil,
11.
both containing the same limiting point,

is
one
entirely interior to the other. 12. For any angle, 4. ABC, there
is
one and only one pair
I,
I'
of orthog-
onal lines through

line,
I,
B -which
separate the lines
BA and EC
harmonically.
One
interior to of the pair contains points is conABC, and gruent to 4-PBC. The line I is called the interior lisector, and the line I' the
&ABP
exterior bisector, of the angle
4 ABC.
13.
The asymptotes
of
an
equilateral hyperbola bisect
any pair
of conju-
gate diameters.
14.
The bisectors
of the angles of a triangle
ABC
meet in four
26.
points,
one
in each of the four regions determined by ABC according to
These four
and are the vertices of points are the centers of four circles inscribed in is the diagonal triangle. The mida complete qxiadrilateral of which
ABC
ABC
point of the pair BC is the mid-point of the points of contact of either pair of inscribed circles whose centers are collinear with A.
15.
line,
formed to V. There exist two points A, two points A', B' such that
be the vanishing points ( 43) of a projectivity on a Let V and the notation being so assigned that the point at infinity is transB which are transformed to
180
CONIC SECTIONS
The discussion in Chap. IX,
77, Ruler-and-compass constructions .
reduces any quadratic problem to the problem of finding the points of intersection of an arbitrary line with a fixed conic. According to Theorems 5 and 9 the necessary and sufficient condition that
Vol.
I,
a line coplanar with a conic meet it in two points is that the line pass through an interior point of the conic. Hence this condition will serve to determine the solvability of any problem of the second
degree in a real space. Thus the discussion of linear and quadratic constructions, under the projective meaning of these terms, may be
regarded as complete. When we adopt the Euclidean point of view, the fixed conic may be taken as a circle; and therefore every problem of the second
degree
is
section of
reduced to the problem of determining the points of interan arbitrary line with a fixed circle (cf. 86, Vol. I).
of elementary Euclidean
The constructions
geometry which are
known
or of
as ruler-and-compass constructions involve the determination
(whenever existent) of two arbitrary lines, an arbitrary line with an arbitrary circle, or of two arbitrary circles. The last of these problems has been shown in 65 to be
of the points of intersection
reducible to the
struction
first and second. Hence any ruler-and-compass conbe reduced to the problem of finding the intersection of an arbitrary line with a fixed circle.
may
On
account of the special character of the line at infinity, there
is
not a perfect correspondence between the linear constructions of projective geometry and the Euclidean constructions by means of a ruler.
The
operations involved in the linear
constructions of projective
geometry are
(a) to join
(5)
two points by a
(projective) line
to take the point of intersection of
any two
lines.
These are evidently equivalent to the folio wing Euclidean operations:

(1) to join
(2) to
(3')
two ordinary points by a line ; take the point of intersection of two nonparallel lines to draw a line through a given point parallel to a given
first
line.
The
of these operations
corresponds to the proposition that
may
be thought
is
of as carried out
with a straightedge or ruler whose
length
not limited.
(3')
The
operation
can be effected by means of

:
(1)
and
(2),
together
with the following operation

(3) to lind
on any ray through a point -4, a point

is
C such
that the
point pair
AC
congruent to a preassigned point pair
AB*
For let A be the given point and let BC be the given line. Let be in the order {ABO} such that BA is congruent a point on the line to BO. Let A be the point of the line OC in the order OCA such that
AB
CO
is
congruent to CA. Then

(1), (2),
AA
is
evidently parallel to
BC.
Thus
and
(3) serve as a basis for all linear operations in
the projective sense.

constructions.
They obviously
;
yield also a certain class of

suffice for all quadratic
for, as
quadratic constructions
but they do not The latter may be provided
by adjoining the operation
of taking the point of intersection
explained above, with a
fixed circle of an arbitrary line through an arbitrary interior point. For the proof that (3') is not a consequence of (1) and (2), and
that
(]), (2), (3)

is
do not provide
for all quadratic constructions, the
reader
referred to Hilbert, Grundlagen der Geometrie, Chap.
VII
(4th edition, 1913).
EXERCISES
AB is congruent to BC, show how to construct a parallel to the line AB through an arbitrary point P alone. and of the (2) operations (1) by means 2. Given two parallel lines, show how to find the mid-point of any pair of the lines by means of (1) and (2) alone. of either on points 3. Given a parallelogram and a point P and a line I in its plane. Through P draw a line parallel to I, making use of the ruler only.
1.
Given three collinear points A, B, C such that
* It is important to notice that the pairs and have the point A in comfor drawing a circle through a given point and (3) provides merely with a given other point as center. The drawing instrument to which this correso sponds is a pair of compasses which snaps together when lifted from the paper, to a point pair A'B' unless that it cannot be used to transfer a point pair A A'. This will be understood by anyone reading the second proposition in Euclid's Elements, which shows how to lay off a point pair congruent to a given of point pair on a given ray. The operation (3) may be replaced by the operation C is congruent to a fixed a point C such that the point pair finding on any ray measuras a point pair OP. The instrument for this operation may be thought of the reference ing rod of f xed length (say unit length) without subdivisions. (Of.
AB
AC
mon. Thus
AB
AB
4. Given a point pair construct the point pair
AC and its mid-point AD such that AC n

'
B, using the ruler
alone,
-AD~
5.
Given four collinear points A, A', B, B', construct the fixed point of
the parabolic projectivity carrying A to A' and B to B'. 6. Given a projectivity on a line, find a pair of corresponding points and A' such that a given point is the mid-point of the segment A A'.
7.
8.
Inscribe in a given triangle a rectangle of given area. Given four tangents of a parabola, construct a tangent parallel to a
line.
given
9.
find the point of intersection of the
Given three points of a hyperbola and a line parallel to each asymptote, hyperbola with a line parallel to one of
the asymptotes. 10. Construct by ruler and compass any
number
of tangents to a conic
given by
five of its
points
also
any number
of points of a conic given
by
five of its tangents.
11. Construct
any number
of points of a parabola through four given

of a parabola touching three given
points.
12. Construct
any number of points
lines
and passing through a given point.
Through a given point construct an orthogonal pair of lines conjugate with regard to a conic. (If the point is exterior to the conic, these lines are the bisectors of the angles formed by the tangents to the -conic
13.
from
this point.)
78. Conjugate
imaginary elements. It has been shown in
6 that
a real projective space S can be regarded as immersed in a complex projective space S' in such a way that every line of S is a subset of a unique line of S'. Certain additional definitions and conventions have been introduced in 70. But in both these places little use was
made
of the properties of imaginary-
elements beyond their existence
and the
fact that S' satisfies
prove some of
DEFINITION.
Assumptions A, E, P. We shall now the most elementary theorems about the relation beS'.
tween elements of S and
points, lines, or planes are said to be conjugate relative to a real one-dimensional form of the first or second
Two imaginary
if
degree if and only in the real form.
they are the double elements

2
of
an involution
to
As an example
The
conic
consider a real conic

line have in
and the
C and a line I exterior common the double points of an
it.
ellip-
78]
CONJUGATE IMAGIN ARIES

2
183
to
the involution on
<?
whose
axis
is
/.
Hence the points common
2 are conjugate imaginaries both with respect to C' and to I. Since any one-dimensional form of the first or second degree whose
C 2 and
elements are points
is a line or a point conic, and since the double points of any involution on a conic are the intersections of the axis of the involution with the conic, we have
THEOREM
real line.
10.
Any
two conjugate imaginary points
are on
By
duality
line.
we have
that any two conjugate imaginary planes are
on a real
line conic, cone of lines, or regulus.
conjugate imaginary lines are by definition on a real point, If they are on a real line conic, the plane dual of the argument above shows that they are on a real
point.
Two
By
dualizing in space
we
obtain the same result for conjugate
imaginary
lines of a cone of lines.
Hence we have
two conjugate imaginary coplanar lines are on a real point and any two conjugate imaginary concurrent lines are on a, real plane.
11.
THEOREM
Any
Conjugate imaginary lines on a regulus will be considered in a

later chapter.
THEOREM
12.
The
lines joining a real point

it
imaginary points not collinear with

Proof.
to two conjugate are conjugate imaginary lines.
The conjugate imaginary

on a real
line.
points are double points of an
elliptic involution
From any
point not on this line

of lines
this involution is projected into
an involution
whose double
lines are the projections of the given points.
THEOREM
point
13.
If AÂ Z and
B^^
nary points on
different lines, the lines
are two pairs of conjugate imagiand meet in a real A

1
A^B Z
and
are conjugate imaginary
lines.
Proof. By hypothesis the lines AÂ 2 and BJE> Z are real and hence they meet in a real point G. Let B be the conjugate of G in the elliptic
the two real points which are paired iii this involution and separate A and C harmonically. Since any two harmonic sets are protective,
CBPQ == CARS
The centers
of these
and
CBPQ = CASK.
two
real points
two
perspectivities are
C and
1
(7 , 2
and since each perspectivity transforms two pairs of the elliptic involution on the line A t A 2 into two pairs of the elliptic involution on the
line
points
B^BZ) it transforms A l and A z to Bl and J? 2 Hence one of the Cl and C2 is the intersection of the lines A B and A 2 B 2 and
.
the other that of the lines

these pairs of lines
is
AJBZ and A
B^.
By Theorem
12 each
of
a pair of conjugate imaginaries.
The complete quadrilateral whose pairs of opposite vertices are AÂy BJ32 and C^C^ is analogous to the quadrilateral considered in 71 whose vertices were IJ,2 and the limiting points of two orthog,
onal pencils of circles
(cf. fig.
57).
With
regard to the existence of
such quadrilaterals
we have
AÂ Z BJ3Z C^C^
, ,
THEOREM
of
a,
14. Let
lie
the
pairs of opposite
vertices
complete quadrilateral. If
A^
and
B^
are pairs of conjugate
C^ and z the complete quadrilateral is real.
imaginary points, then
are real
B
,
are conjugate imaginaries, then

the diagonal triangle is real.
and the diagonal triangle of If A and A are real and B and C and Cn are conjugate imaginaries
1 l
and
Proof.
In the
first
case
C and Cz
t
of the last
theorem and hence are
real.
are determined as in the proof The diagonal triangle has for
In the second
illy
,/?
case let a be the line through
A 2 which
is
harmon-
conjugate to A Z A^ with respect to the pair of lines A Z B 1 and Since the latter two lines are conjugate imaginaries and A A
real,
is
is real.
transforms
The harmonic horuology with A as center and a as 1 and B^ to C1 and Cf Hence C^ and C are conjugate l a
line
taginaries
and the
C^
is real.
Relatively to a real frame of reference a real involution is repreated by a bilinear equation with real coefficients ( 58, Vol. and
I),
double points appear as the roots

efficients.
Hence the
of a quadratic equation with real coordinates of a pair of conjugate imaginary
ints are expressible in the form
XQ x lt x2 x & y y v ya> y g are real Like remarks can be made th regard to the coordinates of a plane or a line, and Theorems 10-14 n easily be proved analytically on this basis. The following theorem
iere
,
,
pears to be easier to prove analytically than synthetically
THEOREM
%l
15.
complex
line
on a real plane contains at
least one
point.
Proof.
Let the equation of the line be
^+^+^,=
lis
0.
may
u
f
be expressed in the form

(u[
+ iu'l)
be
-f (u(
+ iu'J] x + (u[ + *'<)

l
x*
>
here
Q,
u'J, etc.
are real.
real, to
This equation
is
equivalent,
if
xa x v x3
,
Q required to
u' x u' 'x

iô
+ ufa + u'yK = 0, + u' 'x + u'Jx = 0,

2
1
1
equations which are satisfied by at least one real point.
EXERCISES
1.
A
A
conic section through three real

pair of conjugate
and two conjugate imaginary points
real.
ith
imaginary points cannot be harmonically conjugate regard to another pair of conjugate imaginary points. 3. An imaginary point is on one and only one real line and has one and
2.
186
79. Projective, affine,
CONIC SECTIONS
and Euclidean
classification of conies.
[CHA.V
Let us
regard a real plane TT as immersed in a complex plane TT', and consider all conies in TT' with respect to which the polar of a real point is always
a real line.*
Throughout the
in this sense.
conic
if
is
rest of this chapter the word "conic" shall be used The involution of conjugate points with regard to such a one in which real points are paired with real points. Hence,
a conic contains one real point, every real nontangent line through
this point contains another point of the conic,
and the conic
is real.
The
conies under consideration therefore fall into tsvo classes, the real
conies t and those containing
no
real point.
any two real conies are equivalent under the group of projective collineations. The same proposition holds also for any two conies of the other class, as we shall now prove. Let two such conies
By
76, Vol.
I,
be denoted by Cf and C\. On an arbitrary real line I they each determine an elliptic involution of conjugate points. By Cor. 4, Theorem 9, there is a projectivity of the line I carrying the involution determined
by Cl into that determined by C%. Any projectivity of the real plane which effects this transformation on I will carry G\ into a conic C%
which has the two conjugate imaginary points
with
of
I
I}
A^ on
in
common
G\.
collineation leaving
invariant will
now
;
carry the pole
with regard to C% to the pole of I with regard to Cl and therefore 2 (7 to a conic 8 Cl which has A v A 2 and the tangents at these points in common with Cf. Let L be the pole of I with regard to Of
carries
and L^
points
"be
MM which are conjugate with respect to G\ and harmonically

l
any real point
of
I.
By
Cor. 3,
Theorem
9,
there
is
a pair of
separate
and
Cl and harmonically separating

axis,
L as center, and carrying M' to M carries

16.
and also a pair M'M[ conjugate with respect to L and L v The homology with I as
Cl to
2
.
Hence we have
two real conies or any two imaginary conies with real polar systems are conjugate under the group of real pro-
THEOREM
Any
jective collineations,
* In
85 this condition
is
tion of the conic relative to a
seen to be equivalent to the condition that the equaframe of reference in TT shall be expressible with
If
the line
be taken as the line at infinity of a Euclidean plane
the argument above shows that any two imaginary conies are also conjugate under the affine group. Since these conies do not meet any real line in real points, they are analogous to ellipses no matter how the line at infinity is chosen. Hence we make the definition
:
DEFINITION.
An
imaginary conic with a real polar system
is
called
an imaginary
The
76,
ellipse.
results just established, together
with those stated in Ex.

as follows:
7,
and Exs. 14 and

17.
15,
37,
may be summarized
group
the conies
THEOREM
imaginary
Under
the affine
classes,
with real polar
systems fall into four

ellipses.
Any
parabolas, hyperbolas, real ellipses, two conies of the same class are equivalent.
conies
Under the Euclidean group

not meet
/
must be
characterized
by
their
relations to the circular points 1^
I2
Since a real conic which does
in real points meets it in conjugate imaginary points,
any real conic through 1^ also contains JT and is therefore a circle. For the same reason the imaginary conic determined by an elliptic polar system must contain /2 if it contains J1
.
imaginary ellipse with respect to which the pairs of conjugate points on ln are pairs of the absolute involution is called
DEFINITION.
An
an imaginary THEOREM
are similar.
Proof.
circle.
18.
Any
two real
circles
or
any two imaginary
circles
Let the centers, necessarily

2
be
and
respectively.
2 real, of two circles C and The center O l may be transformed to Oz by a

2
translation
T r This carries C to a circle C*. Any real line I through 2 meets Cf in two points C: and C2 and If in two points JK"a and 2 Since each of these pairs is harmonically conjugate with respect to 2 and the point at .infinity # of I, the homology T 2 with 2 as center
O2
and
real.
as axis
which
carries
C:
to
J^
also carries
C2 to
if
This homol(7
2,
ogy evidently
If
(
sider
are z K^, are pairs of conjugate imaginary points, conC^C^ and K^KZ with 77) the real pair of points PP' harmonically conjugate
carries all real points to real points
C1?
regard to
and 00* and the real pair QQ' harmonically conjugate and 00*. The homology T 2 must carry P and P' with regard to to Q and Q' and therefore carries all real points to real points in
C^
K^
Now
its
the conic
2
<7
2
is
fully determined
by
its
,
center
carries
and
is
fully
determined by
Ix Ja> 1^,
points Ilf I2 Ifa and

to If
2
.
Cv
a
.
<72
and
Hence
T2
C? to
19.
K\
The product
T2T
carries
C2
THEOREM
Proof.
Any
two parabolas are similar.

2
Let
C 2 and
be two parabolas
and
let
CM
and If, be their
points of contact with /.
and
let
2 1 ((7
)=
(7f.
Let
K
;
be any rotation carrying Cm to If* m be the conjugate of If, in the absolute
Let
involution and let c be the ordinary line
and C
its
point of contact
also let
Kx
tangent to
2
,
and
through IfM tangent to C% k be the ordinary line through
its point of contact.
The translation
FIG. 62
carrying
C to JT carries
J
Any
C'
axis,
line
and
2 c to Tc and C? to a conic C touching Zw at a through K, not containing JfM or &., meets (7 2 in a 2 point
.
in a point JT'.
and carrying
C' to JT' carries
The homology T with .BT as center, L as g q to K\ The product T 8T 2T i is a
similarity transformation carrying
Cz
to Jf 2 .
No
bolas.
theorem analogous to the last two holds for ellipses Suppose an ellipse or a hyperbola C* meets l in C a
another ellipse or hyperbola
JP meets
it
(7 2
in
and hyperand C and If and K" In oie a
similarity transformation carries
and
into
and If
^
existsTfr'
t7
^ and
an
*
"to to
o a conic cj
^ whi v ch ?passes
Car
and
through
^ ^
This
analogous to the proof of Theorem 18 it can "be shown that if Cf and K* are both real ellipses, or both imaginary ellipses, or both
bolas, there is a similarity transformation carrying G\ to K*.
hyper-
Hence
real ellipses or two imaginary ellipses or two are hyperbolas which meet lx in pairs of points C^CZ and
20.
THEOREM
Two
similar if and only if
B (J/
K^K
2,
C C2 }=
1
B (1
EXERCISE
hyperbola for which
B (/!/
2,
^1^2) =
1 is rectangular (Ex. 3,
73).
2 80. Foci of the ellipse and hyperbola. Let C be any hyperbola or real or imaginary ellipse, and let l v 1 2 be the tangents to (7 2 through J and ls l^ the tangents to (7 2 through / The circular points I , I
t
,
.
are one pair of opposite vertices of the complete quadrilateral 1 1 II Let the other two pairs of opposite vertices be F^FZ and F[F respecthe point tively (fig. 63), let a be the line F^FZ , I the line F[F'^, and of intersection of a and &. Also let x and JB X be the points at infinity
of the lines
a and
2
.
6 respectively.
The
triangle
2
OA n Bm
.
is
self-polar
Hence is the center of C and is therefore real. with respect to C 3 be any real point not on l v 12 l s l^ or C Let By the dual of the
2 Desargues theorem on conies ( 46, Yol. I) the tangents to C through are paired in the same involution with XIlt XIZ and XFV XFZ and
XF[, XF'y The double lines xlt xz of this involution are harmonically 2 Hence conjugate with regard to XI^ XI2 and to the tangents to (T
.
and they are paired both in the involution of orthogonal lines at 2 the involution of lines conjugate with respect to C at X, Hence
by Cor.
2,
Theorem
if
9,
orthogonal lines on
In particular,
X which are conjugate with regard to C X follows that a and 5 are real and
it
z
.
xl and xz
are real, and are the
unique pair
2
.
of
are the
Hence A* and only pair of orthogonal and conjugate diameters of C Bx are also real. If is not on a, "b, or L, the lines a^ and # 2 meet a
in a pair of real points

are
v z distinct from A a and 0. Since : and 2 and A n 0, harmonically conjugate with respect to the real pairs z
X X
X^
they are either real or conjugate imaginaries.

conjugate imaginaries,
But
since
Jx and I2
are
by Theorem 14
if
one of the pairs
F^FZ and
F[F' Z
is
a pair of real points, the other is a pair of conjugate imagi-
naries,
and conversely. Hence the notation may be so assigned that and a are real and F[ and F' are conjugate imaginaries. z
Let
points in
and A^ be the points in which a meets C* and B1 and Bz the 2 which b meets C By construction neither of the lines a
.
and
can be tangent to
C2
so that each of the pairs
AÂ Z and B^B^
is
either real or a pair of conjugate imagiuaries.
In case
C2
is
an imaginary
ellipse,
both
AÂ Z and B^BZ

is
are neces-
sarily pairs of conjugate imaginaries.
2 In case C
a real ellipse, the
Hue
is
L does not meet it in any real point, and hence 0, the pole of ?,, z an interior point. Hence both a and b meet C in real points. 2 Hence if (7 2 is an ellipse, A I} A 2 B^, B2 are all real. Whether C is
,
are conjugate a hyperbola, the tangents to <7 from 1 imaginary lines since they join the real point 1 to the conjugate
an
ellipse or
FIG. 63
imaginary points
Ix and J2
Hence
is
interior to
C2
as is also
z by a like argument. Hence the line F^FZ meets C in real points. Hence if C 2 is a hyperbola, A^ and A z are real. But if <7 2 is a hyperis an exterior point, and hence A*,, which is bola, harmonically separated from by A^ and J 2 must be an interior point. Hence 5,
,
the cole of A*, does not meet
C2
in real points, and consequently

!
and B^ are conjugate imaginaries. Let the polars of Z\, v F[, z relative to
(7
be denoted by dv dz ,
d',
A[ respectively.
;
Then d and d2 being the

v
polars of real points are
real
and since their point
of intersection is polar to a, it is -#,,
and
hence they are parallel to
b.
In like manner d[ and d' z pass through
A* and
are conjugate imaginaries.
conic
(7
nor, or secondary, axis.
a being called the major, or principal, axis and & the Each of the points Fv fz F[, F' is called a z
,
;s, and each of the points oh of the lines d v d2 d[ } d' z

,
A A
I}
2,
B^,
a vertex, of the conic
C2
is called a directrix of
C2
TIG. 64
EIG. 65
In the course of the discussion of the complete quadrilateral have established the following propositions
:
THEOREM
,icli
21.
If C
is
is
not a
circle, its
a hyperbola or a real or imaginary ellipse axes are the unique pair of conjugate diam-
which are mutually perpendicular. Two of the foci and two of the are real. The real foci lie on the major axis and the real rectrices are perpendicular to it. The other two foci are conjugate
rs
rectrices
.aginaries
terior
points
and lie on the minor axis. If C is real, the real foci are 2 and the real directrices are exterior lines. If C is a
2
%l ellipse, all
v vertices
e
the four of the vertices are real; if C is a hyperbola, on the major axis are real and those on the minor axis
conjugate imaginaries.
The two tangents to C z through

iirs
pass also through
It and J2
of conjugate lines at JP are separated harmonically
?o
tangents and hence meet
nible points are

is
Jx and Jg
If
we
in pairs of limit attention to real elements,
by these the involution whose
lines with expressed by saying that the pairs of conjugate which pass through a focus are orthogonal. Conversely, are conjugate with the pairs of orthogonal lines at any point 2 lines at spect to <7 , the double lines of the involution of orthogonal
may be
spect to
(7
192
CONIC SECTIONS
to coincide
[CHAP.V
of
would have
with the double
lines of the involution
conjugate lines, and hence
P would be a focus.
Hence
THEOREM
22.
ellipse are the
The real foci of a hyperbola or a real or imaginary unique pair of real points at which all pairs of contangent to the four minimal lines l^ lz la> ^ Hence the pairs of tangents to these conies
,
jugate lines are orthogonal.
The
set of all conies

(
form a range
47, Vol.
I).
through any point P not on the sides form an involution among the pairs
of the diagonal triangle of
OA n Ba
which
are the pairs of lines
i s O n C\ there is and PF[, PQ- Now if Z Z V only lt one tangent to C" at P, and this tangent is therefore a double line of the involution. This and the other double line have to be harmon-
PI PI
PF PF
2
conjugate with respect to PI^ and PI2 that is, if C and P two double lines have to be orthogonal. These double lines must be harmonically conjugate also with respect to PFl
2
ically
are real, the
and
PFf
Thus we have a result which may be expressed

76):
as follows
(ctEx. 12,
THEOREM 23. The tangent and the normal to a real ellipse or hyperbola at any real point are the bisectors of the pair of lines joining this point to the real foci.
In the proof of this proposition we have excepted the vertices of the conic, but the validity of the proposition for these points is selfevident.
Another proposition which follows directly from the

is
discus-
the following, in which we make use of the fact that the pair of real foci determines the pair of imaginary ones, and vice versa.
sion above
24. DEFINITION. The system of all conies having two imaginary foci in common is a, range of conies of Type I. The, two conies of the set which pass through any real point have orthogonal tangents at this point. Such a range of conies is called a real or two
THEOREM
system of confocal conies or of confocals.
The construction for the foci which has been considered in this section, when applied to a circle, reduces to a very simple one. The tangents to the circle at Ti and 7 meet in the center of the circle. The center of the circle 2 is therefore sometimes referred to as the focus and the line at infinity as the
directrix.
In the
section.
rest of the chapter the foci, center, directrices,
aud axes of
aii ellipse or a hyperbola will be denoted
by the same
letters as in this
The notation has been assigned so that for an ellipse the points
are in the order
^
2
and
for the hyperbola in the order
where J^ and
denote the points
of intersection of
t.Ti
axis with the directrices d l and d z respectively. 2 81. Focus and axis of a parabola. Let ^'
1
it is
and /2
with
tangent to L, there are two let these be respectively

;
orc l;>
Let their point

2
of intersection be de
<7
by L I and
respectively,
the point
a,
of contact of
C 2 with L
than A*, in
be
\
and the
point, other
is called tne jocus, me HUB M, me DEFINITION. The point irX the line the axis, the point ^ the vertex, of the parabola
I,
A*
as follows:
2
That the focus, directrix, etc. of a parabola are real may be proved The transformation from pole to polar with regard to
transforms the absolute involution to an involution of the lines
(7
through
Ax
and transforms /x and J2 into

at
A L
ao
and
AnL
respectively.
The involution in the lines
Aa
is
l
perspective with an involution

2
among the points
of
C which has L and
as
double points. Hence
L and L^
are conjugate
10 the imaginary points. Hence by Theorem
line
is
real.
is real,
Hence
its
pole, F,
is
real.
Hence
the line a joining
F to A*,
and also the point A. 2 Since the two tangents to G through
pass through
/ and /
a
2,
perpendicular. Conversely, if the pairs of conjugate lines at any point are orthogonal, the tangents
any two conjugate lines through

this point
F are
1^
through
must contain
and J2 respectively. Hence
the only such point.

is
Since the tangents through

all real points
is parallel to d,
F are
imaginary,
F is F
of
interior to
G'
and hence
on d are
exterior.
The tangent
of
at
and hence by the construction
d perpendicular
C2
2 G''
of
Since the tangent at any other ordinary point is not parallel to d, it follows that the line a is the only diameter which is perpendicular to its conjugate lines. These and other
to a.
obvious consequences of the definition
may be summarized as
is
follows
THEOREM
is
25.
The axis of a parabola

its
real
lines.
and
is
the only
a,
diameter perpendicular to all
conjugate
The focus of
parabola at which
the
real
and
lies
on
the axis.
The focus
is the
unique point
all
pairs of conjugate lines are orthogonal.
It is interior to
the focus,
parabola.
The directrix
to the axis.
is real, is the
polar of
and
is
perpendicular
All real points of the directrix are exterior to the parabola. The vertex is real and is the mid-point of the focus and the point in which the directrix meets the axis.
The system
of all conies
tangent to ^ and
I)
I,,
and
to
at
A* forms a
range of Type II
(47, Vol.
axis.
which consists
of all parabolas having
The pairs of tangents to these conies the plane are by the dual of Theorem 20, Chap. V, Vol. I, the pairs of an involution in which PIV is paired with PIZ and PF with PA n The tangents to the two conies of the range
as focus
and a as
through any real point
P of
which pass through P are the double lines of this involution and hence separate P/t and P/ harmonically. Thus we have 2 THEOREM 26. The parabolas with a fixed focus and axis form a
range of Type II. The two parabolas of the range which pass through a given point have orthogonal tangents at this point.
The tangent
other.
to either parabola through
is
therefore normal to the
Since these two lines separate
PF and PA X harmonically, we have
THEOEEM 27. The tangent and the normal to a parabola at any point are the bisectors of the pair of lines through this point of which One 'OO.S.tes tJh.fntl.nll. fnr.i/.s n.in.d. tJia ntJi.oif no n rliinvnotaw
tJh.t>.
81]
FOCAL PROPERTIES
EXERCISES
196
of
P is any point of an ellipse, the normal at P is the interior bisector 4- F^PFV If P is any point of a hyperbola, the tangent at P is the interior
1.
If
bisector of
2.
4 t\PF
lines
At any
orthogonal
of a
nonfocal point in the plane of a conic there is a unique pair of which are conjugate with regard to the conic. In case of an
ellipse or a hyperbola these lines harmonically separate the real foci. In case parabola they meet the axis in a pair of points of which the focus is
the mid-point. 3 For any point P of an axis of a conic there is a unique point P' on the same axis such that any line through P is orthogonal to its conjugate line through P'. The pairs of points P and P' are pairs of an involution
whose double points are the foci of the conic, or, in (called a focal involution) case of a parabola, the focus and the point at infinity of the axis. If P and P' are on the minor axis, 4-PF^P' is a right angle. If the conic is a parabola,
F is the
4.
mid-point of the pair PP'. confocal central conies having a real point in common, one an ellipse and the other a hyperbola.
Of two
is
5. The tangents at the points in which a conic a focus meet on the corresponding directrix.
is
met by
a line through
6. If
two conies
7.
two conies have a focus in common, the poles with regard of any line through this focus are collinear with the focus.
to
the
Let
a
P be
any
point of a conic,
If
and Q the point in which the tangent at
P meets
directrix.
F is the corresponding focus,
&PFQ,
is
a right angle.
through the two real foci and a point P of a conic, it will have the two points in which the tangent and normal at P cut the other
8. If a circle passes
axis as extremities of a diameter.

9. If a variable
respectively,
and
F is
t
tangent meets two fixed tangents in points a focus, the measure of &PFQ, is constant.
and Q
the pair of lines
to
be two tangents of a central conic meeting in a point T; is congruent to the pair TFZ t v l 11. The line joining the focus to the point of intersection of two tangents a parabola makes with either tangent the same angle that the other tangent
10.
Let ^ and
TF
makes with the

12.
axis.
Let p be a variable tangent of a parabola, and P a point of p such that the line PF makes a constant angle with p. The locus of P is a tangent to
the parabola.
13. 14.
The
foci of all parabolas inscribed in a triangle lie

is
on a
circle.
circumscribed to a triangle which parabola passes through the focus.

circle
circumscribed to a
15.
The
circles circxim scribing four triangles
whose
sides
form a complete
16.
Let
The
to
lines
P be any point coplanar with, but not on an axis of, a conic C2 which are at once perpendicular to and conjugate with regard to C"
.
The axes
the lines through of C 2 are tangents of this parabola. 17. If P and P' are a pair of one focal involution of a central conic, and
are the tangents of a parabola (the Steiner parabola).
Q and Q' a pair of the other, P, P', Q, Q,' are on an equilateral hyperbola, which may degenerate into a pair of orthogonal lines. 18. Given five points of a conic, construct by ruler and compass the center, the axes, the vertices, the foci, and the directrices. Construct the same
elements
when
five
tangents are given.
82. Eccentricity of
conic.
Let
2
F he
a real focus,
and d the
cor-
not a circle. Let a he the responding directrix, of a conic C which is 2 major axis of <7 , aud h the line parallel to d such that if a meets d
in a point D,
Then d
is
is the mid-point of the pair and h in a point H, the vanishing line ( 43) of the harmonic homology T with
FK
F as
center and
as axis.
Since
F
2
the circular points.

circle
_BT
a focus, the tangents to (7 through pass also through 2 Hence the transformation F changes (7 into a is any point of the circle, Now if with as center.
is
P' the point of C to which P is transformed "by T, and D' the meets d, it follows point in which the line through P' parallel to by Cor. 2, Theorem 21, Chap. Ill, that
z
FD
Dist (P'F]
~ Dist (PF}
Dist (FD}
'
Dist (P'D'}
Since Dist
(PF} and Dist (FD} are constants, it follows that THEOREM 28. DEFINITION. The ratio of the distances of a point of a conic to a focus and to the corresponding directrix is a constant
The
to
conic C" is a parabola if and only if the circle d, the vanishing line of F. In this case
Dist (FD)
is
tangent
= Dist (PF),
The conic
(7* is
and hence the and only if K 2
eccentricity
is
unity.
points.
a hyperbola
if
meets d in real
In this case
Dist (FD)
< Dist (PF),
and hence
remark
the eccentricity is greater than one. Applying a like to the ellipse we have
29.
THEOREM
according as
unity.
conic section is
less
an
ellipse,
hyperbola, or parabola
its eccentricity is
than, greater than, or equal to
circle is
said to have eccentricity zero, because
if
and
be held
be moved so as to increase constant, and Dist (PF)/ Dist (FD) approaches zero.
FD
"without limit, the ratio
The
relatively to either of
eccentricity of a hyperbola or an ellipse is evidently the same its real foci, because the two foci and the
corresponding directrices are interchangeable by an orthogonal line reflection whose axis is the minor axis of the conic.
As an immediate
THEOREM
30.
corollary of the definition of eccentricity real conies are similar if
we have
they have
Two
and only if
the same eccentricity.
On comparing
eccentricity
is
this
theorem with Theorem 20,
it is
evident that the
a function of the cross ratio of the double points of
the absolute involution and the points in which the conic meets fc. As an example of this relation we have (by comparison with 72)
the theorem that any two hyperbolas whose asymptotes make equal angles have the same eccentricity. The formula connecting the eccentricity of a
hyperbola with the angular measure of its asymptotes is given in Ex. 7, below, and the formula for the eccentricity in terms of the cross ratio referred to in Theorem 20 is given in Ex. 9.
Since a real focus of
a real focus
(e.g.
any conic
is
an interior point, the line through
64) perpendicular to the principal axis meets the conic in two points, Q^Q^ The number Dist (Q^Q^) is evidently
2 , fig.
the same for both foci of an ellipse or hyperbola, and hence
is
a fixed
DEFINITION.
The number p
2
Dist (Q^Q^)
.
is
called the parameter,
or latus rectum, of the conic
(7
In the following exercises e will denote the eccentricity and p the parameter of any conic. For an ellipse or hyperbola a denotes Dist (OAJ and c denotes Dist (OFJ. For an ellipse & denotes
Dist (OB^.
For a hyperbola
& denotes
Vc 2
a\
In
all cases
a radical sign indicates a positive square root.
EXERCISES
1.
of a hyperbola, Dist (F1 P) = and 2 = //2 + c 2 ellipse Dist (-8 1 .F 1 ) 4. Dist (.-1 jFj) Dist (7^,4 2 ) = A 8
2. 3.
P is If P is
If
any point any point
of
an
ellipse,
Dist (F^P}
Dist
Dist
(F2 P) = 2 a. (F2 P) =
2 a.
In an
In an ellipse or hyperbola e = - and p = In a parabola Dist (A.F) =jt>/4. 7. The measure (67) of the pair of asymptotes of a hyperbola determined by the equation o
5.
6.
is
cos
8.
9.
For an equilateral hyperbola e = V%. The cross ratio R (C^Cg, /t /2 ) = ^2 referred to in Theorem 20
is con-
nected with the eccentricity by the relation
4*
in case of an ellipse, and by v J
in case of a hyperbola.
10.
e
2
1-2/fc+yfc2
circles
Let A* and
be the
with
as center
and passing through the
A^ and B v respectively, of an an angle of measure with the ray

vertices
ellipse,
respectively.
Then the
line
X
of
through
OA Y parallel
and let a variable ray making meet these circles in X and Y

to
OA
meets the line through
parallel to
OB
in a point
of the conic.
If x
and y are the coordinates
relative to the axes of the conic,
x
6
is
a cos
0,
I sin 0.
called the eccentric anomaly of the point P. 11. Relative to a nonhomogeneous coordinate
cipal axis of a conic is the z-axis,
equation of a parabola,
the
ellipse,
. .2
system in which the prinand the tangent at a vertex the y-axis, the and hyperbola, respectively, can be put in
form
TOX
-4-
12.
Relative to the asymptotes as axes, the equation of a hyperbola
may
be written
_
13. Relative to
a*
+
,
any pair
of conjugate diameters as axes,

,
an
ellipse
has the
equation
^ ^
\
=1
~t
'
and a hyperbola,
'
X2 a' 2
1J
Z 2
= 1.
b'
r If A is a point in which the x-axis meets the conic, Dist (OA ) = a'. In the case of an ellipse, if B' is one of the points in which the y-axis meets the = b'. conic, Dist (OB')
14. 15.
The measure of the ordered point The numbers a' and b' satisfy the
ellipse
triads
OA'B'
is a constant.
conditions a' 2
b /2
a2
6 2 in case
in case of a hyperbola. 16. The equation of a system of confocal central conies relative to a system of nonhornogeneous point coordinates in which the axes of the conies
I"*
of
an
and a"2
a2
b2
are x
and y
is
^rx + jTTx- 1
i
x2
'
where A
u^
[w
,
a parameter. In the homogeneous line coordinates such that + u 2 y + U Q = gives the condition that the point (x, y') be on the line z 2 u v 2 ], the equation of a system of confocals is u z = (a 2 A) w + (b A)w|.
is
17. Relative to point coordinates in which the origin is the focus, y = the axis of the parabolas, and x = perpendicular to the axis, the equation of a system of confocal parabolas is
2 (p
A) x
+ A (p
A)
0.
is (cf.
In the corresponding homogeneous line coordinates this
Ex. 16)
83.
literature will convince
Synoptic remarks on conic sections. An inspection of the one that it would not be practical to include
a complete list of the
known
metric theorems on conic sections in a

derived, however,
book like this one.
The theorems which we have
are sufficient to indicate
how the
rest
may
be obtained either directly
as special cases of projective
and
theorems or as consequences of the focal theorems given in this chapter and Chap. III. The theorems on conic sections have been classified according to
affine
the geometries to
tary
which they belong. The most general and elemenwhich we have considered are those which belong to the proper
geometry
(
projective
17), the
geometry corresponding
to the projeo
200
CONIC SECTIONS
particularly in Chaps.
[CHAP.V
of this class are given in Vol.
I, V, VIII, X. second large class contains those theorems which belong to the
affine
geometry in any proper projective space.

in Chap.
III.
These are treated
somewhat fully The theorems
of the class considered in 74, 75 of this chapter belong to the projective geometry of an ordered space. The theorems 76 belong to the projective geometry of a real space. Finally, of
in
80-82 we have been considering theorems

of a real space. It is quite feasible to
of the
Euclidean
geometry
make a much
finer classification of
theorems
This would mean, for example, distinguishing those properties of foci which hold in a parabolic metric geometry in a general space, then those which hold in an ordered space, and then those
on
conies.
which
are peculiar to the real space.
The theorems which have been under discussion in the remarks

above refer in general to figures composed of one conic section and a
finite
number of points and lines. Theorems regarding more than one conic at a time have not been considered in any considerable
number, and the theory of families of conies has not been carried beyond pencils and ranges. For an outline of this subject the reader
is
referred to the Encyclopadie der
Math. Wiss., Ill 01,
56-90.
EXERCISES
diagonals of the rectangle formed by ths tangents at the vertices of an ellipse are conjugate diameters for which a" = //. The angle between this pair of conjugate diameters is less than that between any other pair of
1.
The
conjugate diameters. For this pair of conjugate diameters a'+ mum. It is a minimum for a' = ,&' = I.
2. If
b' is
a maxi-
two orthogonal diameters of a conic meet
it
in
P and
Q,
tor
an
,,.
ellipse, p
and
OP OP
OQ2 OQ2
a2
for a hyperbola.
3.
locus of a point from which the two tangents to a conic C2 are 2 orthogonal is a real circle in case C is an ellipse or a hyperbola for which a > ?; is a pair of conjugate imaginary lines through the center and the cir;
The
cular points in case

in
-isii
C2 is
C2 is a hyperbola for which a = b; is an imaginary circle a hyperbola for which a < b is the directrix in case C z is a parab:
tangent to a central conic is met by the lines through a focus a fixed angle with it in the points of a circle. In particular, the us of the foot of is a circle. perpendicular from a focus to a
4.
A variable
ich
make
If
t
tangent
5.
"
is
a variable tangent of a central conic, Dist (7^) Dist the other tangent parallel to t, Dist Dist (F/) = b 2 (7<\/)
is
.
(F2 t) =
I".
6.
If /' is
a,
focus of a conic
and
Pv P
the points of intersection of an
itrary line
through
with the conic,
constant.
7.
the tangent to a conic at a variable point meets the axes in two uts J\ and 7'2 and the normal at meets them in N^ and z then
If
,
Dist (P7\)
Dist (P7'2 )
= =
Dist Dist
(PNJ J (7 F )
t
Dist Dist
(PN2 )
(PP2 ).
8.
at
'er
There is a unique circle which osculates* a given conic at a given P. This is called the circle of curvature at P. Its center is called the
of curvature for P and lies on the normal at P. Construct by ruler and compass the center of the
circle of curvature
9.
in arbitrary
LO.
y
The
circle of
point of a given conic. curvature of a conic
one other point, Q.
The line
PQ is
.
at a point meets C 2 in one and the axis and the point P the center of an
C2
ion which transforms

this elation into
into
2
.
C 2 The
center of curvature
is
transformed
the pairs of
the center of the involution on
C2 in which
C2
logonal lines at
LI.
meet C
The tangent and normal
at
any point
of a conic
are both tangent
;he Steiner
:ontact of the
ire of
C2
parabola (Ex. 10, 81) determined by this point. The point normal with the parabola is the center of the circle of curat P, and the point of contact of the tangent with the parabola
.
he pole of the normal with respect to C2 (For further properties of the le of curvature, cf. Encyclopadie der Math. Wiss., Ill Cl, 36.) .2. The polar reciprocal of a circle with respect to a circle having a
it
as center is a conic
having
as a focus.
(A
set of
J.
theorems related
Russell referred
his
one will be found in Chap. VIII of the book by

73.)
W.
14.
Focal properties of collineations.
The
focal properties of conic
ions are closely related to a set of theorems on collineations ie of which are good treatgiven in the exercises below. of it of the subject is to be found in the Collected Papers
J.
S.
Smith, Vol.
I,
p.
545,
and
further references in the

5,
9.
jyclopadie der
Math. Wiss., Ill
AB
202
CONIC SECTIONS
[CHAP.V
Let II be any real protective collineation which does not leave la so that II (p) = ln invariant, and let p and g be its vanishing lines
;
and
1
II (/)
= Qit
If Ii
and J2 are the circular

2)
points, let II"
(/j)
=P v
II- (Iz )
lines
=P
II (J ) a
= <^, n (J =
Q,.
By
the theorems of
78 the
.PI/!
and
PI
S
2
.
in a real point
meet in a real point A v and J^/2 and P^ meet = J52 it is clear that If II (A^ = JB 1 and II (A 2 )
,
the
JjZ,,
complete quadrilateral whose pairs of opposite vertices are P P is transformed into one whose pairs of opposite 1 2) AjA 2
vertices are
Q^^ f^
2,
^^
The
following propositions are

exercises.
now
easily verifiable,
and are stated as
EXERCISES
that any ordered pair of lines meeting at A is transformed by II into a congruent pair of lines. A z is such that any two lines meeting in A 2 are transformed by II into a symmetric pair of lines. No other points
1.
l
is such,
have either of these properties.

2.
2?!
Every conic having a focus at
A:
or
A2
goes to a conic with, a focus at
or
3.
having A^ and A z as foci is transformed by II into the range of conies with B i and B z as foci and this is the only system of confocals which goes into a system of confocals. 4. The pencil of circles with A v A 2 as limiting points is transformed by II into that having B v Bz as limiting points and these are the only two pencils of circles homologous under II. The radical axes of the two pencils are the two vanishing lines. 5. If P is any point and II (P) = P', then the ordered point triad A^PA^
The range
respectively. of conies
similar (but not directly similar) to the ordered point triad B^P'BZ 6. At a point of a Euclidean plane there is in general one and only one pair of perpendicular lines which is transformed into a pair of perpendicular
is
.
lines
7.
by a given affine collineation. In any two projective pencils
of lines there is a pair of correspondline pairs
ing orthogonal pairs of lines.
The
which are homologous with

l<*
congruent line pairs form an involution. 8. Any projective collineation which does not leave
ible as a product of a displacement
invariant
is
express-
and
a homology.
85.
Homogeneous quadratic equations in three variables. Eevers-
ing the process which is common in analytic geometry, it is possible to derive certain classes of algebraic theorems from the theory of conic sections. shall illustrate this process in a few important
We
The general homogeneous be written in the form

o
>
equation
of
the
second degree can
a oi a'o
fl
'i ~"~
ô
(6)
where a tj
=a
jt
Let us
first
suppose that
(7)
% a
98, Vol.
I, it
aoo a oi ao 2 au a u w a n a 22
In
has been shown, from the point of view of general
projective geometry, that every protective polarity is represented
by
a bilinear equation of the form

-J-
a Q1 x Q xl
-f-
ô
(8)
where a y
It
= a^.
also
and where
3= 0.
was
shown
to the condition
that every bilinear equation of this form, subject that the equation in 3= 0, represents a polarity
;
point coordinates of the
fundamental conic
of the polarity is (6),
which
obtained from (8) by setting x\ of this conic in line coordinates is

is
f ;
and that the equation
(9)
where
A^
is
the cof actor of
a.j
in A.
The
coefficients
are elements of the geometric
number system.
and
Therefore in the case of the real plane they are real numbers,
we have
THEOEEM
such that a
-
31.
Every equation of
aj{ and
the
form
a,
(6)
with real
coefficients
A^
represents
lines.
conic whose polar system
Conversely, every conic with transforms real points into real regard to ^vhich real points have real polars has an equation of the
In
in one of
79 we have seen that any conic having a real polar system is two classes, and that any two conies of the same class are
projectively equivalent.
(10)
is
Now
2
tf
it is
obvious that
2
-t-* 2
=0
the equation of an imaginary conic, and that

(11)
otf+atf-aÔ
Hence we have
is
the equation of a real conic.
THEOREM
32.
Any
quadra-tic equation in three homogeneous vari-
ables whose discriminant
does not vanish is reducible

the variables to the
~by
real linear
(10) or to
homogeneous transformation of
the
form
form
(11).
2
Algebraic criteria to determine whether a given conic C whose is in the form (6) belongs to one or the other of these classes may easily be determined by the aid of simple geometric considera-
equation
tions.
In case
(7
contains
no
real points, the line x
has no real
point in
(which is on the line XQ = 0) is on no real tangent to it. On the other hand, if the line XQ = z 2 contained no real point of C and (7 were real, this line would consist
with
it,
,
common
and the point
u^
entirely of exterior points,
through the point

conditions that
of C,
2
^=0.
is
and hence there would be a tangent to (7 2 Hence a pair of necessary and sufficient
Cz
contain no real points are (1) #

.
is
on no point
wx = = Substituting x
and
(2)
on no tangent of C z in (8), we have the equation of an and
#=
involution
and only if A w 0. By a dual argument the necessary and sufficient condition that there be no 2 real tangents to (7 through the point u^= is
which, by
4, is elliptic if
>
applied to
(9),
^ 00
(13)
By a well-known theorem on determinants

this
Tfirlnr.p.s t,n
(or a
simple computation)
5]
ALGEBRAIC THEOEY
205
Hence we have
THEOREM
33.
The imaginary
conies are those
for which
AM >
and
the real ones are those
and & U
-A>
0,
for which not both of

3= 0.
these conditions
are satisfied
and for which A

it is
In these conditions
obvious that
0, 1, 2,
by Ay and a^ where
Let us
i,
j=
A m and
^4
a
i
=
may
/.
be replaced
the case in
provided that
now
all
investigate the cases
where
= 0,
and
first
which not
by x
all
the cofactors
=
A m A n A 22
,
are zero.
To
fix
the notation,
is
suppose that A QO x1 00
0.
Ql ,
Then the
,
bilinear
=A
=A
x2 = A 02 no matter what values

from (A OQ
equation
(8)
satisfied
are taken
by
x' , x' v x' r
Hence
in this case (8) determines a transformation, F, of

,
the points
2 ) distinct (x' , x[, x!
(Aqp
A A
ol ,
02 ).
A collineation which transforms (A^ An AJ to (1, 0, 0)

,
A 01 AQZ
,
into lines through
must reduce
(14)
It is to
(8) to
\/
i7 h W
A
11^1+ 18*1*2 1,1 A _ =

SYI'
I-
/Yi
rtn'
II
/\
13 *
"21 *A
*,
*4b
4t
be noted that
n
l
&"
because
(x'
,
if
this determinant vanished,
F would
transform
all
points
x(, x' 2)
into a single line,
and hence
Am
would vanish. Hence F
transforms any point
into (x' , x[, x' 2)

,
the line paired in a certain invo,
lution with the line joining (x' to (A , A z} Q x[, x' m ol A Q2 ). The double lines of the involution must satisfy the quadratic equation (6).
A=Q
Comparing with the definitions in 45, Vol. and not all the cofactors A m A n A^ are
,
I,
we have
that
when
zero, (6) represents a
degenerate conic consisting of two distinct lines and that (8) represents the polar system of the conic. Since the lines represented by (6) are the double lines of a real involution, they are either real or a pair of
conjugate imaginaries.
In the
first
case (6) can evidently be trans-
formed by a collineation to
n (10)
f)\
-v.2 X 1
2 rz X
ft \J,
QTin in f/ho oor>nnrl
r>aaa
+-.A
points
if
and only
if
AH ^
0.
Hence the case where

if
(6)
represents a
if pair of real lines occurs Finally, suppose that A QO
and only
22
A ^0,
i{
for
i=
0, 1, 2.
= A n = A =A =
0.
In view of the identity,
this implies that all the cofactors A {J are zero, and hence that (8) represents the same line, no matter what values are substituted for Hence (6) represents a single real line (i.e. two coincident x' ,'x[, x! z
.
real lines),
and the polar system
(8)
transforms
all
points not on this

0, (6)
line into this line.
If tin's line be
transformed to z1 =
obviously
becomes
(18)
as*
0.
A degenerate
may
point conic
is
two
distinct or coincident lines.
These
always be represented by a quadratic equation which is a product of two linear ones. For such a quadratic A = 0, because if A 3= 0, the
equation has been seen to represent a nondegenerate conic.

the.
Hence
theory of degenerate point conies
is
equivalent to that of homo0.
geneous quadratic equations for which
A=
The complete
protective classification of conies, degenerate or not,

:
may now
ables
be stated as an algebraic theorem in the form

34.
THEOREM
may
v '/
be
homogeneous quadratic equation in three varireduced by a real linear homogeneous transformation,

2
vO
-
Any
-L \ ^
x X; J=
OC"\Aj'+
y J*
i'b
\J* "
J.
"
<
//
i .
QC" y
5-*
\J
to one of the normal forms (10), (11), (16), (15), (18). The criteria which determine to which one of these forms an equation (6) is reducible
may
be
summarized in
the
following table
Since the algebraic expressions in the above criteria determine conditions on the conic which are independent of the choice of coordinates and thus are invariant under the projective group, it is natural to inquire whether they are
algebraic invariants in the sense of 90, Vol. I. readily verify that A is a relative invariant of (6).
direct substitution will
Suppose
, , ,
we regard
,
(a 00 a u a 22 a ov a 1Q
the coefficients of (6) as homogeneous coordinates a 12 ) of a point in a five-dimensional space. Then A =
determines a certain cubic locus in this space the points on which represent degenerate conies. Now if there were any other invariant of (6) under the
would represent a locus projective group, say < (%), the equation < (a^-) in this five-dimensional space. But since each nondegenerate conic is projectively equivalent to every other nondegenerate conic, this locus would have to
theory of
A = 0. From this it can be proved, by the general represented by algebraic equations, that the locus of $ (0^-) = = coincides with that of A 0, and that hence < (a^-) is rationally expressible
be contained in the locus of
loci
in terms of A.
Thus
is essentially
the only invariant of (6) under the
projective group.
The question, however, arises whether there are not other rational func0. If there tions of the coefficients of (6) which are invariant whenever were such a function, say <f> (a y-), the conies for which would (a#)
A=
<
be a subclass of the degenerate conies which

all
is
transformed into
itself
by
complex projective collineations.
The only
class of this sort consists
of the coincident line pairs which are given by two conditions, A 00 = 0, n = 0. In view of the theorem that a locus represented by two inde-
pendent algebraic equations cannot be the complete locus of a single algebraic equation, this shows that there is no other invariant of (6) even
for the cases in
which
A = 0.
This reasoning could be expressed still more briefly by saying that, while the set of all conies is a five-parameter family, and the set of degenerate conies a four-parameter family given by one condition, the only invariant
pairs
subset of the degenerate conies is the two-parameter set of coincident line which have to be given by two conditions and so cannot correspond to
a single invariant in addition to A.
EXERCISES
1.
In case
,
A = 0,
(VI 00 VA 1V VZ 22 ),
the lines represented by (6) intersect in the point unless the three cofactcrs A ti vanish, in which case (6)
represents the coincident line pair

i.
+ -Sa^y = 0.
2.
of
In case (6) represents a pair of distinct lines, (9) represents their point intersection counted twice. In case (6) represents a pair of coincident
208
86.
affine
CONIC SECTIONS
[CHAP.V
Nonhomogeneous quadratic equations in two variables. theory of point couics corresponds to the theory of
The
(20)
-f
where the aâ satisfy the same conditions as in the last section. The theorem that any nondegenerate conic is an imaginary ellipse, real ellipse, hyperbola, or parabola, and that any two conies of the
class are equivalent under the affine group, translates into the following Any quadratic equation in two variables, for which A =(= 0^ is transformable by a transformation of the form
same
x
(21)
y'
= =
cijX
-j-
o^y
-f-
Cj,
=0,
into one of the following four forms

2
(22)
(23)
2/
(24)
(25)
To know
this it is
merely necessary
to observe that these equations
represent couics of the four types respectively. The criteria to determine in which class a given conic
C2
if
belongs
may
be inferred from the discussion in the last section

.
we
set
x = X /CG and y X /X It is then evident that A W > for an ellipse, A m = f or a parabola, and A m < for a hyperbola. Hence the
I Q Z
affine classification of cases
where
may
be summarized in the
following table
.4 = correspond, as hi the last section, to degenerGeometrically the types of figures are obvious, and to obtain the algebraic criteria we need only combine with considera-
The
cases
where
ate conies.
_ 'J.T
86]
ALGEBRAIC THEORY
209
As normal forms
/9G\ ^"0/
for the first six cases
we may take
X2
*
2
-J-
= =
r\
U,
(27) (28)
(29)
+l=0,
2
?/
a2
0,
"-! =0,
=0, **=0.
(30)
(31)
The
case of coincident real lines at infinity does not correspond to
any equation in nonhomogeneotis coordinates. Summarizing these results we have the following algebraic theorem
THEOREM
reduced
35.
Any
quadratic equation in two variables
may
be
to one and only one of the normal forms (22) (31) by a transformation of the form (21). The normal form to which it is reducible is determined by the criteria in the two tables above.
The question
of invariants of (20)
under the
affine
group may be
investigated in the
of the last section. The results problem in of such an investigation are given in the exercises below. There are no absolute invariants of conies under the nroiective
manner indicated the fine print at the end
for the corresponding projective
EXERCISES
L A
2.
and
A 00
In case
are invariants of (20) under the affine group. A w = 0, ^ n /a 22 an ^ ^22/ a n are invariants of (20) under
coordinates of the center of (20) are
the affine group.

3.
The homogeneous
(.4 00 ,
A ov A
0!1 ).
4. If
A 00
0,
the translation x
z
^j y
=y
A^
-^
transforms (20) into
^00
_ _ ax + 2 ax
5. If
and
A OQ 5*
0,
the asymptotes of (20) are given by the equation
a nxz
6.
2 a i2 xy
is
asz y 2
0.
Any
diameter of a parabola
parallel
to
a lt x
a 12 y
and
tc
87. Euclidean classification of point conies. With, respect to a non-
and homogeneous coordinate system in which the pair of lines x = is orthogonal and bisected by the lines x = y y &nd x = y, the transformations of the Euclidean group take the form (21) subject to
the conditions
(32)
a?
+ 0," = & + &,
0& +
1
*,&,-(),
and the displacements are subject to the additional condition

(33)
Since any ellipse or hyperbola

axes are
x=
and y
0,
parabola with the origin follows that any conic is congruent to a conic having one of the
following equations:
(34) I \
is congruent to one whose principal and since any parabola is congruent to a as vertex and y = as its principal axis, it
2!
-.2
+ + i =()i
!
'
7.2
'
(36\
/O>7\
y.
io
px =
n
0.
(o
i)
y
to
conies can be reduced by displacements are evident when one recalls that two pairs of non-
The normal forms
which degenerate point
if
the circular points and that two pairs of parallel lines are congruent the lines of each pair are the same distance apart.* By comparison
with the second table (A

(38) '
v
-+* y
a
2
0) in ~2
86 we find
2
=0,
(39)
a2
?L
+c
=0,
(40) \ )
(41)
(42)
tf=Q y
c
2
xz -
= 0,
x=Q,
*2 =0.
(43)
The group of displacements is extended to the group of similarity transformations by adjoining transformations of the form
~~
'
(44)
y
normal forms in which
= Ky
0.
>
Transformations of this sort will reduce the equations (34)

b, c,
(43) to
and
are all unity.

of these
The
conic
is
criteria for
determining to which
normal forms a
reducible under the group of displacements or that of similarity transformations are the same as those already found for the affine group. Two conies whose equations can be reduced to the same
normal form are evidently equivalent under the group of displacements if and only if they determine the same values for a and & or c
or p,
for a.
and under the Euclidean group if they determine the same value The numbers a, &, c, p are evidently absolute invariants of the
corresponding conies under the group of displacements, and (40) also under the Euclidean group.
and a in (38)
The problem
of (20) presents
of
determining
a, I,
c,
in terms of the coefficients
and will be left to the reader to be considered in connection with the exercises below and those at the
no special
difficulty,
end of the next section.
When 5, c,p are all unity, a is a function of the eccentricity given by 82. The same reference gives the conthe equations in Exs. 7 and 9, nection between the eccentricity and the invariant M /(a -f- a
V A
).
* The distance apart is the distance of an arbitrary point on one of the parallel lines from the other line. The formula for distance is applied to the case of a
pair
EXERCISES
1.
If
A?
and
A QO ^ 0, the
.
angular measure of the asymptotes

-
is 0,
where
tan 6
Moreover,
=|
log
R ( C C2
x
J^),
are the points in which the conic meets la,, and 7t and 72 are and J. 00 5* 0, these formulas give the angular the circular points. If A = the lines measure of represented by (20). Derive from this the formula for a
where
C and C2
x
in (38) and (40) in terms of the coefficients of (20). 2. A 00 and o n + o 22 are absolute invariants of (20) under the group of .'I a 22 ) under the Euclidean group. If A ^t. displacements, and 00 /(a n +
and
+ a 22 = 0, (20) represents an equilateral hyperbola if A = and n + G 22 = i* represents a pair of orthogonal lines or I* and an ordinary line. A 5* and A m & 0, the axes of (20) are If 3. 2 + 2 ) + (22 ~ a il) W = 1 2 (^
an
;
>
2/
where x and i/ are defined as in Ex. 4, 86. 4. For an ellipse the constants a and b are \l-r~r~ 00/Vl and A, are the roots of
<
>
,.
an(i
\/~j
r~ Aa
where A
(45)
X2 -(u+
ib
22
)*+ôo=0?
and %/
and for a hyperbola a and

(45)
is
are
(a u
-\|
--
The discriminant
l, at (0,
of
a 22 ) 2
4 a? s
5. If Jl 56
which
(46)
is
and -4 00 = 0, the parabola (20) touches the same as (0, a 22 a 12 ). The axis is
,
a la
a i;L )
a nx
+ a 12 y +
a oi
fl
n+
11
a o a ia a 22
88. Classification of line conies.
The
projective classification of line
conies
is
entirely dual to that of point conies
and so need not be con-
sidered separately. The affine classification, however, corresponds to a new algebraic problem. If the line coordinates are chosen so that
is
the condition that the point (XQ x^ x^ be on the line [u u^ wj, the point coordinates being the same as already used, we have the problem of reducing equations of the form (9) to normal forms by means of
,
transformations of the form
2122'
=
In
a!
&
:
Uo
1L -\- GL
These
'invariant.
are
If
the
transformations which
leave the line
[1,
0,
0]
and
6.
c.
d.
c
i
a
i
cn
a.
(47)
is
the same collineation as (21).

affine classification of
The
same
nondegenerate line conies
is
of course the
as that of nondegenerate point conies.
in terms of the equation (9)
express the criteria regarded as given primarily,* let us write
To
(48)
=.
A
where the Ay's are the
cofactor of
A
;J
in a.
(9), and let a y denote the The. point conic associated with (9) must have
coefficients
of
the equation
(49)
2JW;=is
the criteria already worked out, this parabola according as the value of
By
an
ellipse,
hyperbola, or
is
greater than, less than, or equal to zero
and, in the case of an
ellipse, real or
imaginary according as a n
>
or
an
^ 0.
Thus we have
The normal forms

(50) (51) (52)
(53)
for these four classes are respectively
<+<+<=
0, 0, 0,
0.
<-<-< =
w2
-< + < = <-u =

2
conies is dual to projective classification of degenerate line that of degenerate point conies, and therefore yields the following
The
three cases
(1)
two
distinct real points,
= 0,
a{i
0,
one at least
of
cc
aQO a n 22 being different from zero (2) coincident a a n = a22 = a.M (3) conjugate imaginary points,
, , ; ;
real points,
0,
a it
>
0]
for at least
one value of
affine
i.
For the
is
classification let us observe
that since
[1,
0,
the line at infinity, the condition that at least one factor of 0. The following criteria (9) represent a point at infinity is A m
are
now
evident.
a=Q
The normal forms

(54) (55) (56) (57)
for these cases are respectively
<+<=
ul
0,
+ u* = 0, <-< = 0,
^=0,
(58)
(59)
^ =0,
2
<=
0,
(60)
<=0.
EXERCISES
1.
The two
pairs of foci of (9) are the degenerate conies of the range

2 ^00 W + ^010 U 1 + yl 02 U U 2 + A lQUl u + (A^-p) + A lz u t u 2 + ^20 M2 M + ^21 M2 U + ( A 22~ P) M2 =
1
(61)
which are given by the values of p satisfying

(62)
A 00 p 2 -
(a u
a22 ) p
+a=
0.
The
discriminant of this Quadratic
is fa.,
n..-^ 2 4-
4nA.
2.
In case a
and
A 00 ^
0,
the distance between the points represented
3.
The normal forms
for line conies under the group of displacements are

a
(63)
2
a?u%
tfuj
0,
(64)
2
(65)
Uo
flX a2 u2
&X = 0,
0,
+ Wul = +
(67)
u2
fl
(68) (69)
ttf
2
&X = 0, cX = - &X =
+
0, 0,
ottl
(70) (71)
(72)
f-
C2 M|
(73)
Wl
= 0, = 0, = 0, = 0.
Here a, b, p have the same significance as in (34) -(37); 2 Td is the distance between the two points represented by (67); 2 k is the distance between the two points represented by (69); c is expressible in terms of the cross ratio of
the circular points and the two points represented by (68) or (71).
* 89. Polar systems.

(
The theorems on
the classification of conies
79)
may be
regarded as completing the discussion of projective polar

is,
systems in a real plane. There

est in
however, a certain amount of
inter-
of
making the discussion of polar systems without the intervention complex elements, and basing it entirely on the most elementary
for a
S, P.
theorems about order relations. This treatment will hold good

projective space satisfying Assumptions A, E,
THEOREM
36.
the involutions
are all direct,
In any projective polar system in an ordered plane of conjugate points on the sides of a self-polar triangle or else one involution is direct and the other two opposite.
Proof.
Let
ABC be
the self-polar triangle
(fig.
68),
and
let
PP'
be
a pair of points on the side and QQ' a pair on the side CA. Let R that of AP' and be the point of intersection of the lines and AB, BQ', and R' that of CO and AB. Then AP' is the polar of P, BQ' of Q,
BC
PQ
PQ
of
of 0,
and
CO
of
R.
conjugate points on
Hence R and R' are paired in the involution AB. Let R" be the point in which P'Q' meets
If the involutions
,
on
BO
and
CA
are direct,
and P' separate
JB
and C and Q and Q' separate C and A. It follows by Theorem 19, Chap. II, that R and R" do not separate B and A. Hence by Theorems 7 and 8, Chap. II, R' is separated from R by A and .Z?, and hence the
involution on the line
AB is
if
direct.
On the
other hand,
the involutions
onBC
and CA
are not direct,
P and P'
-4
do not separate
and
and A. Hence
and
.ft"
do not,
and $ and $' do not separate C and therefore R and R' do, separate
C,
is direct.
and B. Hence again the third involution
We
direct
;
have thus shown that at least one
of
the three involutions
is
and that
if
two are
direct, so is also the third.
From
this the
statement in the theorem follows.
The reasoning above

COROLLARY
of
a,
is
valid in
any ordered projective space.
Specializing to the real space,

1.
we have
a,
projective
The involutions on the sides of polar system in a real plane are
self-polar triangle
all three elliptic, or
else
two are hyperbolic

37.
and
the third is elliptic,
on the sides of one self-polar triangle of a projective polar system in an ordered plane are direct, the involution of conjugate points on any line is direct.
THEOREM
If
the involutions of conjugate points
Proof.
Let the given self-polar triangle on the sides of which the
which has a conjugate point NonJBC. By the propwhich we are supposing proved, the involutions on the sides of the self -polar triangle, AMN, are direct and by a second application of the same proposition, the involution of conjugate points on I is
a point
osition
;
BC in
direct. Thus the proof of the theorem reduces to the proof that the involution of conjugate points on any line through A is direct. Let such a line meet in a point P', and let be the conjugate
BC
of
P' in the involution on BC,
Let
Q and
and
Q' be a conjugate pair
distinct
from
and C on the
line
A C,
same meaning
tion of
as in the proof of the last
let 0, R, R', JR" have the theorem (fig. 68). Also let
0' be the conjugate of
ABP'
on the line AP', i.e. let 0' be the intersecwith PQ. Applying Theorem 19, Chap. II, to the triangle and the lines O'R and OR', it follows that, since C and P do
AP
not separate
and
P',
and
and R' do separate
and B,
and
0'
are separated by and P'. on the line AP' is direct.
Hence the involution
of conjugate points
COROLLARY 1. If the involutions on two sides of a self-polar triangle, of a polar system in an ordered plane are opposite, then two of the involutions on the sides of any self-polar triangle are opposite and
the third is direct.
If there
Proof.
clusion of the theorem, this would,

all
were any self-polar triangle not satisfying the conby Theorem 36, be one for which
three involutions were direct.
By Theorem 37
it
would follow
that the involutions on all lines were direct, contrary to hypothesis. The propositions stated in the last two theorems and in the last corollary may evidently be condensed into the following
:
COROLLARY
2.
Any
projective polar system in
an
ordered plane
is
either such that the involution
of conjugate points on any line is direct, or such that on the sides of any self-polar triangle two of the involutions are opposite
and
the third direct.
Applying this result in a real plane, we have that every projective polar system is either such that all involutions of conjugate points or such that on the sides of any self-polar triangle two are
elliptic,
involutions are hyperbolic and the third elliptic. In the latter case and let be a self-polar triangle, being the sides upon
ABC
AB
AC
involution on
which the involutions are hyperbolic. Let the double points AB be Cl and C2 and those of the involution on
,
of the
AC be
section
and
B^.
The
polar of
,
Cl
2
is
then the line C^C.
The conic
K* through C^ C2
polar system in given involutions are involutions of conjugate points. By these conditions are sufficient to determine a polarity.
B and tangent to the line C^C at C^ has a which ABC is a self-polar triangle, and in which the
B^,
93, Vol.
I,
Hence the
given polarity
is
the polar system of
z
.
Thus we have
THEOREM
is either the
38. DEFINITION.
A protective polar system in a real plane

A
polar system of a real conic, or such that the involution of polar system of the latter conjugate points on any line is elliptic. type is said to be elliptic.
The existence
Let
of elliptic polar
systems
is
easily seen as follows:
any point not on a side of this triangle, P' the point of intersection of OA with BC, Q' the point of intersection of OB with CA and P and Q any two points separated from P
be any triangle,
}
ABC
and
Q'
by the pairs
BC
and
CA
is
respectively.
By
the theorems in
93, Vol. I, there exists a polar

is self-polar
system in which the triangle

is elliptic.
ABC
the
and the point
the pole of the line
PQ, and by
theorems in the present section this polar system
CHAPTER VI
INVERSION GEOMETRY AND RELATED TOPICS*
90. Vectors
of vectors
and complex numbers. The properties of the addition have been derived in 42 from those of the group of
If the operation of multiplication is to satisfy
translations.
the
dis-
tributive law.
= ab + ac, (b + c)
.,
.
multiplication by a vector, a, must effect a transformation on the vector field such that b -j- c is carried into the vector which is the
sum
of those to
is
which
Z>
and
are carried.
Since the group of trans-
lations
a self-conjugate subgroup of the Euclidean -group, any similarity transformation of the vector field satisfies this condition. Let us then consider the transformations effected on a vector field
by the Euclidean group.

of a translation
Any
similarity transformation is a product
by a similarity transformation leaving an arbitrary invariant. But a translation carries every vector into itself. point Hence any similarity transformation has the same effect on the field invariant. Hence of vectors as a similarity transformation leaving
the totality of transformations effected on the vector field by the Euclidean group is identical with the totality of transformations
effected
on it by the similarity transformations leaving Since no such transformation changes every vector into
of
invariant.
itself,
any
two
them
effect different transformations of the field of vectors.
Hence we have
THEOREM 1. The group of transformations effected "by the Euclidean group in a plane upon the field of vectors is isomorphic with the group
of similarity transformations leaving an arbitrary point invariant.
To obtain a
definition of miiltinlioation
we
restrict attention to the
one transformation of this group carrying the points
and
to
and
B respectively.
lielative to
DEFINITION,
an arbitrary vector OA, which
is
called
the unit vector, the product of two vectors A" (where =f= 0) and OY Y is carried by the direct similarity transis the vector OZ to which
X= 0, OX-OY
THEOREM
is
formation carrying OA to OX, and is denoted by denotes the zero vector.
OX
OY.
In case
As obvious corollaries of this definition we have the following two theorems

:
2.
The triad of points

the
OA Y
if
directly similar to
triad
OXZ
and only
if
OZ=OX-OY.
THEOREM
3.
The equation
OZ= OX-OY
is satisfied
FIG. 60
if if Dist (OX) -Dist (OY), the unit of distance being OA.
and only
&AOX+4.AOY=4.AOZ
and Dist (OZ)
Since the direct similarity transformations leaving a point invariant form a group, the operation of multiplication must be associative, i.e.
OX-(OY- OZ) = (OX- OY)

and
also
OZ,
such that there

=
which.
B,
i.e.
a unique inverse for every vector there must be a vector such, that
is
OB
for
OF
OB-OY =
The group
commutative because
center.
it
OA.
invariant
is
of direct similarity transformations leaving
consists of the rotations about
a commutative group by
(which form a& 58) combined with dilations with

is
Hence the
operation of multiplication
commutative,
i.e.
OX-OT=OY-OX.
The
fact that the
group of translations
is
self-conjugate under the
group
of displacements translates into the distributive law,
OX- (OY+ OZ} = OXRecalling the definition of a

T7WI
OY + OX- OZ.
in Chap. VI,
1
.'
number system given

^,,U,,
TITO
-rvir.fr
n-ifm-mnmrrn
r-1-.^r.rv
THEOREM 4. With respect to the operation of addition described in 42 and of multiplication defined in this section, a planar vector field is a commutative number system.
In proving this theorem we have made use of no properties of the Euclidean group except such as hold for any parabolic metric geometry for which the absolute involution is elliptic. In case the absolute involution were hyperbolic, exceptions would have to be made corresponding to properties of the minimal lines.
The
conflict
definition of multiplication of vectors as given here does not
with the notion of the
ratio of collinear vectors as developed
in Chap. III.
For the quotient of two collinear vectors is a vector collinear with the unit vector OA, and the system of vectors collinear with OA constitutes a number system isomorphic with the real
number
system. Thus,
if
we
denote the unit vector by
1,
any vector
OX collinear with it may
be denoted by
43, a; is a real number and where, according to the definition of where, according to our present definition, x denotes OX itself, Let us denote a vector OB such that the line OB is perpendicular
to the line
OA and
such that Dist (OS)
= Dist (OA},
by
i.
Then by
the definition of multiplication,
Any
is
vector collinear with

1,
is

8,
xf,
where
a?
is
vector parallel to
and by Theorem expressible uniquely in the form

al
-f
Chap. Ill, any vector whatever
bi.
The product of two vectors may be reduced by the distributive, and commutative laws as follows
:
associative,
(al
+ bi) (cl + di) = (al + bi) el + (al + bi) di = (ac bd)l + (be + ad) I
3
By comparison with
THEOREM
5.
and 14
this
shows that
is
A planar field
of vectors
i.e.
a number system isomor-
fhic with the complex number system,
the geometric
number system
in question is that by which the complex number II Supposing that the fundacorresponds to the vector al Px) this l} mental points of the scale on the complex line are
The isomorphism
4-
P P
,
means that there
a correspondence between the complex line and the Euclidean plane in which corresponds to 0, P^ to A, and every x is relative to the scale Q , l} P point whose coordinate
is
P P
a
corresponds to the point
-{-hi
of the
Euclidean plane such that
One obvious property
of this correspondence
which we
line
shall
have
real
to use later is that the points of the
complex
which have
coordinates relative to the scale

the line OA,
a,
or,
n
correspond to the points of * in other words, that the points of the chain C (P^Pf*},
j^,
P^
P a
T>
McmnmA
//)
tfiQ
points on the real line OA.

f
a method for the investigation of

+
,hose relating to n-lines
and
id
as the coordinates of the
isting
theorems obtained by
re the articles
>ciety,
by
Vol.
I, p.
F. Morley, 97; Vol. IV,
;ain
if.
an
elementary parts of the article by F. N. Cole,
line
s
and the real Euclidean
has been so denned that
iable points,
"
may
be taken
of
to X'.
The operation
ity
;ral
transformation carrying
direct similarity trans-
OQ.
starred sections in Chap. I
may
ine
last tlieorem
6.
I,
THEOREM
ean plane
is
IT
Let
may therefore be stated in the Q Q v Q x be three arbitrary

Q
,
tollowmg form
protective line
and
let
and
J\
lie,
points of a complex two arbitrary points of a Euclid-
in whose line at infinity
L an
elliptic absolute
involution
given.
There exists a one-to-one and reciprocal correspondence

corresponds
I
in
which
of
to
TT to
a point of
Q i} L distinct from Q x
to
P to
:
to Q*,,
and
every ordinary point

is
This correspondence
I
such that
i.e.
every protective transformation of
leaving
Qv
invariant,
to
every transformation of the

(1)
form
b,
x'
= ax -f
a 3=
0,
TT,
there
corresponds
direct
similarity
transformation of
and
conversely.
The question immediately

of
TT
arises,
What
group of transformations
I,
corresponds to the general protective group on of transformations

(2)
iff
i.e.
to the set
= ax -}-b
ex
-f-
The transformation
(3)
of TT corresponding to
x'
= I/a?
'
must change any point
P to
a point P' such that
= ^P r
Therefore 3, APIffi congruent to to a as axis must carry the orthogonal line reflection with P P t f If be repoint P" of the line
is
.
Hence, by Theorem
AP^P
PP
garded
as
,
variable
point
of
line
through P it follows that the correspondence between P' and P" is projective. In
corresponds to the P', and each point at infinity of the line of the points in which this line meets the
this correspondence
as center correwith circle through P t sponds to itself. Hence the correspondence between P' and P" on a given line
70
through
is
an involution, and P' and
P"
respect to the circle.
Hence
71),
if
are conjugate points with be a variable point of the

is
plane, the correspondence
between P' and
P"
an inversion. Hence
the transformation of
TT
of the orthogonal line reflection
corresponding to x' = 1/x is the product with P P as axis and the inversion L
with respect to the
circle
through
P
1
with
P
Q
as center.
Now
(1)
i.e.
any transformation
(2) is evidently (cf.
54, Vol. I) a product
of transformations of the forms (1)
and
(3).
But the transformation
to a product of a dilation
has been seen to correspond to a direct similarity transformation, and a displacement. A displacement has
been proved in Chap.

reflections
;
IV
to be a product of
and a
dilation will
now
two orthogonal line be shown to be a product of two
or four inversions
For consider a dilation

a point
to
8,
and orthogonal line reflections. A with a point as center and carrying a point B. If is not between A and B, there exists
Chap. V) a pair of points C^ C2 which separate A and B as mid-point. Let \ be the inversion with harmonically and have l as center and passing through C respect to the circle with r The
(Theorem
transformation
I
X
with
as center,
and
leaves invariant all points of the circle through effects a projectivity on each line through
at infinity. The projectivity on therefore the involution carrying each point to a conjugate point with regard to the circle through A with as
which interchanges
each line through
center. Hence From IjA = I
and the point
is
Â
is
an inversion,
I
2
follows
IJjj.
If
is
with respect to this circle. between A and B, let A be
as center. The product the point reflection with is not between A and such that (A). Hence
AA
2
AA is a dilation AA is a product of
two and
two inversions
Ij,
and
orthogonal line reflections,
A = AI I r Since A is a product of A is a product of four inversions

line
I
orthogonal line reflections.
Hence any protective transformation of a complex

under T
to
corresponds
a transformation of a real Euclidean plane tr which is a product of an even number of inversions and orthogonal line reflections.
In order to prove it that the product of two orthogonal line reflections in IT corresponds to a projectivity of I, (/3) that the product of an and an inversion P of TT corresponds to a orthogonal line reflection
of this proposition is also valid.
(a)
The converse
we need
only verify
projectivity of
I,
and
(y)
that the product of two inversions

/.
P^
of it
is
corresponds to a projectivity of
The
first of
these statements
SS "if
~J
V^V-'.m.X.lJJQ/.A.
J-iJL.njii
J^V
1J
JCJ
l^JjJLl^Jl/ja.^N
JL
JUAIN Ji
To prove
(/3)
let
us
first
consider the case where the axis of
of P. Let be one of the points iu which passes through the center 1 the axis of meets the invariant circle of P, be any point of TT, and X' (X). The considerations given above in connection with the
= AP
transformation (3) show that
OX
and hence that
type as
AP
corresponds to a transformation of
of the
same
obviously the same as PA. In case the axis of A does not pass through the center of P, let A' be an orthogonal line reflection whose axis passes through the
to (3), i.e.
is
an involution. Moreover,
AP
center of P.
Then
AP = AA'
A'P
and
PA =
PA' A' A.
The products AA' and A'A correspond to projectivities by Theorem 6, and PA' = A'P corresponds to an involution by what has just been proved. Hence AP and PA correspond to projectivities. To prove (7) let A be an orthogonal line reflection whose axis contains the centers of P x and Pg Then
.
The products
t
P A and AP 2 correspond to projectivities by (@). Hence P P2 corresponds to a projectivity. Thus we have the important result
X
THEOREM
7.
protective transformation
sponds under T to which is a product of an even number of inversions and orthogonal

line
on a complex line correa transformation of the real Euclidean plane
real reflections, and, conversely, any transformation of the Euclidean plane of this type corresponds to a, projectimty of the
line.
complex
92.
The
inversion group in the real Euclidean plane.

its
DEFINITION. The transformations of a Euclidean plane and

line
at
infinity
which
are products
of orthogonal line reflections
transformation which
and inversions are called circular transformations, and any circular is a product of an even number of inversions and orthogonal line
reflections is said to
set
be
direct.
THEOREM
8.
DEFINITION. The
of
all circular
transformations
226
involution
group.
is
INVERSION GEOMETBY
given constitute a group which
is called
[CHAP.VI
the inversion
The
set
of direct circular transformations form a subgroup

is
of
the
inversion group, which, if the JSuclidean plane

line.
real, is
isoitwrphic with the protective group of a complex
The
first
part of this theorem

is
is
an obvious consequence of the
definition,
equivalent to Theorem 7. That not all circular transformations are direct is shown by the special case of an inversion. An inversion is not a direct circular transformation,
and the second
because
it
correspond under
leaves invariant all points of a circle and hence cannot F to a projectivi'y. Combining Theorems 8 and 6
we have
CoROLLAEY. In a
real
fc
transformations leaving
invariant
Euclidean plane the group of circular is the Euclidean group, and the
lx
direct circular transformations leaving
invariant are the direct
similarity transformations.
tions
of
The isomorphism between the group of direct circular transformaand the projective group on the line may be used as a source theorems about the former. Thus the fundamental theorem of
projective geometry (Assumption P) translates into the following theorem about the real Euclidean plane:
THEOREM 9. A direct circular transformation which leaves three ordinary points, or two ordinary points and L, invariant is the identity. There exists a direct circular transformation carrying any
three distinct ordinary points A, B, C respectively into three distinct points A', B', C' respectively, or into A', B' , and L respectively.
Now
consider a circular transformation II which
is
not direct and

definition
which, leaves three distinct points A, B,
C invariant. By
n = A 2n+1 A2n
where
Aj,
(i
reflection.
Let
= 1, 2, A be
-,
-f 1) is
an orthogonal
an inversion or an orthogonal line line reflection whose axis contains
A, B, C, if these points are collinear, or an inversion with respect to the circle containing them in case they are not collinear. Then is a direct circular transformation leaving A, B, C invariant. Hence
AH
ATI
Sinf.fi
= 1.
is nf -nprinrl t.wn
f.Tiia irn-rVIioa
replaced by
The same argument applies in Hence we have I*.
case one of the points A, B,
is
THEOREM
10.
circular transformation which is not direct
and
leaves invariant three distinct ordinary points A, B, C, or two
ordinary
points A, B, and /, is an orthogonal line reflection or an inversion according as the invariant points are collinear or not.
THEOREM
11.
If
II is
a circular transformation and
sion or orthogonal line reflection,

line reflection.
UAH"
is
A an inveran inversion or orthogonal
Proof.
Let A, B,
C be
three of the invariant points of
then
HAHwhere
leaves II (A), II (B),
H (C)
invariant.
If
n^v-A,,
A
,
AB
are orthogonal line reflections or inversions, then
HAH- = A A
1
An AA
A^,
is
and
is
thus a product
of
an odd number
last
of
orthogonal line reflections

it
or inversions.
reflection or
Hence by the an inversion.

of
theorem
1
an orthogonal
line
The invariant elements

elements of
IIAII" are those to which the invariant

II.
A are
1.
carried
by
Since
HAH"
is
an inversion or an
orthogonal line reflection,
we have
any circle, into an ordinary line and on l x It and an ordinary line into a set of this
.
COROLLARY
a
circle
Any
circular transformation carries
or into the set of points on
carries the set of points

sort or into
on
ln
circle,
COROLLARY
there exists
2.
2 If C and
_ST
are
any two
circles
and
I
z
any
to
line,
a direct circular transformation carrying 2 one carrying C to the set of all points on I and lv
.
K* and
be any three points of C let A', B', C' be any z and let A', B' be any two points of I. By Theorem 9, three points of , there exist direct circular transformations II and II' such that
Proof.
Let A, B,
U(ABC)=A'B'C'
Since A', B',
2
'
and
IF
which
II carries
are not collinear, the set of points into

circle
;
<7
must be a
and since there
is is
only one circle containing
A', B', C', this circle is
z
.
Since there
;,o,
no
circle containing A',

D
',
TT/
/^,i
points on
I*
A' and B',
it
and an ordinary must be

I.
line.
Since the ordinary line contains

through, its center into Hence, by the last
An
inversion
(71)
transforms
all lines
themselves and interchanges the center with L. two corollaries, we have at once
COROLLARY
the set
3.
An
of points on l x and a
inversion carries a circle through its center into line not passing through the center.
COROLLARY 4. pair of circles which touch each other is carried by an inversion into a pair of circles which touch each other, or into a circle and a tangent line together with lx or into two parallel lines
,
and
1M .
z be two circles which touch each othei Proof. Let C and Since an inversion is a one-to-one reciprocal correspondence except z 2 for the origin and lx , if neither C nor passes through the origin,
z
they must be carried into two circles having only one point in common and which therefore touch each other. If C 2 passes through
the origin and 2 line I, while
K*
is
does not,
C2
is
carried into
and an ordinary
carried into a circle
point in common with the line pair a real point, I meets it in a single point and therefore
K* which has one and only one LL Since lm cannot meet K? in

is
and
into
tangent.
If
C*
both pass through the center of inversion, they are transformed and a pair of ordinary lines I, m. Since C 2 and 2 have only
the center of inversion in

the lines
I
and
L, can have no ordinary point in common. Hence I 90

(just before
common and
this is transformed into
and
It
are parallel.
was remarked in
the
fine print at
the end)
that the correspondence
Euclidean plane is with the exception of J^, correspond to the points of a certain Euclidean lineZ. Since J corresponds to l n the chain C (P^K) corresponds
,
between the complex line and the real such that the points of a certain chain C (l^P^),
to the line pair 11^.
Under
the projective group
on a line any two
chains are equivalent ; and under the group of direct circular transformations any circle is equivalent to any circle or any line pair 11*
(Cor. 2).
Hence we have
12. The correspondence T is such that chains in the complex line correspond to real circles or to line pairs llm) where I is
THEOREM
The theory of chains on a complex line is therefore equivalent to the theory of the real circles and lines of a Euclidean plane. In view of this equivalence we shall freely transform the terminology of tho
complex
line to
the Euclidean plane, and vice versa.
Thus
we
shall
speak of the cross ratio of four points in the Euclidean plane and of a pencils of chains in the complex line. The exercises below contain
number
of important theorems some of which can be obtained directly 71 and some of which can be proved most from the definitions in simply by translating protective theorems on the complex line into
the terminology of the Euclidean plane. DEFINITION. An imaginary circle is an imaginary conic the circular points such that
into real lines.
its
through
points
polar system transforms real
The
definition of
an inversion given in
circles.
71 applies without
change
to the case of
imaginary
On the geometry of circles in general the reader is referred to the to those by Stein er papers by Mobius in Vol. II of his collected works;
in Vol. I (especially pp.
16-83, 461-527)
of his collected
works
to
Vol. II, Chaps. II, III, of the textbook
Ex. 4
and
to
by Doehlemann referred to in. the forthcoming book by J. L. Coolidge, A Treatise on
the Circle and the Sphere, Oxford, 1916.
EXERCISES
inversion with respect to an imaginary circle is a product of inversion with respect to a real circle and a point reflection having
1.
An
an
the
same center
as the circle. 2 of a circle are 2. The inverse points on any line through the center as center. If A^ and A z are any two the pairs of an involution having inverse points, OA l OA 2 is a constant, which in case of a real circle is equal
to
(0C)
3.
2 C being a Two pairs of

,
,
point of
points
C2
A A'
and BE'
are inverse with respect
to a circle
with
as center
if
and only
ordered
BE', and
4.
(2) the
(1) triads
if
is
collinear with the pairs
A A'
and.
OAB
and OB'A' are similar, but not
directly similar.
DA,
of six hars OA, OC, AB, BC, linkage which consists of a set CD, and such that Dist ) =r jointed movably at the points 0, A, B, C, D, "Peaucellier a is called a is inversor." If Dist (OC) and rhombus, is held fixed and B varies, the locus of D is inverse to that of with
(OA
ABCD
with
B
.
respect to a circle
i;_i_
as center. If
-t-'ViT-rmrrk
be constrained, say by
dpanriViP.R
a.
i_
n ;- n l,->
7")
linfi.
On
an additional
t.lic*
subject of linkages, cf. K. Doehlemann, G eornetrische Transformationen, Vol. II, p. 90, Leipzig, 1908, and A. Emoh, Projective Geometry, 62-67,
New
York, 1905.
If
5.
B, C,
D
=
are four points of a Euclidean plane,
U
where
,
(AB, CD} =
ke { <>,
Dist (4 C)
.
Dist,
-s-
Dist(SC)
; f
(AD)
and
ft,
Dist(73D)
where a and /3 are the measxires of 4. CAD and respectively. The number k is invariant under the inversion group, and 6 under the group of direct circular transformations. The four points are on a circle or collinear
if
&CBD
=
6.
7.
0.
Constrict a point having with three given points a given cross ratio. 1 any circular transformation, the points = II"" (i= ) and 0' =!!(/) are transformed by II into are called its vanishing points. The lines through
If II is
the lines through 0'.
If
is
is
any point
of the plane,
Dist (0A")
(cf.
Dist (O'X')
a constant, called the
and X' = II (X), then power of the transformation
43).
8.
n
[I
B be two points not collinear with and let II (/I) = A', ordered point triads OAB and O'B'Af are directly similar if is direct, and similar, but not directly so, if II is not direct. 9. The equations of an inversion relative to rectangular nonhomogeneous
Let
and
(7?)
B'.
The
coordinates, having the center of inversion as origin, are

s
_ ~
x*
kx
or imaginary according as k > circle of inversion or k < 0. 10. The coordinate system for the real Euclidean plane obtained by means of the isomorphism of the Euclidean group with the projective group leaving
The
is real
a point invariant on a complex line is such that the coordinate z of any point is j: + iy, where x and ?/ are the coordinates in a system of rectangular non-
homogeneous coordinates and

condition
z
=
d'
1.
The
a b
c
points z of a circle satisfy the
at-}- I
d
fixed. If c
where
circle
is real
and variable and

line.
a, b,
c,
reduces to a
11.
The
circles orthogonal to z
ct
d are complex and

are
0, this
d
&
ft
(c
+ +
d(3)it
z2
+5+ + d+
aa
ca.
where a and
12.
f3
are real.
The
circles
through two points z v
are given by
13.
circle
with
z l as center is
z
given by
z1
ke^,
where
14.
==
<
TT
and k
is
a real constant.
centers of the circles circumscribing the four triangles formed by the sides of a complete quadrilateral are on a circle. This circle is called the
center circle of the
The
complete quadrilateral.
The
centers of the center circles

sides of a complete five-line
of the. five complete quadrilaterals
formed by the
are on a circle called the canter circle of the five-line.
Generalize this result.
the corollary of Theorem 8 the set of direct circular transformations leaving L invariant is the
93.
Generalization
by
inversion.
By
group of direct similarity transformations, and the set of all circular transformations leaving fc invariant is the Euclidean group. This is
the basis of a method of generalization "by inversion entirely analo73. gous to the generalization ~by projection employed in In case a figure which is under investigation can be transl
formed by one or more inversions into a known figure FZ) then such of the relations among the elements of F as are invariant under z
circular transformations
must hold good among the corresponding

it
elements of F^. In order to apply this method
is
necessary to
know
relations
which are
left invariant
by the
circular transformations.
The most
elementary of these are given in the last section, but perhaps the most important property of an inversion for this purpose is that of
" isogonality, or preservation of angles."
DEFINITION. If Cf and Cl are two circles having a point Q in com2 mon, and m 1 and m 2 are the tangents to (Tf and (72 respectively at Q, line pair mjn z is the ordered of the measure (according to 72)
called the
angular measure of the ordered pair of

it
circles at Q, or
simply the angle between the two circles at Q.
If C\ is
any
circle,
}
w
m
a line meeting
in a point Q,
and
m
1
the measure of the ordered line pair

>2
mm
2
is
the tangent to C\ at Q called the angle between
and
C*,
and the measure
of
mjn^
is
z
called the angle between C[

is
and
wij
mf
The measure
of a line pair
mjm
called the angle*
between
and
mf
An
THEOREM
line
and a 13. angle a between two circles or a circle a "by an inversion or or between two lines is changed into TT
*In accordance with common usage, we are here using the term "angle" to denote a number, in spite of the fact that we use it in 28 to denote a geometrical
an orthogonal
line
reflection
and
is
left
unaltered by any direct
circular transformation.
Proof. The statement with regard to direct circular transformations is an obvious consequence of the one with regard to inversions
and orthogonal
line reflections. What we have to prove is, therefore, the following: Let II be an inversion or an orthogonal line reflection, and let ^ such that II (P) = Q is an and 12 be two hues meeting in a point
ordinary point. If
l^
is
carried
by
II into a line, let this line be

is
denoted
x
by
m^
and
if
(together with fc)
carried to a circle
(7
let
denote
the tangent to (7f at Q; likewise, if lz is carried by II into a line, let and if lz (together with L) is carried to a this line be denoted by 2
circle Cl, let
m.2 denote the tangent to Cl at Q. The two ordered pairs
of lines l^.2
In case II
and m^m^ are symmetric. is an orthogonal line reflection,
m = II (IJ and ra = H
t 2
(/ 2 )
and the proposition is a direct consequence of the definition of the " term " symmetric ( 57). Suppose, then, that II is an inversion havas center. ing a point One of the lines l^ Z 2 say can be transformed into Zj,
,
itself if
and only
if Z
is
on
0.
By hypothesis 0^=P; hence

if II (Z ) x
1
1}
the line
goes
into the set of points different
from
through
is
on a circle Cl and Q. Then m z
the tangent to Cl at Q.
line
1
Any
point
meets
which through in an ordinary meets C\ in the
EIG. 71 which corresponds to under the inversion. Hence the line n z through and tangent to Cl cannot meet lz in an ordinary point, and is therefore parallel to lz Hence the line pair IJ2 is congruent to the pair l^. The line m i is the tangent to Cl at Q. Since l^ is carried to ^m 2 by the orthogonal line reflection whose axis is the perpendicular bisector of OQ, the pair
point
NIKE-POINT CIRCLE
If neither of the lines
l
l}
233
itself,
1 2
is
transformed into
neither passes
II
^
rough 0.
e
Let
denote the line OP. Then by the last paragraph

ll
z
symmetric with Im^ and

.
with Im 2
is
symmetry which carries II i lich carries IL to Im Hence LI * 1

fi
But hy Theorem 13, Chap. IV, to Im^ must be identical with that
.
with symmetric *
let us
m,m 12
As an
exercise in generalization
14.
by inversion
C| meei
prove the following

in such a
THEOREM
'.h
If three
circles Cf, Cj,

,
in a point
way
thai
o Q, R, the circle (or line) through

cles
pair of them makes an angle
and
P,
also meet by pairs in three other points
Q and
makes with each of
the other
an angle
Proof.
;h of
The
pair of circles
which meet at O obviously make the angle - at

o
the points P, Q, R. An inversion II with respect to a circle having O center must therefore change them into the sides of an equilateral triangle.
circle
.e
circumscribing this triangle makes the angle

since this circle
is
with each of the
es.
But
3 the transform of the circle
PQR
by
II,
the
iclusion of the
theorem follows.
a second application of the theory of inversion, in combination with jjective methods, we may consider the theorem of Feuerbach on the nine
.nt circle
(cf.
As
Ex.
2,
73).
The nine-point circle of a triangle touches the four inscribed circles. Proof. Let the given triangle be ABC, and let the mid-points of the pairs be A v B v Cx respectively. The nine-point circle is the circle 1, CA, AB
itaining
THEOREM 15.
Av
UN
is
V JiJKDIUIN
UJ^VJlVlJCjXJttl
[^HAP. VI
one
line,
I,
besides
A B, EC, CA, which
touches both
Kf and
/Cf.
Let A'B'C'
be the points in which I meets the sides BC, CA, AB respectively. Then AA', BB', CC' are the pairs of opposite vertices of a complete quadrilateral circumscribing both Kf and 7C|, and the diagonal triangle of this quadrilateral is a
self-polar triangle both for Kf and Kj ( 44, Vol. I). Since the side A A' of this triangle is the line of centers of Kf and Jff, the other two sides, BB' and CC", are parallel to each other and perpendicular to A A'. Let their points of
A A' be J5 and C respectively. These two points are conjugate with respect to both circles, and hence must be the limiting points of the pencil of circles containing Kf and Kj. The radical axis of the pencil of circles is the perpendicular bisector of the pair B Q CW and hence ( 40) passes through the mid-points of all the pairs BC, B'C', BC', B'C, B Q C In particular the radical axis of Kf and 7f| passes through A v the mid-point of BC.
intersection with
.
Hence there
is
a circle G'z with
as center
.
and passing through
B and C
Let r be the inversion with respect to G 2 Since this circle passes through B Q and C it is orthogonal both to Kf and Kj (Theorem 34, 71), and hence F transforms each of these circles into itself. We shall now prove that I
,
transforms
it is
Let B2 not parallel to
into the nine-point circle. be the point in which A i B l meets

I /,
/.
Since
A l Bl
it
is
parallel to
AB,
and hence
is
an ordinary point. Since A^B^ contains

is parallel
the mid-point
Al
of the pair
CB
and
to BC',
contains the mid-
point
C of the pair CC'. The involution which F effects on the line A 1 B l must have C as one of its double points and A^ as its center hence the other double point must be the point B 3 in which A^B V meets BB', because A l is the
j
mid-point of the pair But since
C B S Thus G 2
.
passes
&,
through B a
2
as well as
through
B A'C
B
i
A=B B
s
C B lt
C
and
are harmonically conjugate with respect to transforms B 2 to Bv
and
B2
Bs Hence P
.
if C is the 2 point in which A 1 C1 meets C2 to Cr Since any line whatever is transformed by T to a A v it follows that is transformed to the circle through A v B v and Cv i.e. to the nine-point circle. By Theorem 11, Cor. 4, since is tangent to Kf and Kj, the nine-point circle touches Kf and Kj. Since it has not been
In like manner
transforms
it
can be shown that
I,
circle
through
contains the centers of Kf and K\, specified which of the bisectors of 4this argument shows that the nine-point circle touches all four inscribed circles.
CAB
EXERCISES
1.
Any
three points can be carried

circles
by an inversion
carried
into three collinear
points.
2.
Two
Anv
nonintersecting
rlirp.nt
can be
by an inversion
rrf
into
concentric circles.
mrmila.r t/rpmsfrvrmatirm is
a.
-nrnrlnrvf-.
an
in-woToin-n
anrt
4.
product of two inversions
is
an involution
if
and only
if
the circles
are orthogonal. 5. Of four circles mutually perpendicular
by
pairs, three
can be
real.
6. The nine-point circle meets the circle through C having A as center : in points of the line A'B'. 7. The nine-point circle of a triangle touches the sixteen circles inscribed
to the triangle or to orthocenter.

8.
any
of the triangles
2 8
formed by pairs
a point 0,
2
,
of its vertices
with the be the
Let three
circles C*, Cf,
C meet in
and
let
Pv P P
z
,
other points of intersection of the pairs C|C3 CjCf, CfCf respectively. Tf Q 1 be any point of Cf, Q 2 the point of Cj collinear with and distinct from Q 1 and P s and (2 8 the point of Cg collinear with and distinct from Q 2 and P v then Q 3
, ,
and Q 1 are collinear. 9. The problem of Apollonius. Construct the circles touching three given circles. Cf. Pascal, Repertorium der Hoheren Mathematik, II 1, Chap. II, on this and the following exercise. 10. The problem of Malfatti. Given a triangle, determine three circles each of which is tangent to the other two and also to two sides of the triangle.
,
P2
94. Inversions in the .complex Euclidean plane.

dealt only with a real Euclidean plane.
Thus
far
we have
The
definition of
an inver-
sion given in
2
71, however, applies

;
Euclidean plane
to a circle
(7
,
i.
e.
two points A^
without change in the complex A z are inverse with respect

2
collinear
with
its center.
provided they are conjugate with respect to (7 and The transformation thus defined is obviously
one to one and reciprocal for all points of the complex projective is the plane except those on the sides of the triangle 01^, where 2 center of C , and 7X and 72 are the circular points at infinity. Any
point of
is carried to every transformed to every point of the line OJj, and every point of the line 07a is transformed to 7r In like manner 72 is transformed to every point of the line 072 and every fc is
carried to
by the inversion, and
point of L.
The
circular point
is
point of this line
is
carried to
72
DEFINITION. The sides of the triangle

lines of
07^
are called the singular
the inversion with respect to
C2
and the points on these lines
are called its singular points.
The principal properties of an inversion may be inferred from the following construction If A^ is any point not on a side of the be the points distinct from 7t and 7, and triangle 07 7 let
:
a,
B^
(fig.
73) in
which the
lines
A^
and
AJ
respectively meet
(7
Let
A. be the point of intersection of
IB
and
IB
The
points
and A,
are mutually inverse because,
are conjugate with regard to C*
by familiar theorems on and collinear with 0.

evident in the
first
conies,
they
From
line
,
this construction
it
is
place that all
points, except
/x
of the line
A^,
are transformed into points of the
A 2 fa and
vice versa.
Hence an
inversion transforms the minimal
lines through 1^ into the minimal lines through J
3,
and vice
versa.
More-
over, the correspondence between the two pencils of
minimal
if
lines is
is
such
that
B
2
a variable
point of
<7
the line I^B
corresponds to other words, the IJB. In
always
correspondence
effected
by an inversion between
the two pencils of minimal lines is a projectivity generating the invariant circle
FIG. 73
2
(7
The
definitions of circular
and
of direct circular transformations,
given in
92, apply without
change in the complex Euclidean plane.
The
result just obtained therefore implies that
any
direct circular
transformation transforms each pencil of minimal lines protectively into itself, and any nondirect circular transformation transforms each pencil of minimal lines protectively into the other.
Now suppose that A^ ing /j or Iz

.
is
a variable point on any line

l
not contain-
(4)
Since B^ and J5a are always on the conic

/K\
\
<7
i
2
,
rr/fi
i
r
7\T
L-îJ
r 2 L
iJ>
and
(6)
Hence
(7)
But corresponding
lines of these
two pencils
intersect in the vari-
594]
COMPLEX INVEESIONS
line.
237
corresponds to
or
on a
In the projectivity
(5)
the ]ine
I2
/;
in (4)
corresponds to itself; and in (6) l a corresponds to 10. Hence in (7) the line J2 corresponds to ^0, and so the circle or line
generated by (7) passes through 0. This result may be stated in a form which takes account of the
singular elements, as follows
:
Any
degenerate conic consisting of 1K
and a nonminimal line is carried by an inversion with respect to <7 a into a conic (degenerate or not) which passes through I J and 0. 2 Next suppose A^ to be a variable point on any nondegenerate conic
lt
,
through Ii and J2
(8)
In
this case
and hence by the
projectivities (5)
and
(6)
we have
Hence
only
if
is
again on a conic through 1^ and
J2 which can
,
degenerate
corresponds to itself under (9). The latter case implies, by (5) and (6), that 1^0 and IZ correspond under (8) or, in other words, that the locus of A 1 passes through 0. Hence any nondegenerate conic
through 1^ and J2 corresponds by the inversion with respect a conic through Jx and I2 which degenerates into a pair of z one of which is L, only in case passes through 0.
to
,
to
C*
lines,
This result, together with the other statement italicized above, amount to an extension of Cors. 1 and 3 of Theorem 11 to the com-
plex Euclidean plane. From our present point of view we can also establish the following theorem, which did not come out of the
reasoning in
92.
THEOREM
16.
The correspondence between two

is protective.
circles
which are
homologous under an inversion
Proof. If A^ is a variable point of one circle and A z of the other, I2A Z and hence by (5) then, in the notation above, IZB^
r\A
on -PR nio-n<-
/x-mrl'it.-irYn
^'hot'. ^.Tio
r>nTTQev\rvnrl a
238
INVERSION GEOMETRY
[CHAP.VI
The same reasoning also applies in case one or both of the conies which are the loci of A and A 2 degenerate. We thus have
I
COROLLARY.
projective correspondence
is
version between
its
any two homologous

set
lines
established by an inor between a line and
homologous
of points on
a,
circle.
The proof of Theorem 13 on the preservation of angles under a circular transformation applies without change in the complex Euclidean plane. This theorem can also be proved by the use of considerations
with regard to the circular points.
shall give the argument for the case of orthogonal circles, leaving it as an exercise for the reader to derive the proof along these lines for the general case. 7 1 that the circles through two points A It has been proved in
lt
We
Az
if
if and only are orthogonal to the circles through two points V Z the pairs AÂ^ Bfl^ and /a /2 are pairs of opposite vertices of a
B B
complete quadrilateral (cf. fig. 73). The sides 7^, 7^, 7^, I2 A y of such a quadrilateral are transformed by an inversion relative to any Hence the points A I} A Z Blf circle into four lines through Ii and 72
.
are transformed into four points A[, A! 2, B[, I>[ such that
7 72
a
are pairs of opposite vertices of a complete quadriZ A[A[, and B[B' Hence the pencils of circles through A^, A z and V BZ lateral.
respectively are transformed into two pencils such that the circles of one pencil are orthogonal to those of the other.
With
this result it is easy to prove that
Theorems 811,
13,
and
complex Euclidean plane, proper exceptions being made so as to exclude minimal lines and pairs of points on minimal lines. This is left as an exercise.
95. Correspondence between the real Euclidean plane
pencil of lines. sional
/,,
their corollaries hold in the
and a complex
The correspondence between a complex one-dimenreal
form and the points of a
Euclidean plane, together with
can be established in a particularly interesting way if the onedimensional form be taken as the pencil of lines on one of the circular
points of the line at infinity of the Euclidean plane.
line through
/15 except L, contains one and only one real point of the Euclidean plane. Let us denote by T' the correspondence by which l m corresponds to itself and the other lines through 1^ correspond each
to the real point
which
it
contains.
By
lines on
94 a
direct circular transformation transforms the pencil of projectively into itself. Hence every direct circular transr
formation corresponds under T to a projectivity of the lines on / r By Theorem 9 there is one and only one direct circular transformation carrying an ordered triad of distinct points to an ordered
triad of distinct points
and by Assumption P there is one and only one projectivity carrying an ordered triad of lines of a pencil to any ordered triad of the pencil. Hence a given projectivity of the pencil of lines on / can correspond under T' to only one direct circular transx
;
In other words, T' sets up a simple isomorphism between the projective group of a complex one-dimensional form and the group of direct circular transformations.
formation.
The correspondence between the points of a real line and the lines joining them to 1^ is evidently projective. Since the cross ratio of four points of a real line is real, so is the cross ratio of the lines joining them to /. Hence any real line together with / corresponds under T' to a chain. Since any two chains of a one-dimensional form
are projectively equivalent, and any circle of the Euclidean plane is equivalent under the inversion group to an ordinary line and l m , it follows that under F' any chain corresponds to a circle and any circle
to a chain.
The correspondence F' may be used to transfer the theory of invofrom the complex pencil of lines to the Euclidean plane. Let AA', BB', CO' be pairs of opposite vertices of a complete quadrilateral of the Euclidean plane. The pairs of lines joining these point pairs to JT are pairs of an involution. Hence X
lution
THEOREM
17.
lateral are pairs
The pairs of opposite, of an involution, i.e.
vertices
of a complete quadri-
they are pairs
of homologous
points in a direct circular transformation of period two.
rilateral constitute
In other words, the pairs of opposite vertices of a complete quadthe image under F' (and hence under F) of a quadrangular set. "While the converse of this proposition is not
the proposition can be generalized by inversion so as to give
true,
240
INVERSION GEOMETRY
[CHAP. VI
a construction for the most general quadrangular set in which no four of the six points are on the same circle or line (cf. Ex. 1, below). shall state the construction in terms of chains.*
We
C
THEOREM 1 8. Given two pairs of points A A such that no four of the Jive points are on chains C (AB'C) and C (A'BC) either meet in a or touch each other at C. In the latter case let
1
and BB' and a point the same chain. The
point D other than C D denote C. The chains
C (DAB) and C (DA'B

are pairs of
meet in a point C' such that AA', BB',
CC
an
involution.
Proof. Consider the figure in the Euclidean plane (together with I*) corresponding under F to the figure described in the theorem. If F' (D) =f= /, I" (D) can be transformed to fc by an inversion I.
:
Under IF the four chains C (AB'C), C (A'BC), C (DAB), and C (DA'ff) correspond to Euclidean lines (with l a), and hence AA', BB', CC' correspond to the vertices of a complete quadrilateral so that the theorem reduces to Theorem 17. If T (D) = 1 M the theorem reduces directly
; 1
,
to
Theorem
17.
line AA',
COROLLARY. Three pairs of points on a complex

CC', such that the chains
BB'
(A'B'C'}, C (A'BC}, C (AB'C}, C (ABC'} are distinct, are pairs of an involution if and only if the four
chains have a point in common.
EXERCISES
if
Three pairs of points of the same chain A A', BB', CC' are in involution for any point D not in the chain the chains C (DAA'}, C (DBB'}, C (Z>C*C")
1.
are in the
2.
same
pencil.
81, from the theory of involutions in a plane. BB', CC' are pairs of opposite vertices of a complete quadrilateral, the three circles having A A', BB', CC' respectively as ends of their diameters belong to the same pencil, and the radical axis of this pencil passes
Derive Ex. 15,
3. If
A A',
through the center of the

quadrilateral.
4L
circle
circumscribing the diagonal triangle of the
Construct the double points of an involution in a Euclidean plane with
ruler
and compass.
* This puts in evidence the fact that "while the geometry of real one-dimensional forms depends essentially on constructions implying the existence of two-dimen-
96.
The
real inversion plane.
In a real Euclidean plane an inver-
sion has been seen to be a one-to-one
and reciprocal transformation
except in that
to
.
it
transforms
la
to the center of inversion, and the center
An inversion, therefore, is strictly one to one if we regard it as a transformation of the set of objects composed of the points of the real Euclidean plane together with I* regarded as a single object.
with the line at infinity regarded as a single object,
inversion plane
plane.
;
DEFINITION. The set of points in a real Euclidean plane, together is called a real
fc
is
called the point at infinity of the inversion
The
is
set of points
on a real
circle, or
on a
real line
together
with
I*,,
called a circle of the inversion plane.

of
An inversion is either
or
an inversion in the sense

circle or
71 with respect to a real

reflection.
imaginary
etc.
an orthogonal line
92.
Circular transformations,
are defined as in
The
set of
theorems about the inversion plane,
which remain valid when the
figures to
which they
refer are subis
jected to every transformation of the inversion group, real inversion geometry.
called the
definition,
to
Although the point at infinity receives special mention in this from the point of view of the inversion geometry it is not be distinguished from any other point of the inversion plane. For
any point of the inversion plane can be carried to any other point of it by an inversion. In a set of assumptions for the inversion geometry as a separate science, there would be no mention of a point at infinity
just as there is
no mention
of a line or a plane at infinity in our
assumptions for projective geometry. The inversion geometry has a relation to the Euclidean geometry which is entirely analogous to the relation of the projective geometry
namely, the set of transformations of the inversion a group which leaves one point of the inversion plane invariant is parabolic metric group in the Euclidean plane obtained by omitting
to the Euclidean
;
this point
A
ple,
large class of
from the inversion plane. theorems about

of the
circles
can be stated with the

of inversion.
utmost simplicity in terms
geometry
For exam-
the propositions that three noncollinear ordinary points determine a circle and that two ordinary points determine a line combine into
the single proposition
THEOREM
19.
In
the inversion
7
77.
plane any three
distinct points are
The theorem that there

circle ('" at
on
(7
may
is one arid only one circle touching a given a given point A, and passing through a given point B not be put in the following form, which also includes the
I
not on a given line proposition that through a given point
there
is
one and but one line parallel to
I.
THEOREM 20. There is one and but one circle through a point A on 2 and a point B not on C z and having no point except A in common with C'2
a
circle C'
,
.
The theory of pencils of circles makes no special mention of is merely one the radical axis ( 71), for the radical axis (with ) circle of the pencil and is indistinguishable from the other circles.
In like manner the center of a
circle
is
not to be distinguished
from any other point; for the center is merely the inverse of l xt with respect to the circle, and the inversion group does not leave
lx
invariant.
Thus the theory of pencils of circles in the inversion geometry involves no reference to the radical axis or to the line of centers.
pencil of circles
may
be defined as follows
DEFINITION.
through two
pencil of circles is either (a) the set of all circles distinct points, or (b) the set of all circles orthogonal
to the circles of a pencil of
through a point
of a given circle
Type (a), or (c) the set of all circles 2 2 (7 and meeting (7 in no other point.
pencil of circles is said to be hyperbolic, elliptic, or parabolic, according as it is of Types (a), (b), or (c). Any point common to all circles of a pencil is called a base point of the pencil.
By comparison with the theorems in the preceding sections it is evident that the pencils of circles of these three types include all the 71 and also certain pencils of circles which pencils referred to in
from the Euclidean point of view. Thus, consider a pencil of lines through an ordinary point of a Euclidean plane. Each of these lines, with L, constitutes a degenerate circle, and the set of degenerate circles is a pencil according to the definition
are regarded as degenerate,
above. Again, a pencil of parallel lines in the Euclidean plane deter2 mines a set of circles [7T ] in the inversion plane which have in common only the one point L. By Theorem 11, Cor. 3, any inversion
with a center
transforms f^T
2 ]
into a set of circles
rZT, !
Since there
is
one and only one
circle
of the
set
[E]
through
of
every point of the Euclidean plane, [Kf]

circles of
must be a pencil
Type (c). The fundamental theorems about circular transformations may be

:
stated as follows
circular transformation is a one-to-one transformation of the inversion plane which, carries circles into circles. There is a unique direct circular transformation carrying three distinct
21.
THEOREM
points A, B,
C to three distinct points A!, B', C' respectively. circular transformation leaving three points invariant is either an inversion
through these three points or the identity.
circles in
circles,
relative to the circle
The theorems on orthogonal

corresponding propositions
reflections,
71, together with ths
on
lines,
and orthogonal
line
become:
THEOREM 22. Two circles are orthogonal if and only if one of them passes through two points which are inverse with respect to the other. COROLLARY 1. Two circles are orthogonal if and only if they belong respectively to two pencils of circles such that the limiting points of one pencil are the common points of the circles of the other pencil.
COEOLLARY
all circles
2.
through
If A I and A z are inverse with respect to a A 1 and orthogonal to C* pass through A^.
circle C*,
The correspondence F, which was established in 90, 91, between the Euclidean plane and the complex projective line, is one to one and reciprocal between the inversion plane and the complex line. Since circles and chains correspond under F, the inversion geometry
is identical
with the geometry of chains on a complex
line.
The
direct
circular transformations of the inversion plane correspond to the

iprojectivities of
the complex line. 90 that the inversion with respect to the chain C (Qo^Q*) transforms every point z x iy into the conjugate imagito any chain nary point z = x iy. Hence an inversion with regard is a transformation projectively equivalent to that by which each
It follows
from
= +
point goes to its conjugate imaginary point we make the definition

:
(cf.
78).
For
this reason
DEFINITION.
Two
noints are said to be coniuaate with respect to
It is easily seen that
any nondireot circular transformation
is
product of a particular inversion and a direct circular transformation Hence any nondirect transformation may be written iii the form
,
= 0%+ b cz + a
,
We
shall return to this subject in
99.
EXERCISES
1.
Construct a set of assumptions for the inversion geometry as a separate
science.*
2.
Work
out the theorems analogous to those of
71,
00-96 for the
parabolic metric group in a modular space. Thus obtain a modular inversion 1 geometry. The number of points in a finite inversion plane is p + 1 if the
number of points on a circle is p + 1. 3. The double points of an involution leaving a chain
invariant are inverse
>n
plane.
The more elemen-
version plane follow readily

3
r
Euclidean and projective

TT'.
e plane
By
leaving out
:ermined; and by regarding termined. Any line I of TT'

;le
of the inversion plane TT
this circle as identical with

it
taking the place of the
ations on
10.
any
circle
which
veen any two circles by an lows that the order relations

Itered
by
inversion.
Hence
3y circular transformations.
n a chain are identical with

18, 19, 21-24. eloped in that the order relations of
i
(<?!<?.) and the
circle
on
circles are unaltered
by
circular transformations,
and order
relations
is
on chains
are unaltered
by
projectivities, it follows that
such that the order relations
on any chain and the corresponding Therefore the theory of order in the inversion plane applies also to the complex line. Returning to the Euclidean plane TT', we know by 28 that the
of corresponding sets of points
circle are identical.
points not
on an ordinary
line
fall into
two
classes such that
any
I,
I.
two points of the same class are joined by a segment not meeting whereas a line joining two points of different classes always meets
By
/
64 any
circle containing
two points
of different classes
meets
in
two
points.
We thus have
,
on a
DEFINITION. The points of an inversion plane not C* fall into two classes, called the two sides of C 2 such 2 that two points on the same side of C are joined by a segment of a 2 circle which does not contain any point of C and such that any circle
THEOREM
circle
23.
containing two points on different sides of
C2
contains two points of
C2
Since order relations on circles are not altered by inversion,

follows
:
there,
COROLLARY
tlie
1.
If two points are on
points
to
which they are transformed
opposite sides of a circle C*t an inversion II are on l>y
opposite sides
ofIi(C}.
a complex line the points on one side of the chain C (<$$$<*) are evidently those whose coordinates relative to the scale Q0> Q I} Q*
On
x
are
+ iy,
where x
is
real
and y
real
and
positive,
iy.
and those on the

Hence, in general,
other side are those whose coordinates are x
COROLLARY
2.
through A, B, following two equations
The points D and D are on opposite sides of a circle if and only if y and y are of opposite sign in the
! 1
U (AB,
where
x, y, x', y'
CD)
= x + iy,
real.
B (AB,
CD')
= x' + iy',
are all
DEFINITION.
is real.
A throw T (AB, CD) is said to be neutral if B (AB, CD)

y and
of y' are
Two throws T (AB, CD) and T (A'B',
as oppositely sensed according
C'D') are similarly or the same or of opposite
signs in the equations
B (AB, CD) =1 +
st
iy
and
R (A'B', C'D') = x' -f
iy',
&.
cr!
1J 1
ViAincr TP.nl
From
this definition
it is
obvious that a direct circular transforma-
tion transforms
It is also
any non-neutral throw into a similarly sensed throw. obvious that an inversion which, reduces in the Euclidean
an orthogonal
line reflection changes non-neutral throws
plane
TT
to
into oppositely sensed throws.
Hence we have
circular transformation carries non-neutral throws into similarly sensed throws, and a, nondirect circular trans24.
THEOREM
A direct
formation carries them into oppositely sensed throws.
EXERCISES
intersecting in two distinct points separate the inversion plane into four classes of points such that two points of the same class are 2 2 , whereas joined by a segment of a circle containing no points of C and
1.
Two
circles
C2
any
2 z containing points of different classes contains points of C and 2. Two points which are inverse with respect to a circle are on opposite
circle
sides of
3.
it.
What
is
the relation between the sense of throws as defined above and
the sense of noncollinear point triads in a Euclidean plane as defined in 30? 4. In a Euclidean plane if a triangle ABC is carried to a triangle A'B'C'
by an inversion, the sense S (ABC)
is the same as or different from S (A'B'C') according as the center of the inversion is or is not interior to the circle ABC. 5. In the notation of Ex. 7, is interior to a circle C 2, then 0' 92, if
is interior to
II(C
),
and every point
interior to
C2
is
transformed by II to
a point exterior to
0'.
98. Types of circular transformations. By 5 every projectivity on a complex: line has one or two double points. On account of
the correspondence F the same result holds for the direct circular transformations of the real inversion plane.
first a transformation II having but one double In the theory of projectivities such a transformation has been called parabolic; and it has been proved that there is one and but
Let us consider
point.
which goes to A Q R (MA 0i AÂ_^ = 1. Hence A_ v A A^ are on the same chain through M. Since A_ lt A Q are transformed into A 0) Av
,
one parabolic projectivity leaving a point invariant and carrying a point A to a point A r "We have also seen that if A_ is the point
,
respectively, this chain is left invariant
by
II.
In like manner any other point B not on the chain C Q (A^M) determines a chain which is left invariant by II. These two chains
would have
set of chains
to be left invariant
through
M no
is
by
of
two
II. Thus II leaves invariant a which have a point in common,
and such that there

any point except M,
If II
one and only one chain of the
set
through
be regarded as a transformation of the inversion plane, this means that II leaves invariant each circle of a pencil of circles of the parabolic type. In the Euclidean plane e, obtained by leaving
out of the inversion plane, this pencil of circles is a system of parallel lines and II is a direct similarity transformation. Now let us regard
e from the projective point of view. The transformation II leaves all points of the line at infinity of e invariant, because it leaves each of
the circular points invariant as well as the point at infinity of the system of parallel lines. Hence II is a translation in the Euclidean
plane e. This result

follows
:
may
be expressed in terms
of the
THEOEEM
25.
Any
direct
.,./,
..,
invariant point transforms
-VIMU wooy &</<>/ y jj^, u ^^ VJ ^^ v parabolic type having this point as base point. One and only one of these pencils is such that each circle of the pencil is invariant.
Eeturning to the Euclidean plane
we have
and only one
THEOREM
26.
Any
direct similarity transformation which is not
a translation or the identity leaves invariant one

ordinary point.
Proof. Regard the Euclidean plane as obtained by omitting one direct similarity transformation point from an inversion plane. effects a transformation of the direct inversion group and leaves this
point invariant. In case it leaves only this point invariant, it has just been seen to be a translation in the Euclidean plane. If not, by
the
first
invariant point unless
paragraph of this section it has one and only one other it reduces to the identity. invariant similarity transformation leaving an ordinary point
into itself the pencil of lines through this point and the pencil of circles having this point as center. Two important special cases arise, namely, a rotation about and
must transform
as center. Moreover, since there is one and only a dilation with, invariant and carrying one direct similarity transformation leaving
as center invariant,
line
through
0.
and changes every line through into another dilation which is not a point reflection leaves
as invariant, and changes every circle with every line through center into another such circle. Hence a product of a dilation and a
rotation, neither of
which
circle
is
of period two, leaves invariant
no
line
through
and no
falls
with
is
as center.
Since either a rotation
or a dilation of period
two
a point reflection,
transformation
under one
of the three cases just
any direct circular mentioned or
else is a point reflection.
Stated in terms of the inversion plane these
results
become
(cf. fig.
56, p. 158):
direct circular transformation having two fixed THEOREM 27. points transforms into itself the pencil of circles through the fixed
points
circle
and
also the pencils of circles a~bout these points. either leaves invariant every circle
it
The trans-
formation
of the other pencil, or
of one pencil and no leaves invariant no circle of either
pencil, or it leaves invariant every circle of both pencils
and
is
of
period two.
DEFINITION.
if it
A direct circular transformation is

;
said to be parabolic
if it
;
leaves invariant only one point
to be hyperbolic
leaves in-
variant two points and all circles through these points to be elliptic if it leaves invariant two points and all circles about these points ; to be loxodromic if it leaves invariant two points and no circle through
the invariant points or about them. The theorems above are all valid for the complex line if circles be replaced by chains and direct circular transformations by projecdefinition is to be understood to apply in the same Since every nonidentical projectivity on the complex line has one or two double points, the discussion above gives the theorem
tivities.
The
fashion.
THEOEEM
a complex
elliptic
;
28.
direct circular
transformation (or a projectivity on

is
line) is either parabolic, hyperbolic, elliptic, or loxodromic.
COROLLARY.
An involution on
a complex line
is
both hyperbolic
and
and any projectivity which
both hyperbolic
and
elliptic is
an
involution.
vy
u JUJHLJ.X
EXERCISES
whose double points a x and x2 are distinct from each other and from the point P*, of a scale P P v P x and whose characteristic
1.
%
projectivity
cross ratio
7.'5,
Vol. I)
is k,
may
be written
(10)
If
one of the double points
is
Pm
and the other

x:
is
x v the projectivity
may
be written
(11)
x'
= k (x
Zj).
e ie ,
The projectivity is
loxodromic
if
hyperbolic if k is real, elliptic if k = neither of these conditions is satisfied.
where 6
is real,
and
2. The parabolic projectivities with x^ as double point in the form
may
be written.
(12)
or, in case the double point
rL- = _L_ +
at,
is
P,, in the form

x'
=x+
at.
In either case a subgroup is obtained by requiring t to be real. The locus of the points to which an arbitrary point is transformed by the transformation
of this
subgroup
is
a chain, and the
set of such chains constitutes a parabolic
pencil of chains.
3.
The
projectivities (10)
and (11) for which
is constant and t a real variable, form a group (a continuous group of one real parameter, in fact). The locus of the points to which a given point is carried by the transformations of this group or the group considered in Ex. 2 is called a path curve. In the nonparabolic cases, if a is real the path curves are chains through the double points. If a is complex and a = 1,
where a
they are chains about the double points. If a satisfies neither of these conditions, and the double points are P and P&, the path curves are the loci of x = rei9 satisfying the condition
(13)
r
(3
- ae? 9
the double points are not specialized, the path curves are projectively equivalent to the system (13). Diagrams illusof curves will be found in Klein and Fricke's the three trating types path
where a and
are real constants
if
Elliptische
4.
nates,
Modulf unktionen, Vol.
I,
Abschnitt
II.
From the Euclidean point

and the
1
of view the r
and 6 in Ex.
meeting the
3 are polar coordilines
loci (13) are logarithmic spirals
through the
orie-in at
the anffle tan- (1//3X
(A. generalization of the notion of angle
one-parameter group of Euclidean transformations

allel lines
may be a pencil of paror a pencil of concentric circles or a set of logarithmic spirals
congruent to (13).
5. A projecti vity having a finite period must be elliptic. A direct similarity transformation having a finite period must be a rotation. 6. A loxodromic projectivity is a product of an elliptic and a hyperbolic
proj ecti vity.

7.
projectivity leaving a chain invariant is either hyperbolic or elliptic.
9B. Chains and antiprojectivities. The theory of chains on a complex line has been developed in the sections above by combining the
:
general theory of one-dimensional projectivities with the Euclidean theory of circles. It is of course possible, and from some points of
view desirable, to develop the theory of chains entirely independently of the Euclidean geometry. The reader 'is referred for the outlines of
such a theory to an article by J. W. Young in the Annals of Mathematics, 2d Series, Vol. XI (1909), p. 33. Many of the properties of
may be generalized to n dimensions, an n-dimensional chain or an n-chain being denned as a real ^-dimensional space contained in an n-dimensional complex space in such a way that any three
chains
points on a line of the real space are on a line of the complex space. 6 and 70.) discussion (This is the relation between S and S' in
of the theory of these generalized chains will
by
C. Segre
and
C. Juel referred to
be found in the articles below, and also in those by
Young, Transactions of the American Mathematical Society, Vol. XI (1910), p. 280, and H. H. MacGregor, Annals of Mathematics,
J. "W.
2d
Series, Vol.
XIV (1912),
p. 1.
The transformations,
az-\- 1
(14)
cz
* a
=0
of
of the
complex
line
which were mentioned at the end
96 are
analogous to the following class of transformations of the complex

protective plane
:
(15)
=a
-4-
where x denotes the complex number conjugate to x

{
These trans-
to collinear points,* but they are not projective collineations.

x[, x' 2
If x' Q
gives the equation of a nonprojective correlation. The analogous formulas in four homogeneous variables will define nonprojective collineations and correlations in
,
be replaced by u'
z , (15) u[, u'
space.
DEFINITION. nonprojeetive collineation or one-dimensional transformation of the type (14)

projectivity.
correlation
is
or
called
an anti-
The theory
of antiprojectivities
has been studied by

1,
Mathematica, Vol. Torino Atti, Vol.
XIV
(1890), p.
(1890), pp. 592. Their role in projective geometry may be regarded as defined by the following theorem due to G. Darboux, Mathematische Annalen,
Vol.
XXV
0. Juel, Acta and more fully by C. Segre, 276, 430 and Vol. XXVI, pp. 35,
XVII (1880), p. 55. In this paper Darboux also points out the connection of the geometrical result with the functional equation,
THEOEEM
projective
projective.^
29.
Any
line
which carries harmonic
one-to-one reciprocal, .transformation of u real sets into harmonic sets is
Proof. Let II be any transformation satisfying the hypotheses of the theorem, A, B, C any three points of the line, JI (ABC)=A'B'C'> and IT the projectivity such that U'(A'B'C')-=ABC. Then ITU (ABC)
ABC.
II
If
If
1
= If7 "
we can prove
,
that II 'II
is
is
the identity,
it
will follow that
and hence that II
a projectivity.
n'll were not the identity, it would transform a point to a point Q distinct from P, while it left invariant all points of the net of
rationality
R (ABC). Let L^
L L
z,
be points of this net in the order
By Theorem 8, Chap. V, there would exist two real points S, T which harmonically separate the pairs PL^ and L^L^. The transformation n'll must carry S and T into two points harmonically separating
the pairs
* Cf.
QL and L2 L a But
.
since the latter
two
pairs separate each
28, Vol. i.
8,
t Von Staudt, Geometric der Lage (Ntirnberg, 1847), of a real line as a transformation having this property.
We
defined a projectivity are using Cremona's
252
other,
INVERSION G-EOMETBY
[CHAF.VI
by Theorem 8, Chap. V, there is no pair harmonically. Hence the assumption that II'II leads to a contradiction.
separating them both

is
not the identity
COROLLARY
space
1.
Any
collineation or correlation in
a real protective
is projective.
Proof.
Since a collineatiou transforms collinear points into col-
linear points, it transforms nets of rationality into nets of rationality
in such a
way
-ive
that the correspondence between any two homologous 33-35, Vol. I). Hence, according to the theo(cf.
projective.
sorrespondence effected by the collineation between Hence the collineation is projective.

nt proves that a correlation is projective. The reason'iange in a real projective space of n dimensions.
-"'"<
reciprocal transformation of the

\ts into points rsion group.
and
circles into
Proof.
Regard the inversion plane

TT'
;
TT,
minus a point R,
as a Euclid-
ean plane
let II
of the corollary, let
be any transformation satisfying the hypotheses = P', and let H' be an inversion carrying (-S;)
is a transformation satisfying the hypotheses of the corollary and leaving JE, invariant. Since H'll carries circles through n into circles, it effects a collineation in TT. By the first corollary this collineation is projective. Since
P' to
P*.
Then H'H
it
carries circles into circles, it is a similarity transformation.
Hence
Since
H'H
H=
a transformation, say H", of the inversion group in H'" 1 !!", is also in the inversion group.
is
TT'.
Translated into the geometry of the complex projective line the

last corollary states
:
COROLLARY
is either
3.
Any transformation which
carries chains into chains
a projectivity or an
antiprojectivity.
In the light of Corollary 2 it is clear that the whole theory of the inversion group can be developed from the definition of a circular transformation as one which carries points into points and circles
J
4-
Trt.
TV
/T"T~ *
"
EXERCISES
formulas for antiprojectivities in a modular geometry. Of. 0. Veblen, Transactions of the American Mathematical Society, Vol. VIII
1.
Derive the
(1907),
2.
p.
360.
if any of the following propositions are true? Any one-to-one and reciprocal transformation of a complex projective line which carries harmonic sets of points into harmonic sets of points is either projective or antiprojective. Any one-to-one and reciprocal transformation of a complex projective line which carries quadrangular sets of points into quadrangular
Which
sets is either projecLtve or antiprojective.
Any
collineation or correlation of
either projective or antiprojective. 3. An antiprojectivity carries four collinear points having an imaginary cross ratio into four points whose cioss ratio is the conjugate imaginary.
is
a complex projective space,
100. Tetracyclic coordinates.
The general equation
of a circle in
a Euclidean plane in Chap. IV is

(16)
TT
with respect to the coordinate system employed

2
(x
+ 7/
4- 2
ap +
2 ajj
+ a = 0.
%
DEFINITION.
degenerate
circle is either a pair of lines joining
an
ll x ,
ordinary point to the circular points at infinity or a pair of lines where 1M is the line at infinity.
Thus (16) represents a nondegenerate following condition is not satisfied

:
circle,
provided that the
(17)
=
=
/
The condition aQ
clearly
find
means
that (16) represents a degenerate
an ordinary line, unless a^= which case (16) reduces to a 3 = 0. The condition
circle consisting of
=
2
also, in
Qg\
means
in case
Q, ft
.2
^2 __ Q
aQ
=f=
that (16) represents a pair of ordinary lines

,
as are real, these two through the circular points. In case a a' 2 lines must be conjugate imaginaries. In the rest of this section the
t
,
,
a' s
( Q a^ az a s ) as homogeneous coordinates of a point in a projective space of three dimensions, S.,. For every point of S 3 except those satisfying (18), there is a unique circle or line pair lln where I is ordinary, and vice
, , ,
,
are supposed real. Let us now interpret the ordered set of numbers
between the points or S 8 not on tiie locus (lo) and the circles or the inversion plane TT obtained by adjoining l x (regarded as a point) to TT. The points of S 8 which are on the locus (18) and not on <XQ
represent pairs of conjugate imaginary lines joining ordinary points

of TT to
It and J2
respectively.
There
is
one such pair
of conjugate
imaginary lines of TT through each ordinary point of TT. The points of S 8 on the locus (18) and not on a = may therefore be regarded as
corresponding to the points of TT, with the exception of L. The only and (18) is (0, 0, 0, 1), and this point point of S 3 common to # may be taken to correspond to la Thus the points of S 8 not on (18)
represent circles of the inversion plane

represent the points of
TT.
TT,
and
the points of
S 8 on (18)
Stated without the intervention of S 8 this

,
means that the ordered
set of
numbers
(a
relation (18)
may
a^ a 2> a g ) taken homogeneously and subject to the be regarded as coordinates of the points of TT. When
,
not subject to the relation (18) they of the circles and points in TT.
may
be regarded as coordinates
c
DEFINITION. The ordered sets of four numbers (a Q , a^, az , to (18) are called tetracyclic coordinates of the points in TT.
8)
subject
The same
term
is
applied to any set of coordinates

3
(/3 Q , /3 1} /? 2 , /3 g )
such that
A=
The
circles (real or
a v aJ>
av
(*
>
! 2 > 3)
j=<>
imaginary or degenerate) represented by

are called the base or
(1, 0, 0, 0),
(0, 1, 0, 0), (0, 0, 1, 0), (0, 0, 0, 1)
fundamental
circles of the coordinate system.
A second particular choice

The points
of Sj
of tetracyclic coordinates is given below.

of all real
on (18) evidently constitute the set
points on the lines of intersection of corresponding planes of the

projective pencils
(19)
o
two
where the planes determined by the same value of or are homologous. For (18) is obtained by eliminating a- between these two equations. The lines of intersection of homologous planes are all imaginary, but
a regulus,
each contains one real point This system of lines is, by 103, Vol. I, and the set oi points ofi the lines, by 104, Vol. I, a quad-
ric surface.
whose
The locus (18) is therefore- a real quadric surface all of rulers are imaginary (cf. also f 105, Vol. I).
OJJUJ-IU.O
LU
Ct
JJCJJLUJJ.
VI
Uli UJLCO.
_L'
UJL
1/11O
JJfJJ.il/i3
Ui
WHO
JLULIO
JUJLUlJLlg
K> v
az> as) and
(ft' ft' ft' ft)
correspond to the circles given by
the equation
(\aQ + tfj
(x*
+ f) + (7U*! + /*ft) x + (\a + /*,) y + (\a + /*) = 0,

z
a
which represents a pencil

in case the latter are real.
of circles, together
with
its
limiting points
Any collineation F of S 8 which carries the quadric (18) into itself must correspond to a transformation F of TT which carries points into points, circles into circles, and pencils of circles into pencils of circles. F therefore has the property that if a point P of TT is on a circle (7 2 2 of 7?,then F(P) is on F(<7 ). By Theorem 29, Cor. 2, F is a circular
transformation.
Conversely, any circular transformation of
TT
carries
points to points, circles to circles, and pencils of circles to pencils of circles, and therefore corresponds to a collineation of S 8 which carries
projective.
the quadric into itself. By In other words,
Theorem
29, Cor. 1, this collineation is
THEOREM
projective
30.
The
real inversion geometry is equivalent
to
the
geometry of the quadric (18). COROLLARY. The projective geometry of the real quadric (18)
is
equivalent to the complex projective geometry of
a one-dimensional
form.
one-to-one correspondence between a complex line and the real Let I be any complex quadric (18) may also be set up as follows line in the regulus conjugate to that composed of the lines (19).
:
Each of tnese and one point,
lines contains
Q, of
I.
The correspondence required
one real point, P, of the quadric (18) is that in whic
Q
a
corresponds to P.
By
circle,
properly choosing the constants which enter in the equation o. we may set up the correspondence between the circles of the
inversion plane and the points of an S B in such a way that the equation of the quadric surface corresponding to the points of the inversion
plane has a particularly simple form. may be written

(20)
The equation
of a circle in TT
The
points (|
^, | 2
3)
which correspond
to points of the inversioD
plane
now
satisfy the equation
(21)
#=
circles
+#+&
(1, 0, 0, 0), (0, 1, 0, 0),
and the
corresponding to the four points

(0, 0, 0, 1)
(0, 0, 1, 0),
and
imaginary. The a 2 <x s } by the equations

,
are mutually orthogonal, one of them being are connected with (# coordinates ( x 1? | 2 a)
,
a o=
+i>
a i=
2>
a2
3>
=fo-i'
which represent a
quadric (21).
If
collineation carrying the quadric (IS) into the
^/ô, ^2/^0'
3/^0
are regarded as
nonhomogeneous coordinates with
respect to a properly chosen frame of reference in a Euclidean space of three dimensions (cf. Chap. VII), (21) is the equation of a sphere. Hence the real inversion geometry is equivalent to the projectiA e geometry of a sphere. The latter equivalence may be established very neatly, with the aid of
r
theorems of Euclidean three-dimensional geometry, by the method of stereographic projection. This discussion, would naturally come as an exercise in the next chapter. It is to be found in books on function theory. On the
whole subject of inversion geometry from this point of view, compare Bocher, Reihenentwickelungen der Potentialtheorie (Leipzig, 1894), Chap. II.
DEFINITION.
A. circle
C*
is
linearly dependent on
two
circles
Cf
and
2 (Tj
if
(7
and only
2
if it is
in the pencil determined by C\ and C$.
A
C
circle
is
of
circles Cf,
if it is
member
some
finite set of circles
C; +1
,
C% if and only C* +k such that

,
+i is linearly
dependent on two
is
of Cf,
if
+i _ 1
A seb of n circles
dependent on the
linearly independent
no one
of
k). (i= 1, 2, them is linearly
rest.
The
set of all circles linearly

is
dependent on
three linearly independent circles
called a bundle.
EXERCISES
tetracyclic coordinates of a point are proportions,! to the powers of the point with respect to four fixed circles. If the four circles are mutually orthogonal, the identity which they satisfy reduces to (21).
1. 2.
The
A.
homogeneous equation of the
first
degree in tetracyclic coordinates
represents a circle. 3. What kind of coordinates are obtained by taking as the base (a) two orthogonal circles and the two points in which they meet? (b) four points? 4. Two points of S, correspond to orthoeronal circles if and onlv if thev are
is
5. What set of circles corresponds to the comes in "which the quadric (L>1 j met by the planes of a self-polar tetrahedron ? 6. The direct circular transformations of TT correspond to collincations of
S 3 which leave each imaginary regains of ('21) invariant, while the others correspond to colliueations interchanging the two reguli. The direct circular transformations of 7? correspond to direct collineations of S 3 in the sense of
31,
Chap.
II.
7.
8.
The The
to a
circles of a circles
lints
common to two
-
bundle correspond to the points of a plane of S 3 bundles constitute a pencil and hence
corre-
spond
circles
of S 3
Determine the projectively distinct types of pencils
of
on
this basis.
c2
dependent on four linearly independent circles. which has the same power, 10. For any bundle of circles there is a point with respect to every circle of the bundle. The radical axes of all pairs
9. All circles are linearly
of circles in the
0, the radical
11.
of all
bundle pass through 0. In case there is more than one point axes of all pairs of circles of the bundle coincide. bundle of circles may consist of all circles through a point (the set lines in a Euclidean plane is a special case of this). In every other case
there
circle
is a nondegenerate circle orthogonal to all circles of the bundle. This has the point (Ex. 10) as center and consists of the points C such that Dist (OC ) = c. It is real if and only if c is real. In case c is imaginary let 2 real circle be the C consisting of points C" such that Dist (0C") =c; any
1
circle of
the bundle meets C'2 in the ends of a diameter.
101. Involutoric collineations.
In view
of the
isomorphism between
the real inversion group and the protective group of the real quadric the group of a general quadric will he (21), a further consideration of
found apropos. In this connection
we need
.
to define certain particular
types of involutoric collineatious in any projective space. are all hased on Assumptions A, E, P, II Q
It is
The theorems
proved in
co,
not on
there exists a
provided that 0, on co. It follows
I, that if co is any plane and any point homology carrying any point P to a point P', P, P' are distinct and collinear and P and P' are not
29, Vol.
by the
constructions given in that place that
if
one
with regard to point P is transformed into its harmonic conjugate meets co, every point is transformed and the point in which the line in this way. It is also obvious that a homology is of period two if and
OP
only
if it is
of this type.
Hence
vre
makp
the following definition:
DEFINITION.
if
homology
and
of a three-space is said to
he harmonic
is
and only
if it is
of period two.
is
harmonic homology
also called
is
a point-plane reflection
denoted by {Oca} or {coO}. where
258
DEFINITION. If
,
INVERSION GEOMETRY
I
[CHAP.VI
two nouintersecting lines of a projective space S 3 the transformation of S 3 leaving each point of I and V invariant, and carrying any other point P to the point P' such that the line PP' meets I and I' in two points harmonically conjugate with regard
/'
and
are
to
P and P', is
THEOREM
called a skew involution or a line reflection in

I
and
I'.
It is
denoted by {Z/'},and
31.
and
I'
are called its axes or directrices.
line reflection {II'} is
a product of two point-plane

distinct points of
I,
re/lections {Oca}
a)
{PTT}, where
and P are any two

TT is
is the
plane on
V.
P and
and
the
I,
plane on
let
and
V.
Proof. Consider any plane through

it
and
L be the point in which
harmonic homologies and whose axes are PL and OL respectively. The product is therefore the harmonic homology whose center is L and axis 1. Hence the product {Oao} {Ptr} satisfies the definition of a line reflection whose axes are I and I'.
meets
effect
In this plane {Oca} and {Pir}
whose centers are
and
P respectively
COROLLARY. A.
two,
line,
reflection is
a projective collineation of period
and any
projective collineation of period two leaving invariant

is
the points
of two skew lines
line reflection.
EXERCISES
1.
2.
A projective collineation of period two in a plane is a harmonic homology. A projective collineation of period two in a three-space is a point-plane
Let A, B, C,
reflection or a line reflection.

3.
be the vertices of a tetrahedron and
a, ft, y,
8 the respec-
tively opposite faces.
as products of the three harmonic homologies {Aa}, {B{3}, {Cy} constitute a commutative group of order 8 consisting of four point-plane reflections, three line reflections, and the
If the transformations other
The transformations obtainable
than the identity be denoted by 0, 1, 2, the multiplication table may be indicated by the modular plane given by the table (1) on p. 3, Vol. I, the rule being that the product of any two transformations corresponding to points i, j of the modular plane is the
identity.
3, 4, 5, 6,
one which corresponds to the third point on the line joining i and j. 4. Generalize the last exercise to n dimensions. The group of involutoric
transformations carrying n 4- 1 independent points into themselves is commutative, and such that its multiplication table may be represented by means of
a
finite projective
line.
TYroifictivitv
space of n
1 dimensions in
which there are three points
on each 5. A
of a cnmnlex:
linp. aunli
that for one nointi
whi^Ti is
characteristic cross ratio of
is
an nth root
of unity
in case n
3,
this
cross ratio cross ratio
is
group,
6.
is
said to be equianharmonic, and a set of four points having this said to be equianharmonic. As a transformation of the inversion equivalent to a rotation of period n.
is
A planar projective
set of transforms of
any point
collineation of period n(n>2) is of Type I and the is on a conic, or else the collineation is a
;
homology. In the first case, it is protectively equivalent to a rotation in the second case, to a dilation (in general, imaginary). Consider the analogous problem in three dimensions. (For references on this and the last exercise cf. Encyclopedic des Sc. Math. Ill 8, 14. The statements in the Encyclope'die on the planar case are not strictly correct, since they do not sufficiently take
the existence of homologies of
finite perinr) into account. ">
'
102. The projective group c* ~ in 104, Vol. I, a quadric TL.O.J u^

intersection of the lines of
^&M
**.w.
~~
___,.....
two conjugate reguli. These two reguli may be improper in the sense of Chap. IX, Vol. I, and in the following theorems improper elements are supposed adjoined when needed for
the constructions employed.
DEFINITION.
If there are proper lines
it is
said to be ruled, otherwise
on a quadric, the quadric said to be unruled.
is
THEOREM 32. harmonic homology whose center is the pole of its z plane of fixed points with regard to a quadric surface Q transforms z Q* into itself in such a way that the two lines of Q through any fixed
point are interchanged. be a point not on Proof. Let
Z
2 Q and
,
<a
its
of Q* meets
I'
CD
in a unique point
(cf.
JT.
I'
The plane 01
polar plane. Any line contains one other
line
of
<?
and
104, Vol.
I
I)
ing that
to
I
to a point
of
other than
K must
passes through K. Any line joinmeet V in a point L' such
and L' are harmonically conjugate (104, Vol. I) with regard and the point in which OL meets to. Hence {Oca} interchanges and V. From this result the theorem follows at once.
Comparing Theorems 31 and 32, we have

COROLLARY.
each regulus on
A
Q
line reflection {ab} such that

z
a and
are polar with
respect to a qiiadric Q* transforms Q* into itself in such

is
a way that
transformed into
itself.
THEOREM
on
the
33.
A
fff
projective collineation of
three points of the quadric invariant,
same
n.'ntJ.
ruler, is either the

fi.!r.p.d
a quadric which leaves no two of the three points "being identity or a harmonic homology whose
nnla.r with,
re.s'np.r.t t,n
f~h.t>.
f.e.in.t.pv
inl.n.'n.p.
nni<n.tit are.
M/.n.rJ.wis*.
Proof.
them by
line of
Denote the three points by A, B, C, the plane containing and the pole of &> by O. Since no two of A, J3, C are on a 2 (/, o> contains no line of Q' and hence is not on 0. Since three
o>,
points of the conic in
which w meets the quadric are

is also 0.
invariant, all
collineation
such points are invariant, as

is
Hence the given
either the identity or a homology.
In the latter case
it
must be
harmonic homology, since any two points of the quadric collinear with are harmonically conjugate with respect to O and the point in which
the line joining
them meets
&>.
There exists one and only one projective collineation transforming each line of a re/julus into itself and effecting a given projectivity on one of iJicse lines. Such a collineation is a product of
34.
THEOREM
two line reflections whose axes are lines of the conjugate regulus. Proof. Let -R* be a regulus and li'l the conjugate regulus. projec-
a product of two involutions, and m g are say I and I'. Let {m^nj- be a line reflection such that l lines of R% through the double points of I, and let {m(m^} be a line
tivity
line,
I,
on a
of
J^
is
by
78, Vol.
I,
reflection
points of
I'.
such that m{ and m' are lines of 2t% through the double z The product of {m'jnQ and {m^m^} effects the given
projectivity
on
and transforms each
line of ft? into itself.
Conversely, any projectivity F leaving all lines of JRf invariant effects a projectivity on I which is a product of two involutions I
and F. The
line reflections
defined z } being {m^m^} and {m[in'

'
as before,
{wX>
leaves
all
m mJ
i
T" 1
points of
invariant and hence leaves all lines of
B?
as well
as all lines of
R%
invariant.
Hence
and
{m(m'. } \ i .zj i
{mm
i
!
} 2J
T*~
1, >
{m(m(}
{w^nj
= F.
lines of a regulus by the projeclive collineiUions transforming the regulus into itself is simply isomorphic with the projective group of a line. DEFINITION. collineation of a quadric which carries each regulus
effected
COROLLARY. The yroiip of permutations of the
on the quadric into
itself is said to
be
direct.
direct collineation of a z quadric surface Q carrying an ordered triad of points of Of, no two 2 of which are on a line of Q to an ordered triad of points of Q* no two
THEOREM
35. There is one
and but one
102]
GttOUP
Let
q,
OJf
A QUADEIC
261
Proof.
a, b,
c,
p,
be the given ordered triads of points, let r be the lines of one regulus through the points A, B, C.
ABC and PQR

and
P,
Q,
respectively,
let
a', I
1 ,
c',
p', q
r'
lines of the conjugate regulus
through the same points.
respectively be the By the last
is a projeetive collineation F carrying a, b, c to p, q, r respectively while leaving all lines of the conjugate regulus invariant, and also a projeetive collineation F' carrying a'b'c' to p'q'r' respec-
theorem there
tively while leaving all of the lines a,

of
b, c,
p,
q,
r invariant.
The product
is only one direct collineation having this effect is a corollary of Theorem 33. Let R% be the regulus containing the lines a, b, c, and R% the regulus
r containing a', b', c'. The two collineations F and T which have been used in the proof above are commutative as transformations of R?
and F' carries A, B,
to
P, Q,
R respectively.
That there
because F' leaves all lines of

transformations of
R*
invariant,
R$ because F leaves
FF'
all lines of
and are commutative as R% invariant. Hence
rr'=FT.
By Theorem
where
I,
34,
= {lm}
{rs}
{I'm'}
{r's'},
2
.
m,
r, s
are lines of R?,
and
V, m', r', s' are lines of JB 2
The
collineations {rs}
and
{I'm'} are
commutative
for the
same
reason, that
F and
F' are commutative.
Hence
1
FF' = {lm}
The
{I'm }
{rs}
{r's'}.
pairs lm and I'm' are two pairs of opposite edges of a tetrahedron the other two edges of which may be denoted by a and b. The product {lm} {I'm'} leaves each point of a and b invariant and is involutoric on each of the lines I, I', m, m'. Hence
{lm}- {l'm'}
= {ab}.
R*
,
The
lines
a and b are polar with respect to
because one of them
is
to the point mm' and the other the line the line joining the point of intersection of the plane IV with the plane mm' (cf. 104, Yol. I).
In like manner
{pq}
{p
= {cd}
R.
where
and d are polar with respect to

36.
Hence we have
THEOREM
surface
is
Any
direct projeetive collineation
of a quadric
expressible in the
form
quadric
Since any line reflection whose axes are polar with, respect to a is a product of two harmonic homologies whose centers
(cf.
are polar to their planes of fixed points
Theorem
31), the last
theorem implies
COROLLARY 1. Any direct protective collineation of a quadric is a product of four harmonic homologies whose centers are polar to their
respective
is
planes of fixed points. 2. Any nondirect protective collineation of a quadric a product of an odd number of harmonic homologies who$e centers
COROLLARY
are polar to their respective planes of fixed points.

Proof.
If a projective collineation
interchanges the two reguli,
and
of
harmonic homology of the sort described in the statement the corollary, then FA=Ais a projective collineation leaving each
a
A is
By Cor. 1, A is a product of an even number of harmonic homologies of the required sort, and hence F=AA is a in odd number.
regulus invariant.
103. Real quadrics.
The isomorphism between the
real inversion
group and the projective collineation group of the real quadric (or sphere) (21) may now be studied more in detail. Since a circular
invariant
transformation leaving three given points of the inversion plane TT is the identity or an inversion (Theorem 21), and since a
collineation of
S 3 leaving three points of the quadric (21) invariant the identity or a harmonic homology whose center is polar to its plane of fixed points, it follows that inversions in 7? correspond to homologies of S a Hence the direct circular transformations of TT coris
.
respond to the direct collineations of S s transforming (21) into itself. An involution in TT is a product of two inversions whose invariant
circles intersect circles intersect
and are perpendicular. To say that the invariant and are perpendicular is to say that they intersect in
such a way that one of the circles is transformed into itself by the inversion with respect to the other. Now suppose that {Oa>} and {PTT} are the harmonic homologies corresponding to the two inversions. If the points of the quadric on the plane o> are to be transformed among themselves by {Pir} &> must pass through P. In like
,
manner
TT
must pass through
0.
Hence
where
is
the line OP,
I'
the line
OJTT,
and the
lines
and
I'
are
polar with respect to the quadric. Hence the involutions in the group of direct circular transformations correspond to the line reflections
whose axes are polar with respect to (21). Thus the theorem that any direct circular transformation of TT is a product of two involutions is equivalent to Theorem 36 applied to the quadric (21). Since an involution in TT always has two double
points,
we have
the additional information, not contained in
102,
that every line reflection transforming the quadric (21) into itself has two and only two fixed points on the quadric. The line joining these
two points is obviously one of the axes of the line reflection. Hence the line reflection has two real axes one of which meets the quadric (2 1) and the other of which does not.
These remarks are enough to show how the real inversion geometry can be made effective in obtaining the theory of the real quadric (21),
shall now show that any real nonruled quadric is projectively equivalent to the quadric (21), from which it follows that the real inversion geometry is equivalent to the projective geometry of any
real
We
nonruled quadric. nonruled quadric
is
obviously nondegenerate.
In the complex
space any two nondegenerate quadrics are projectively equivalent, because any two reguli are projectively equivalent. Since (18) repreit therefore follows that every nondegenerate quadric be represented by an equation of the second degree. Now let Q z be any quadric whose polar system transforms real points into real planes, and let the frame of reference be chosen so
sents a quadric,
may
that
(1, 0, 0, 0), (0, 1, 0, 0), (0, 0, 1, 0),
real self-polar tetrahedron.
and (0, 0, 0, 1) are vertices The plane section by the plane
of a
=0
must be a conic whose equation is of the form a ! Xl + a2 X2 + a sXS =

>
and similar remarks can be made about the sections by the planes a? = 0, x = 0, and # = 0. From this it follows that Q 2 has the equation 1 2 8
(22)
aox z
,
+ atf + a^ + a x = 0,
2
s
where a
av
2
a 3 are
real.
The
projective collineation
3
(23) x'
= Vojav
Q
%[
= VT^X, ^ = V~^T*v
*
= >/KK
transforms
into a quadric having one of the following equations

SB** j.
+x2+
'
x'
0.
Any
one
of
lent protectively to
(24) (25) (26)
It is also
the eight quadrics thus represented is obviously equivaone of the following three:
x* + x* + xl + xl = 0, - xl + xl + x* + * = 0,
2 a
-xS-x* + x* + x*=0.
obvious that (24)
is
is imaginary, that (26) has real rulers, equivalent to (21).
and that (25)
EXERCISES
Determine the types of collineations transforming into itself (1) a real unruled quadric, (2) a real ruled quadric, (3) an imaginary quadric having a
1.
real polar system.

2. Discuss
the projective groups of the three types of quadrics enumerated
in the last exercise.
104. The complex inversion plane. projective plane may be obtained from a Euclidean plane (of. Introduction, Vol. I) by adjoining ideal points and an ideal line in such a way as to make it possible to
regard every collmeation as a one-to-one reciprocal transformation of all points in the plane. In like manner the real inversion plane has
ideal point
been obtained from the real Euclidean plane by adjoining a single which serves as the correspondent of the center of each
Similar considerations will now be adduced showing that an inversion in the complex plane may be rendered one to one and reciprocal by introducing two intersecting ideal lines.
inversion.
to
In the complex projective plane an inversion has been seen ( 94) be a one-to-one reciprocal transformation of all points not on the
sides of the singular triangle
01^, and
to effect a projective transfor-
mation interchanging the pencil of lines on Jx with the pencil of lines on I2 In this projectivity the line J^ is homologous both with 0/x
.
and with OI2
In the Euclidean plane obtained by omitting the line JjJ2 from the projective plane, it follows that the inversion is one to one and reciplines, p Q and through 0. correspondence between the set of minimal lines [p] parallel with and distinct from p Q and the set of
rocal except for points

it
on the two minimal
Moreover,
effects a projective
minimal lines
("ml parallel
with and distinct from
n.
The correspondence between any line p and the homologous line is incomplete because there is no point on p corresponding to the intersection of m with p and no point on m corresponding to the Q
intersection of
p with Q This correspondence, however, may be completely one to one and reciprocal by introducing an ideal as the correspondent of the point pm and an ideal x on Q Pa, on p as the correspondent of the point mp Q In order to
.
made
point
point
treat
minimal lines symmetrically, ideal points Pj, and Ml must be introduced on p and m Q respectively, as mutually correspondis introduced as the ing points. Also one other ideal point
all
the
ro
p together with its ideal J point 2 J, is transformed into a set of points consisting of #, Ml, and all the points This set x
correspondent of 0. According to these conventions the line
of points is therefore called
an ideal line
mm
In like
Q
manner the
with
its
line
together
ideal point
M
,
;
is
transformed
into
a set
of
J^',
points consisting of
Om
and
all
the points
this set of points is
and called an
P m
FIG. 74
The Euclidean plane with the lines p a and m*, adjoined is called an inversion plane. Or to state the definition formally and without reference to a particideal line p M
.
ular inversion
Given a complex Euclidean plane TT and in it two minimal lines [p] and [m]. By a complex inversion plane TT is meant the set of all points of TT (referred to as ordinary points) together with a set of elements called ideal points of which there is
DEFINITION.
pencils of
M for each m, x for each^, and one, denoted by by P and w's determining distinct ideal points, and also one other ideal point which shall be denoted by #,. By a minimal line of TT is meant (1) the set of points on a p together with the corresponding Pn, or (2) the set of points on an m together with the cor-
one, denoted
distinct p's
responding
set of all
Q-nrl
x>
or (3) the set of all
Px
's
together with <X, or

lines of
(4\
a.rft
(4)
the
(1)
Mjs
together with
nvrlnm.n.'r'ti
Ox
The minimal
(?>\ fl.nfl
Types
O.\ BT-P
p.nllp.rl
flTirl
thft linP.S
ra.llfid
ideal.
minimal line
;
or (3)
mx
of Type (1) or (4) will be denoted by p, ot lype (2) by m, the minimal lines of Types (3) and (4) are denoted by and p^ respectively.
This definition
p and on
is evidently such that each point of TT is on a unique a unique m. DEFINITION. By an inversion 1 of TT is meant a transformation
denned as follows by an inversion

lar lines of I, I interchanges
I of TT
p Q with
m, m
taining a p with the
m containing the m
p and ra Q are the singup x and each p conto which p is transformed by I.

:
If
with
A point of
points of
circle of
TT
to the point
which is the which is the

invariant
intersection of a
and an
ra is transformed
I (m).
intersection of I (p)
and
The
set of
TT left
TT. pair of called a degenerate circle of TT. By reference to 94 it is evident that every circle of
by an inversion is called a nondegenerate minimal lines, one a p and the other an m, is

rr is
a subset
of the points
on a
circle of
TT.
it
The complex inversion plane is perhaps best understood by setting in correspondence with a quadric surface, the lines of one regulus on the quadric being homologous with [p] and those of the other with
[H], This correspondence may be studied by means of tetracyclic coordinates as in 100, but it can also be set up by means of a
geometric construction as follows: Eegard the complex Euclidean plane
TT
with which we started as
immersed in a complex Euclidean

such that OJT
is
space.
Let
be a quadric surface
(fig.
a line of one ruling and
OIZ
of the other
74).
Through 7X and
there are two other lines of the two rulings which intersect in a point 0,. Any point of the Euclidean plane is joined z to Oa by a line which meets the quadric Q in a unique point Q other
J2
than
Oco
the lines
and, conversely, any point of Q which is not on either of Ox 1^ or O m lz is joined to Ox by a line which meets the
is
Euclidean plane in a point P. Thus there
a correspondence
between the Euclidean plane and the points of Q* not on 0*1^ or This correspondence is such that every minimal line in TT of 0,0/2the pencil 011 1^ corresponds to a line of the quadric which is in the same ruling with 01^ and every line of TT of the pencil on J2 corresponds to a line of the quadric which
is
in the
same ruling with 0/2

TT
From
this
it
is
evident that
if
ideal elements are adjoined to
as
explained above, the ideal points can be regarded as corresponding to
the points of the lines

reciprocal correspondence
X I^
and
X IZ so that there is a one-to-one

TT
between
and Q 2
any nondegenerate circle of TT is a conic through 1^ and /2 This is projected from M by a cone of lines having in common with 2 It follows that the cone and Q have X I2 (f the two lines O a,Il and
.
Now
also a conic section in
common. For
let
Q I} Qz Q a
,
be three
;
of
the
common
QiQ$i
are not on the lines 0^1^ and O m lz 2 meets the cone in a conic X* and Q 2 in a conic
points
which
the plane
These two
conies have also in
0,/j and
O M /2
(if
the points in which they meet the lines these points coincide, X? and If 2 have a common tan-
common
gent at this point), and hence because a nondegenerate cone through

.
K =K
2 2
.
The
conic K* is nondegenerate, Ox can have no other line than
Hence every nondegenerate circle a section of Q* by a nontangent plane. 2 is any noudegenerate conic section which is a Conversely, if 2 plane section of Q , it is projected from 0, by a cone two of whose
0*1^ and
X
IZ
in
common with Q
under
of TT corresponds
T to
lines are 0^1^ and degenerate circle of
O x J2 Hence
.
corresponds under
2
to a non-
TT.
An
.
inversion in
TT
minimal
line of the pencil [p] into that
C 2 Let K 2 be the conic section on Q 2 corresponding under T to C" The inversion corresponds under T to a transformation of Q 2 by which
2
with respect to a circle <7 transforms every one of [m] which meets it on
.
every line of one regulus is transformed into the line of the other 2 This is the transformation regulus which meets it in a point of (Theorem 32) effected by a harmonic homology whose plane of fixed
X
is
points contains
respect to
2
.
and whose center Hence every inversion
K
2
in
the polar to this plane with TT corresponds under T to a
collmeation of
effected
by
a harmonic
homology whose center and
plane of fixed points are polar with regard to

in
2
.
2 such collineation of Q evidently corresponds under
Conversely, every T to an inversion
TT.
Hence (Theorem
36, Cors. 1
and
2) the inversion
group in
IT is
isomorphic under T with the group of projective collineations of Q*, and the direct circular transformations of TT correspond to the projec2 tive collineations of Q which carry each regulus into itself.
EXERCISE
Develop the theory of the modular inversion plane, using improper elements
the two variables x and y are thought of as completely independent of each other. The domain of each is the set of all complex numbers,
including oo. This domain is therefore equivalent to the complex line or to the real inversion plane. Thus the domain of x may be taken
to be a real unruled quadric (in particular, a sphere)
of
and the domain
y another real unruled quadric. Or the pair of values (x, y} may be regarded as an ordered pair of points on the same real unruled quadric.
Now consider a regains in the complex projective space and, adopting the notation of the last section (fig. 74), let a scale be established is the zero in each scale. Let x be on the lines ~v. and Q so that
of
anv point on p Q and y of any point on

\
Jn
Then a
unique point on the quadric, i.e. the m through the point with x as its
wx^ u,.^ j, u^w^gh the point with y as its coordinate. Conversely, the same construction determines a pair of numbers (x, y} for each point of the quadric.
WWWOLÔJWUO,
^^
DEFINITION.
The
set of all
ordered pairs
called
complex numbers, including

plex plane of analysis.
oo, is
the plane of the theory of functions of
(x, y} where x and y are a complex function plane, or complex variables, or the com(x, y]
The ordered pairs
are called points.
oo or y oo is said to be ideal or at infinity, and point for which x all other points are called ordinary. The points at infinity of the function plane can be represented
Any
conveniently by replacing a? by a pair of homogeneous coordinates x9 x^ such that xJxQ x, and y by a pair (y y^ such that yjy - y.
,
Thus the points
of
the function plane are represented by
K'^; y>yd>
and the ideal points
a.re
those satisfying the condition
The set of ordinary points of the function plane obviously forms a Euclidean plane in which a line is the locus of an equation of the form
ax -f- ly -f-
0.
This
is
equivalent in homogeneous coordinates to
2 7) (
^+
fyô + cxy
0,
an equation which is linear both in the pair of variables x x and in the pair y yv The most general equation which is linear
,
,
in both pairs is
(
28 )
ax y
+ #% + 7^0 + î =
This reduces to (27) if the condition be imposed that the locus shall contain the point (<x>, oo) which in homogeneous coordinates is
(0, 1;
0, 1).
DEFINITION.
is
The
called a circle (or a bilinear curve), is called a line.
set of points of the function plane satisfying (28) and any circle of the form (27)
The group of transformations which is indicated as most important by problems of elementary function theory has the equations
y'
or,
in
homogeneous coordinates,
This group of transformations clearly transforms circles into circles.
The subgroup obtained by imposing the
conditions,
transforms lines into lines because
it
leaves
(oo, oo)
invariant.
Returning to the interpretation of the coordinates x and y on a quadric, it is clear (cf. 102) that every transformation (29) represents
a direct collineation of the quadric, the formula in x determining the transformation of one regulus and the formula in y the transformation of the conjugate regulus. Hence the fundamental group of the
function plane
is
isomorphic with
Ae
group
of
direct projective
270
INVERSION GEOMETRY
[CHAP. vi
may
points of a regulus be connected with the three-dimensional coordinates ( %v a 8 ) by means of the following equations
, ,
:
The parameters x and y which determine the
where
1.
For the
set of all points
1}
2,
f 8 ) given by these
equations are the points on the quadric,

(21)
#=# + # +
plane section of this quadric
>
,'.
Any
<>'
i>
among
given by a linear equation in which by (31) reduces to a relation of the form (28) z' 8 the parameters a? o^; y y r Hence the circles of the funcis
, ,
tion plane correspond to the plane sections of the quadric (21). In view of the relation already established between the groups it
follows that the geometry of a quadric in a complex projective space is identical with that of a complex function plane. In view 104 both these geometries are identical with the complex of
inversion geometry.*
The complex projective plane may be contrasted with the complex inversion plane or function plane in an interesting manner as follows : The homogeneous coordinates (ao a i} a z ) may be regarded as the
,
coefficients of
a quadratic equation
(32)
a$ + a z z + Kg* = 0.
i
(
l
Every such equation determines two and only two values
of
which may coincide
or
become
infinite
(if
zjz
0); and,
moreover, two
distinct points of the projective plane determine distinct quadratic
equations and hence distinct pairs of values of
2^/2^.
*If one "were to confine attention to real values, the definition of the plane of analysis given above would determine a set of elements abstractly equivalent to a real ruled quadric. This is distinct from the real inversion plane, because the latter
The numbers
(z
zj
may
be taken as homogeneous coordinates on
is a one-to-one and reciprocal correspondence between the points of a complex protective plane and the pairs of points on a complex protective line. It is important to notice that the pairs of points on the line are not ordered pairs, because a
a protective line.
Thus there
pair of values of z i /z Q
of the
taken in either order would be the pair
of roots
same quadratic.
representing the points of a complex line on a real unruled (e.g. a sphere), we have that the protective plane is in one-
Now
quadric
to-one reciprocal correspondence with the unordered pairs of points of the quadric. On the other hand, we have already seen that the com-
plex projective plane is in one-to-one reciprocal correspondence with the ordered pairs of points of the quadric. In either case the points
of a pair
may
coincide.
of
Tor further discussion

Infinite
the subject of this section see
"The
of the
Regions of Various Geometries" by M. Bdcher, Bulletin

Society, Vol.
106. Projectivities
of
XX
(1914),
p.
185.
general.
one-dimensional forms
in
The
theorems
of
the last four sections have established and made use of
the fact that the permutations effected
among
the lines of a regulus by
projective collineations form a group isomorphic with the projective a regulus is a one-dimensional form of the group of a line.
Now
second degree,* and the notion of one-dimensional projective transformation has been extended to all the other one-dimensional forms
76). It is therefore to be expected (Chap. VIII, Vol. I, particularly that an analogous extension can be made to the regulus. This we shall now make, but instead of dealing with the regulus in particular, we shall restate the old definition in a form which includes the cases
where the regulus

DEFINITION.
is
in question.
correspondence between any two one-dimensional forms whose elements are of different kinds and not such that all
elements of one form are on every element of the other form is said to be perspective if it is one-to-one and reciprocal and such
that each element of either
of the other form.
form
is
on the corresponding element
*The one-dimensional forms of the first and second degrees in three-space are the pencil of points, the flat pencil of lines, the pencil of planes, the point conic,
a pencil of points
Tins covers tne notion or perspectivity as aeimea in v 01. i Detween and a pencil of lines or between a pencil of lines and
It also defines perspectivities
a point conic, etc.

of
between
(1)
the lines
a regulus and the points on a line of the conjugate regulus, (2) the lines of a regulus and the planes on a line of the conjugate regulus, a regulus and the points of a conic which is a plane (3) the lines of
section of the regulus, (4) the lines of a regulus
and the planes of a cone tangent to the regulus. DEFINITION. correspondence between two one-dimensional forms or among the elements of a single one-dimensional form is protective
if
if it is the resultant of a sequence of perspectivities. This definition comprehends that made in 22, Vol. I, for forms of the first degree, and extended in 76, Vol. I, so as to include those of
r
and only
alent under duality to a point conic. In order ^ rt " it is necessary to prove that it does not
" 1of ion of perspectivity
between oneIn other words, we must prove w-dimensional forms of the
jvn><,
iwyrvv
(.*
prujvwvbt
ttn/t// ut/t/y
t,u
<//M>
new
to
definition only if it is
I.
protective according to the definition of
22, Vol.
To prove
this
theorem
it is sufficient
show
that a sequence of
of the first degree
perspectivities beginning
and ending with forms

the
first degree.
and involving forms involving only forms
of the second degree can be replaced

of
by one
This follows directly from the fact that each one-dimensional form of the second degree is
generated by projective one-dimensional forms of the first degree. For example, if a pencil of points [P] is perspective with a regulus and the regulus with a point conic and the point conic with some[I]
thing
else, it
follows
by the theorems
of
103, Vol.
I,
perspective with the pencil of planes [ml],
where
is
that [P] is a line of the
conjugate regulus and [ml} is perspective with the point conic. Thus the regulus [I] in this sequence of perspectivities is replaced by the pencil of planes [ml]. In similar fashion it can be shown by a consideration of the finite
of the
it
number
of possible cases that
however a form
second degree may intervene in a sequence of perspectivities, can be replaced by a form or forms of the first degree. The
enumeration of the possible cases is left to the reader, the argument required in each case being obvious.
spondences of
the protective
any one-dimensional form with
itself is
isomorphic with
group of a line. For let T be any projectivity of a one2 dimensional form F' of the second degree (e.g. a regulus), and let II 2 represent a perspectivity between F' and a one-dimensional form F
]
of
is
the
first
degree
(e.g.
H'T'II
a projectivity of is a projectivity of F*.

.
a line of the conjugate regulus). Then Jiril" In like manner, if T' is a projectivity of l
Hence
II establishes
an isomorphism
between the two groups. *107. Projectivities of a quadric. An involution on a regulus is the transformation of the lines of the regulus effected by a line reflection whose axes are the double lines of the involution. Since
any projectivity of a regulus
is a
product of two involutions,
it
may
be regarded as effected by a three-dimensional projective collineation which transforms the regulus into itself. Conversely, any direct projective collineation
two
line reflections
of
transforming a quadric into itself is a product of (Theorem 36) each of which effects an involution
on each
the reguli on the quadric.
This relation between the theory of one-dimensional projectivities and the projective group of a quadric may be used to obtain properties of the quadric analogous to the properties of conic sections
studied in Chap. VIII, Vol.
tions
I.
The discussion
is
based on Assump-
A, E, P, whenever this
improper points being adjoined to the space
is reqxiired for quadratic constructions. In Chap. VIII, Vol. I, we have seen that any projectivity on a conic determines a unique point, the center of the projectivity, and
that the axes of
any two involutions into which the
projectivity
may be resolved pass through its center. If, now, a projectivity F be given on a regulus, any plane IT meets the regulus in a conic C z on which is determined a projectivity T' having a point P as
center.
and points
This determines a correspondence between the planes IT of space which is a null system ( 108, Vol. I), and
hence the axes of the involutions into which the projectivity Y' can be resolved form a linear complex. The formal proof of this
statement follows.
37. For any nonidentical projectivity of a, regulus there a linear complex of lines [I] having the property that if ^ is any line of the complex not tangent to the regulusf there are three lines l^
THEOREM
exists
such that
is
polar
to
l^
and
to
with respect
to the regulus,
and
such that the collineation
\W vw
'
effects
tlie
yiven projectivity on the regulus.

to the
having this properly belongs
complex,
and
Moreover, every line ^ 1 / so do 1 4 2 8

, ,
.
z If Proof. Let li be a regulus and F a projectivity of It l^2 and ll are pairs of polar lines such that {IJ Z } {l^} effects the given t
2
.
projectivity
to R'
2
.
on
2
,
let TT
be any plane containing ^ and not tangent
is perspective with a projectivity F' projectivity F on 2 on the conic C 2 in which TT meets Moreover, {^2 } and {^4 } effect
The
involutions on
which are perspective with involutions

,-,,_ ,,.
I'
and I"
on
C 2 Thus on C z
.
But
(cf.
77, Vol. I)
l
l
is
the axis of
I'
and hence passes through the

l
t
center of I".
similar
argument shows that
(i
2, 3, 4)
passes
through the center of the projectivity perspective with F on the conic

in
which Hence
is
met by any plane containing

l^,
1 2, l
s,
all lines
l and not tangent to t defined as above are contained in the

.
set [I] of all lines
to
2
,
is also TT
which
I such that if -TT is any plane on I and not tangent on a point P defined as follows Let C 2 he the conic in 2 meets R* and F' the projectivity on C perspective with the
:
projectivity
on
J?
2
;
then
P is
2
the center of F'.
The
set [Z] obviously contains all lines tangent to

(if
at points of the
double lines
existent) of F.
If
,
is
any other
2
line of [I] let TT be a
plane on ^ and not tangent to R let C be the conic in which TT meets R 2 and let F' be the projectivity on C 2 perspective with F. By 79,
,
Vol.
I,
and the
definition of [I], F'

line,
g,
is
having ^ and another
as axes.
-
a product of two involutions Let lz and l^ he the polars of ^
and ls respectively. Then {l^ 2 } {1 3 1 4 } effects the perspectivity F' on C and hence effects F on R z By the first paragraph of the proof lz l s l^
.
are all lines of [I]. ated in the theorem.
Hence
all lines of [I]
have the property enunciis

2
By
TT
definition,
if TT is
It remains to prove that [I] a plane not tangent to

If TT is
a linear complex. the lines of [I] in
form a
TT
flat pencil.
tangent to
let
be the line of
TT.
on
and q the
line of the conjugate regulus

I on TT and the point
on
In case
is
fixed line of F, the lines
are the tangents to

no.
2
,
i.e.
the
pencil of lines on IT
In case v
is
not a fixed line
TT must have a polar line 1 2 be the projectivity on q perspective with T. If T is effected, by {IJJ {IJ,}, then T" is the product of two involutions, I' and I", which are perspective with the involu-
a line ot
[l\.
Any
other line ^ of
[l\
passing through the point pq. Let
T"
tions effected
on
JR*
by {^J and
{? 8 /J
respectively.
Since
l z
must
pass through the point pq, the latter when F" is expressed as a product of
involutions
is
is
a double point of I'. But two involutions, one of these

of its
(cf.
fully determined
by one
the latter
is
not a double point of T"

;
78, Vol.
double points in case Hence the I).
other double point, P, is fixed and since ^ must pass through it, it follows that all lines of [I] on TT pass through P. Moreover, it is evident that if l l is any line (except q) on TT and P, 12 its polar line,
and {yj- any line reflection effecting an involution on -R 2 which is perspective with I", the projectivity T is effected by {Z^} {^JHence [I] contains all lines on TT and P. Hence [I] is a linear
complex by Theorem 24, Chap. XI, Vol.
I.
z of a quadric surface Q which does not leave all lines of either regulus invariant determines a linear
THEOREM
38.
direct projectivity
congruence of lines having the property that if a^ is any line of the 2 congruence not tangent to Q there exist lines a z , b^ b z of the congruence such that
(33)
r = {a
.
as }-{& i y.
to the
Moreover, each line a^ having this property belongs and so do az b lf b

,
congruence,
Proof.
effects a projectivity
on each regulus
of
z Q and
,
each of
these reguli
theorems determines a linear complex of lines. The two complexes are obviously not identical and hence have a linear congruence in common. Any line a of this congruence is either
by the
last
tangent to
2
,
or such that there exist lines a z , & 1 & 2

,
which are in both
complexes and such that {a^} {^J effects the same projectivity as F on both reguli. Hence {a a } T. Moreover, any a^ for {bfi^ : 2 which az 5 l &2 exist satisfying this condition must, by the last theorem,
,
belong to both complexes and hence belong to this congruence.
The congruence referred to in the theorem may be on a point of Q z and on a plane 2 tangent to Q at this point ; or it may be parabolic and have a line of the quadric as directrix ; or it may be hyperbolic and' have a pair
1.
COEOLLARY
degenerate
and
consist of all lines
J.7UUJ.
juci/
i_/
ueuuuc
i/ue
uuiiî
U.CJLKJC
rci.cij.cu
LU
J.UL
uu
and
its
al
be the polarity by which every point is transformed into z This polarity transforms any line polar plane with respect to Q of C into its polar line, and the latter, by the theorem, is in C,
let II
.
Hence
II transforms
into
itself.
I,
According to
107, Yol.
any congruence
is either degenerate,
parabolic, hyperbolic, or elliptic. If degenerate, it consists of all lines on a point ft or a plane p, being on p. If II transforms such a con-
and p, and hence R must be gruence into itself, it must interchange 2 on Q* and p tangent to $ at R. The congruence C will be of this type z if & meets a in a point of Q and does not meet a t l z
.
If
is
parabolic, its one directrix
must be transformed
.
into itself
,
by
and hence must be a line of Q 2 This case arises if a v a z b^, & a 2 2 all meet the same line of $ and do not meet any other line of Q If C is hyperbolic, II must either leave the two directrices fixed
II,
.
individually or interchange them. In the first case each directrix must be a line of Q*, which implies that a lt a z , b lf b z all meet two 2 lines of Q and hence that all lines of one regulus are left invariant
by r, contrary
possible one.
2 Q but If C
,
to hypothesis.
lt
Hence the second

,
It occurs
when a a 2
case is the only v b z do not all meet any line of
are
met by a
pair of real lines.
is
has two improper directrices * and the reasoning the same as for the hyperbolic case.
is elliptic, it
DEFINITION.
A line
if it
is
pq
if
and only
meets both
if
said to meet or to be met by a pair of lines of them. pair of lines Im is said to
meet or cross a pair pq
both
and
in
meet pq.
EXERCISES
1.
The
lines
which cross the distinct pairs of an involution on a regulus
together with the lines tangent to the regulus at points of the double lines (if existent) of the involution form a nondegenerate linear complex.
2. If two pairs of polar lines, a a and 6 of a regulus meet each other, x 2 x 2 the involutions effected by {a^} and {b^b,-,} are harmonic (commutative) and their double lines form a harmonic set.
ft
,
* This
regulus.
may be proved as follows Let 1} 2 Any plane on 4 meets the regulus U 2

:
be lines of
Z
C
Z
meets
tliis
7i 2
gate to
conic in two improper points t which pass through t , P^ meet Z x

,
P P
,
containing
,
1?
not on the same in a conic, and Z4
The two lines of the regulus conju8 4 and hence meet all lines of 0.
,
3.
conic
Let T be a projectivity on a regulus 2 A variable plane meets R in a C 2 on which there is a projectivity F perspective with T. The axes of
.ft
.
the projectivities
4.
are lines of a linear congruence.

(1) in
Enumerate the types of collineations leaving invariant a quadric the complex space, (2) in a real space, (3) in various modular spaces.
* 108. Products of pairs of involutoric projectivities.
direct projective collineation of a quadric surface whose axes are polar, if it interchanges two points of the quadric which are not joined by a line of the quadric.
THEOREM
39.
is
line reflection
Denote the collineatioii by F, the quadric by Q 2 the two on it reguli by Rf and JR%, and the two points which F interchanges by A and B. Let a and b be the lines of R% on A and B respectively, and
Proof.
,
a'
and V those
b it effects
it effects
of J?2
on
A and B
respectively.
Since
it
interchanges a
a' and an involution on R%. Let I, m be the double lines of the involution on JRf, and p, q those of the involution on R%. F is evidently the product of {lm} by {pq} and hence is a line reflection whose axes
and
an involution on R%, and since
interchanges
are the line joining the points Ip
and
mq and
the line joining the

to
planes Ip and mq. These
two
lines are polar
with regard
2
.
THEOREM
not meet the
40.
z which are not on a quadric Q and do same line of Q z are met by one and but one polar pair
Two
lines
of
lines.
Proof.
Let one of the given lines meet the quadric in

it
and A'
and the other meet

to A,
/ I
in
B and B
1
.
By Theorem 35
which
there
carries
is
direct projective collineation of the quadric
a unique to A', A'
and
B to
B'.
By Theorem 39
this is a line reflection {lm}

z
.
and
and and
m are polar with respect to Q m both meet the line AA', and
line BB'.
Since {lm} transforms A to A', to B', I since {lm} transforms
and
m both meet the

= {I'm'}.
If there
were another pair
of polar lines
I',
meeting AA' and
BB', {I'm'} would interchange

{lm}
A and A' and B and B'. By Theorem 35
2 lines which are not on a quadric Q and do not z meet the same line of Q are met by two and only two lines which are conjugate to them both with regard to Q\
COROLLARY. Two
Proof.
This follows directly from the theorem, because two mutu+T-rT--\
14 v* rta
ot-*
nvt
o-nfc
Virvf-.l-i
rt/"vn
TTI/TQ^"^
i".f\
THEOREM
4tl.
J.J
a,
simple nexagon
'in
'inscrwea -m a yruuurw sur-
face in such a
way
that no two of its vertices are on a line of the
of
a polar pair quadric, the three pairs of opposite edges are met each l>y lines, and these three polar pairs of lines are in the same linear
congruence.
Proof.
Let
A^B^C^B^C^
are polar
Z
7,
be the simple hexagon.
By
the last
pair of
like
theorem the pair

lines G
I}
of opposite edges
A^B Z A Z B^
,
is
met by a
which
lt
manner
1
B C B^C ----*
a
with respect to the quadric. In are met by a polar pair a v a a and CÂ Z
, ,
r\^r,iA n -~
fv, q
product
of line reflections,
3 (7
2,
{fc^J carries
Cs
to
I}
and
<.
v
,
carries
to
lf
{bfa} carries
Ct to A z and
{c xc 2 } carries
it is
and by Theorem 39
A Z to jB r Hence F interchanges B^ and -B 2 , a line reflection. Denoting T by {d^d^ we have
By Theorem 38
In view
of
the axes of the four line reflections in this equation
are all lines of the
same congruence. the corollaries of Theorems 38 and 40 this theorem

:
may be
such a
restated in the following forms

1.
COROLLARY
If a simple hexagon
of
lines
is
inscribed in a quadric in
way
that no two of its vertices are
on a
three polar pairs
which meet
the
line of the quadric, the pairs of opposite edges are
met by a polar pair of
lines (which
may
coincide],
If a simple hexagon is inscribed in a quadric surface in such a way that no two of its vertices are on a line of the quadric, each pair of opposite edges is met by a unique pair of lines conjugate to both edges, and the latter three pairs of lines are met by a pair of lines conjugate to each of them. The lines of the last pair
2.
COROLLARY
may
coincide.*
of the American Mathematical Society, Vol. XVI (1909), pp. 55 and 62. A theorem of non-Euclidean geometry from which this may be obtained by generalization has been given by F. Klein, Mathematische Annalen, Vol. XXII
/1883), u. 243.
* Bulletin
This theorem is closely analogous to Pascal's theorem on conic sections (Chap. V, Vol. J). In the Pascal hexagon the pairs of opposite skl<!s determine three points A, B, C which are collinear. In the hexagon inscribed in a quadrie they determine three pairs of lines aâ^ 6.J)2 c^c^ which are in a linear congruence. In case the vertices of the hexagon are coplanar, the
,
theorem on the quadrie reduces directly to Pascal's. The Pascal theorem may be proved by precisely the method used above. For let vl 1 /i 2 C1 .'l 2 -> 1 C'2 be a hexagon inscribed in a conic and let A be the
point
B,
\Bb}, and
be ((\A.,, .-I 1 <" 2 ), and C be (J. 1JB 2 5 1 /1 2 ). Let {Aa}, C\J.> 2 '), B Cc} be the harmonic homologies effecting the involutions having .1. as centers. By construction the projectivity effected by {Cc} {Bb} {Aa}
(-/->!
C'2 ,
{
on the conic carries Bl to B z and J!.2 to 7J 1? and hence its center and axis by D and d, we have
,
is
an involution. Denoting
{Cc} -{Bb}
This implies
{Bb}
{Aa}
= {Cc}
{Dd}.
By the theorems of Chap. VIII, Vol. I, the line AB is the axis of the projectivity effected by {Bb} {Aa} and must contain C and D. Plence A, B, C are
collinear.
Pascal's theorem is thus based on the proposition that the product of three involutions on a conic is itself an involution if and only if the centers of the
three involutions are collinear, i.e. if and only if their axes are concurrent. Let us denote an involution whose double points are L and {LM}, as
Mby
in Ex. 11,
52.
If the involution is represented
on a
conic, the doxible points
are joined by the axis of the involution. The proposition above then takes the form The product {L 3 AT3 } {L 2 2 } {L : r } is an involution if and only if the lines V L 2 2 L S S concur. The concurrence of the three lines
:
L^M
means
either that the three point pairs
are themselves pairs of be stated as follows

:
have a point in common or that they an involution. Thus the theorem on involutions may
three
THEOREM
{LÎ^}, {L 2
Z/2/1/,,
42.
2 },
LS
In any one-dimensional form a product of
involutions
{L S
3 }
is
an
involution
in
case the pairs
of points
L^MV
the
hare a point in common or are pairs of an involution; and
product
is
not
an
involution in
any
other case.
points of the involutions may be either proper or improper In order to state the result entirely in terms of proper elements, the involutions may be represented on a conic and the condition of the concurrence of their axes, as above or it may be stated in terms
The double
(real or imaginary).
expressed by saying that they all belong to the same pencil of involutions, or by saying that they are all harmonic to the same projectivity. This theorem on involutions in a one-dirnensional form is fundamental in the theory of those groups of projectivities, in a space of any number of
it is
essentially the
same as Theorem
8,
Chap. IV, which was fundamental in
the theory of the parabolic metric group in the plane. Corresponding theo114 rems in the Euclidean geometry of three dimensions will be found in
and 121, Chap. VII.
The same principle appears as Theorem 27, Cor. 1, Chap. Ill, in connection with the equiaffine group. These groups are all projective and on that account related to the projecgroup of a one-dimensional form. But the essential feature which they have in common is that every transformation of each group is a product of two involutoric transformations of the same group. On this account, even without
tive
their
common
projective basis, the geometries corresponding to these groups
must have many features in common. In particular, whenever there is some class of figures such that if two of the figures are interchanged by a tranfv formation, the transformation is of period two, there must exist a theorem analogous to Pascal's theorem. As examples of this may be cited Theorem 41 above Ex. 6, Ex. 1, 80, Vol. I 122, below and in the list of exercises below, Ex. 4, referring to the group of point reflections and translations, Exs. 5, 6 referring tc me Euclidean group in a plane, Ex. 7 referring to the equiaffine group. On this subject in particular and also on the general
;
;
theory of groups generated by transformations of period two, the reader should consult a series of articles by H. Wiener in the Berichte der GesellVol.
schaft der Wissenschaften zu Leipzig, Vol. XLII (1890), pp. 13, 71, 245; XLIII (1891), pp. 424, 644; and also the article by Wiener referred
to in
45, above.
Cf also
.
80, Vol.
I.
EXERCISES
1.
(Converse of Theorem 41.)
If
the three pairs of opposite edges of a

i
simple hexagon are met by three pairs of lines aâ z fij&g, e 1 c 2 in pairs of points which are harmonically conjugate to the pairs of vertices with which they are collinear, and if the lines a v a 2 b^ J 2 c v c 2 are in the same linear congruence,
,
then the vertices of the hexagon are on a quadric surface with regard to wnich
aâ z
,
l-J) z ,
c x c 2 are
2.
Two
If
polar pairs of lines. pairs of lines which are polar with regard to the same regulus
cannot consist of lines of a

3.
common
regulus.
two
lines
and
are
met by two
pairs of lines which are polar with

of intersection of
are polar. respect to a quadric, / and 4. In a Euclidean plane let A, B, C
be the three points
pairs of opposite sides of a simple hexagon. If sides containing them, and is the mid-point of
and
it.
are mid-points of the one side containing it, then
C is
5.
also a mid-point of the other side containing
Let
A 1 B Z C A Z B 1 CZ
1
be a simnle hexagon
A. T
perpendicular bisector of the point pair
BZ
a Euclidean plane. If the coincides with that of A Z B V and

ir.
the perpendicular bisector of z O^wiih that of B^CV and the perpendicular bisector of C!.A. with that of d. A tJ.P.n tha t.Tirp.P. nprnp.-nrilir'.nlfl.r Hi'spr-frY!"
joining their mid-points are concurrent.
109. Conjugate
of antiprojectivities
of
imaginary lines of the second kind. The theory ( 99) and the extended theory of projectivities one-dimensional forms ( 106) will now enable us to complete the
which we were not ready to discuss in 78. protective space and let S be a three-chain of
to S' in the
theory of conjugate imaginary elements in certain essential details Let S' be a complex
S', i.e.
manner described
in
and
70,
and
a space related let us use the
definitions
and notations
collineatioii of S' is
'
of 70. The simplest type of antiprojective given by the equations

,
(34)
tf
=z
xi
=
i>
^g=
of
The frame of reference is such that the points nates. The transformation changes each point
S have real coordi-
where the
a's
and
(a
/3's
are real, into the point

,
-*0
if
a^-i/3,,
-*',,
-*
8 ).
These two points

(
distinct are joined

,
by the
,
real line
+ X0 ^ + X^,
a z + Xft
a.
+ X^
3)
double points of the involution determined by the transformation of the parameter X,
and
are the
Comparing with the definition

it is
of
conjugate imaginary points in
78,
clear that (34) is the transformation

its
goes to
as real.
conjugate imaginary point,
by which every point of S' the points of S being regarded
From
point invariant,
the fact that the transformation (34) leaves no imaginary it follows that it cannot leave any imaginary line or
of plane invariant. For the real line through an imaginary point would the given line or plane is left invariant by (34), and hence
be left invariant
by
(34).
On
the other hand, (34) leaves every real
Hence
is its
(34) interchanges any element of S' with the element which 78. conjugate imaginary according to the definition of The definition of 78 defines the notion of conjugate imaginary
all
elements for
one-dimensional forms of the
first
or second degrees,
and the theorems
of that section cover all cases except that of a pair
of conjugate imaginary lines which are the double lines of an elliptic involution in the lines of a regulus.
DEFINITION.
tic
An
involution in a
is
flat
imaginary line which is a double line of an ellippencil is said to be of the first kind, and one
is
which
a double line of an elliptic involution in a regulus
said to
be of the second kind.
THEOREM
second kind.
Proof.
43.
Any imaginary
be an imaginary
line is either of the first or
of the
Let
line.
(
It
cannot contain two real
points, else it
would be a
In the
first
I,
real line
70).
Hence
it
contains one or
I,
no real
point.
case let
be the real point on
one of
the imaginary points on
and
P the imaginary point

plane
conjugate to P.
The
line
PP is real, and hence the

OP
and
OPP is
real.
Hence by
78
are the double lines of an elliptic involution in the pencil of real lines on the point and the plane OPP.
the lines
OP
and
let P, Q and R be three points of and let P, Q R be their respective conjugate imaginary points. The lines PP, QQ, RE are real and no two of them can intersect, for if they did I
In the second case
would be on a
real plane,
the last paragraph. On the real line
and we should have the case considered in Hence these lines determine a regulus R* in S.
is
PP there
P
by
78 an
elliptic involution
having
as its imaginary double points. Hence there is an elliptic involution in the regulus R%, conjugate to R%, having I as one double line and a line I through as the other. The lines I and I are conjugate
and
imaginary lines by definition, and satisfy the definition of imaginary lines of the second kind. Since (34) transforms each element into its
conjugate element,
it is
clear that
contains
Q and
as well as
P.
its
The system
of real lines obtained

I is,
conjugate imaginary point on
by joining each point of I to by the reasoning above, a set
of
lines of the real space S, no two of which intersect. Any four of them determine a linear congruence ( 107, Vol. I) C in S and also a linear congruence C of S'. The congruence C has the property that each of
its
hues
is
contained in a line of
I
C,
and C evidently
1.
is
the set of
all
lines joining points of
to points of
Hence C
is
107, Vol. I, gruence according to the definition of real lines meeting I and I. Hence the system of real lines joining
points of I to their conjugate imaginary points in S, or in other words
:
an elliptic conand consists of all

congruence
is
an
elliptic
THEOREM
of an
44.
An
imaginary
line
of the second kind
is
directrix
elliptic
congruence.
The observation, made in the argument above, that there is one line congruence through each point of an imaginary line of the second kind, shows that an elliptic congruence may be
of a certain elliptic
taken as a real image of a complex one-dimensional form. This of course implies that the whole of the real inversion geometry can be
carried over into the theory of the elliptic
Of. the exercises
congruence and vice versa.
below.
lines of the
The relations between the imaginary
second kind and
the regulus and elliptic congruence are fundamental in the von Staudt 6. In addition theory of imaginaries which has been referred to in
to the references given in that place, the reader
may
consult the Ency3,
clopedie des Sciences Mathematiques, III 8,
19,
and III
14, 15.
EXERCISES
has a pair of conjugate elliptic congruence in a real space lines of the second kind as directrices.
1.
An
imaginary
I
2.
The correspondence by which each point

its
of an
imaginary line
correI
sponds to
its
conjugate imaginary point

line.
is
an antiprojectivity between
and
conjugate imaginary Under the projective group of a real space any imaginary point is transformable into any other imaginary point, any imaginary line of the first kind
3.
into any imaginary line of the first kind, and any imaginary line of the second kind into any imaginary line of the second kind an imaginary line of the first kind is not transformable into one of the second kind.
;
a one-to-one reciprocal correspondence between the points of a complex line and the lines of an elliptic congruence in a real space in which the points of a chain correspond to the lines of a regulus. By means of this
4.
There
is
5.
Let S a be a thi >e-dimensional complex space.
'
Any
five
noncoplanar
.
points of SpJ determine a unique three-chain, which is a real S 8 This S 8 is ' 6 and 70. Through any point P related to S 3 in the manner described in of SB not on S 3 there is ( 78) a unique line which contains a line of S 8
,
(i.e.
a chain C t ) as a subset.
On
this chain
lution having involution.
as a double point.
Let
C 1 there is a unique elliptic invobe the other double point of this
and P are the conjugate imaginary points with regard to the real space S 3 and the transformation of Sg by which each point P not on S 3 goes to P, and each point on S 3 is left invariant, may be called a reflection in the three-chain S 3 Any transformation which is a product of an odd number of reflections in three-chains is an antiprojective collineation, and any transformation which is a product of an even number of reflections
,
.
in three-chains
is
a projective collineation. Every collineation
is
expressible
in this form.
110.
The
principle of transference.
We have seen how the geometry

initially as
of the inversion of the
group in the plane, arising
an extension
Euclidean group, is equivalent to the projective geometry of the complex line and also to that of a real quadric which may be specialized as a sphere. We have also seen the equivalence of the
u ?ional forms in projective groups of all one-diu! any properly projective space. Since the regulus is a o~ie-dimensional form, this gave
a hold on the group of the general quadric. The latter group in a complex space has been seen to be isomorphic with the complex
inversion
group
and
also
with the fundamental group
of
the
we have helped ourselves forward by transferring the results of one geometry to another, combining these with easily obtained theorems of the second geometry, and thus extending our
knowledge of both. This is one of the characteristic methods :: modern geometry. It was perhaps first used with clear understanding by 0. Hesse,* and was formulated as a definite geometrical principle
34. (Uebertragungsprinzip) by !\ Klein in the article referred to in This principle of transference or of carrying over the results of one geometry to another may be stated as follows Given a set of elements
:
function plane. At each step
[e]
and a group G of permutations of

\_T~\
these elements,
and a
G.
fet
\e'~\
of
"be
theorems
which
state relations 'left
invariant
~by
Let
another
set
If
there is
of elements, and G' a group of permutations of \e'~\. a one-to-one reciprocal correspondence between." \e] and [V]
in which
G is simply isomorphic with G', the set of theorems [T} deterT/ mines by a mere change of terminology a set of theorems [^ ] which invariant are e' which elements state relations among by G'. left
This principle becomes effective
when the method by which

it
[e]
and
are
denned
is
such as to make
[e']
which are not so
easily seen for
and
G'.
easy to derive theorems This has been abundantly
illustrated in the present chapter,
bub the series of geometries equivalent to the projective geometry on a hue could be much extended. Some of the possible extensions are mentioned in the exercises below.
From
is clear
[e]
the example of the conic and the quadric surface ( 107) it that in order to carry results over from tbe theor}-- of a set and a group G to a set |V] and a group G' it is not necessary that
the correspondence be one-to-one. The transference of theorems is, however, no longer a mere translation from one language, as it were,
to another,
DEFINITION.
tations of
[e],
but involves a study of the nature of the correspondence. of permuGiven a set of elements [e"] and a group
the set of theorems [J ] which state relations among the elements of [e] which are left invariant by Gr and are not left
1
invariant
by any group
of
permutations containing Cr is called a
mathematics.* generalized geometry or a branch of This is, of course, a generalization of the definition of a geometry 34 and 39. At the time when the role of groups in employed in
geometry was outlined by Klein, the only sets [e] under consideration were continuous manifolds, i.e. complex spaces of n dimensions or loci defined by one or more analytic relations among the coordinates in such spaces. The older writers restrict the term "geometry" of
points
by means of this restriction on the set [e]. But in view of the existence of modular spaces and other sets of elements determining sets of theorems more nearly identical with ordinary geometry than some
of those
state the definition in the
admitted by Klein's original definition, form adopted above.
it
seems
desirable to
in the introduction to Vol.
In case the set of theorems [T] is arranged deductively, as explained I, it becomes a mathematical science. The
of the
problem
foundation of such a science
is
that of
determining,
if
possible, a finite set of
assumptions from which [T]
may be
deduced.
* The generalized conception of a geometry is discussed very clearly in the article in the Encyclopadie der Math. Wiss. Ill AB 46. number of special by G. Fano * -_ii! j J.T-- l_j.j. V1* * + V, otH-inld
>
EXERCISES
a projective collineation interchanges the two reguli on a quadric, homologous lines of the two reguli meet in points of a plane. z * 2. Let R z be a the pole of <a regulus, w a plane not tangent to R , and
1.
If
(<a
may
conveniently be regarded as the plane at infinity of a Euclidean space).

z
T of R may be effected by a collineation F' leaving all lines of the conjugate regulus invariant. This collineation multiplied by the harmonic homology { O<a} gives a collineation T" interchanging the two reguli. By Ex. I, A
projectivity
to
F" determines a unique plane. Let P be the point polar to F" with regard 1 The correspondence thus determined between the projectivities T of z 2 ft and the points of space not on R is one to one and reciprocal. It is such that projectivities which are harmonic ( 80, Vol. I) correspond to conjugate points with respect to R", and all the involutions correspond to points of o>. *3. The construction of Ex. 2 sets up a correspondence between the projectivities of a one-dimensional form and the points of a three-dimensional space which are not on a certain quadric. The same correspondence may be
obtained by letting a projectivity

yf
a x
a^x
+ +
"i
a8
correspond to the point (a a v a 2 a 3 ). The relations between the one-dimensional and three-dimensional projective geometries thus obtained have been
,
by C. Stephanos, Mathematische Annalen, Vol. XXII (1883), p. 299. *4. Develop the theory of the twisted cubic curve in space along the following lines: (1) Define it algebraically. (2) Give a geometric definition.
studied
(3) Prove that Definitions (1) and (2) are equivalent. (4) Derive the further theorems on the cubic as far as possible from the geometric definition. It will be found that the properties of this cubic can be obtained largely from those of
conic sections
and one-dimensional projectivities in view
of
an isomorphism
of
the groups in question. The theorems should be classified according to the 83. principle laid down in
*5.
A
R
rational curve in a space of k dimensions is a locus given paramet:
rically as follows
=R
, x
(f),
*1
(0"- Zn=R*,
, ^
-^
where
is
-,R n (f) are rational functions of t. In case k = n and the locus (t), not contained in any space of less than n dimensions, the curve is a normal
Develop the theories of various rational curves along the lines outlined For reference cf. 28 of the encyclopedia article by Fano referred and articles by several authors in recent volumes of the American. Journal of Mathematics.
curve.
in Ex. 4.
to above
* 6. The linear dependence of conic sections may be defined by substituting " " " point conic or line conic," as the case may be, for "circle in the definition 100. Develop the theory of linear families of conies of given at the end of one, two, three, and foxir dimensions, \ising the principle of correspondence
"
whenever possible and classifying theorems according to the principle
laid
CHAPTER
VII
AFFINE AND EUCLIDEAN GEOMETRY OF THREE DIMENSIONS

111. Affine geometry.
DEFINITION.
Let
TT^
fixed plane of a projective space S.

TT
The
set of points of
be an arbitrary but S not on
is
called a Euclidean space
and
7r n is
called the
of this space.
The plane
7r m
;
and the
all
points and lines
plane at infinity on TT* are said
to be ideal or at infinity
other points, lines, and planes of S
are said to be ordinary. When no other indication is given, a point, line, or plane is understood to be ordinary. Any projective collinea-
tion transforming a Euclidean space into itself is said to be affine ; the group of all such collineations is called the affine group of three dimensions, and the corresponding geometry the affine geometry of three dimensions.
DEFINITION.
Two
ordinary lines which have an ideal point in
are said to be parallel to each other. Two ordinary planes which have an ideal line in common, or an ordinary line and an ordinary plane which have an ideal point in common, are said to be
common
parallel to each other.
In particular, a line or plane is said to be parallel to itself or to any plane or line which it is on. For ordinary points, lines, and planes we have as an obvious consequence of the assumptions and definitions of
Chap.
I,
Vol.
I, 1.
the following theorem
Through a given point there is one and only one line parallel to a given line. Through a given point there is one and only one plane parallel to a given plane. If two lines, I and I', are not in
THEOKEM
the
same plane there

to I to
I'.
is
parallel parallel
and
I'.
If
one and only one plane through a given point I is I and V are parallel, any plane through
Another obvious though important theorem
is
the following
THEOREM
2.
The transformations
effected
in
an ordinary plane
7t
by the affine group in space constitute the affine group of the Euclidean
In consequence 01 tins theorem we nave tne wnoie amne plane geometry as a part of the affine geometry of three dimensions, and we shall take all the definitions and theorems of Chap. Ill for granted
without further comment.

This discussion
is
valid for any space satisfying Assumptions A, E.
The
affine
geometry
31,
considered in
Exs. 5-7 below.
an ordered space (A, E, S) has already been and certain additional theorems are given in
of
EXERCISES
1.
The lines
joining the mid-points of the pairs of vertices of a tetrahedron
',. c-iassuy wie nuaur.ii; sur.ia.utja J.KJIU the point of view of real affine geometry. Develop the theory of diametral lines and planes. The real projective classification of the nondegenerate quadrics has been given in 103. The 19. affine classification is given in the Encyclopedic des Sc. Math. Ill 22,
*3. Classify the linear congruences from the point of view of the real
affine
107, Vol. I. geometry. Cf *4. Classify the linear complexes from the point of view of real Vol. I. Cf 108, geometry.
.
.
affine
respect to the coordinate system used in line joining (b v b 2 , & a ) are (a v 2 a 8 ) and
5.
With
81 the points of the
A=
B=
nr
B
6.
i+x
+x r
1. The segment corresponding to X = oo and the point at infinity to X = .4 B consists of the points for which X > and its two prolongations of those for which X < 1 and 1 < X < respectively.
Two
points
and D' are on the same
side of the plane
ABC
if
and
Only
7.
5 (AB CD) = 5 (ABCny
101 and dealing with an ordered Euclidean Using the notation of or w is at infinity { 0(o} is an affine collineation which alters sense if and {IV} is an affine collineation which does not alter sense if I or f is at infinity. In an ordered projective space {II'} is, and {0o>} is not, a direct
space,
collineation.
elation having
If
is
I
112. Vectors, equivalence of point triads, etc. DEFINITION. An -ir* as its plane of fixed points is called a translation.
is
an ordinary line on the center of the translation, the translation

I.
said to be parallel to
The properties of the group of translations follow in :um the following evident theorem.
large part
112]
AFFINE GEOMETEY
3.
289
THEOREM
The transformations
TT
by the translations leaving

lations of the
effected in an ordinary plane TT invariant constitute the group of trans-
Euclidean plane composed of

of this
the
ordinary points of
tr.
As
corollaries
we have
statements about translations in
space which are verbally identical with Theorems 3-7, Chap. III. Theorem 8, Chap. Ill, generalizes as follows
:
are three noneoplanar lines and T any translation, there exists a unique triad of translations T x T T z parallel to OX, Y, OZ respectively and such that
COROLLARY. If OX,
Y, and
OZ
T = T;r;r s
theory of
The congruence under translations generalizes to space without change, and the contents of 39 and 40 may be taken as applying to the affine geometry in three-space. In like manner the
" replaced by Euclidean plane 42 then apply without change. space." The theorems of We arrive at this point on the basis of Assumptions A, E, Adding Assumption P we take over the theory of the ratio of col-
to space
definition of a field of vectors and of addition of vectors " " Euclidean be if the words
is
carried over
linear vectors
from
43, 44.
Some
of
the theorems to which
it
may
be applied without essential modifications of the methods used

definition of equivalence of ordered point triads in
if
in the planar case are given in the exercises below.
The
48
i
is
such that
plane
IT',
a plane
IT
be carried by an affine collineatio]
to
any two equivalent point

IT'.
triads of TT are carried to
equivalent point triads of of ordered point triads in

point
triads
49
are
Moreover, the definition of is such that if two coplanar ordered

carried
two measure
ABC,
DEF
7?i
by an
affine
collinerôn to
A'
C',
D'E'F'
respectively,
(ABC}
'
m (DEF}
rp.sTilf-.
= m (A'B'C'} ~
m (D'E'F']
'
This
in vip.w nf Thp.nrp.m 39. Cha.n. Til.
defends on the
p.rmv.
DEFINITION.
lent
if
Two
if
ordered point triads

carried
and only
AB C may be
is
ABC and A'B'C' are equivaby a translation to an ordered

.of
triad
A"B"C" which
equivalent in the sense
48, Chap. Ill,
to A'B' C'.
The fundamental propositions with regard to equivalence,

oped in
48, remain valid under the extended definition.
as devel-
Thus
if
and
A B C ^ A B C2 AB C ^ A 2B
i
l
2 C,,
if
AB C *= A^BfJ^
it
111
Ti^
etc.
ovftmoinn nf
thp. nnt.ion
of
i
equivalence carries with

of
a cor-
measure,
i.e.
measure
is
now
proviso that the unit triad in any

^
triad in
of
any
parallel plane,
equivalence of ordered point tot generalize directly to the

jnal space.* shall therenieasures of an ordered set of
>
We
develop the corresponding
synthetic theory (cl J&x. 13 below). DEFINITION. By the measure of an ordered tetrad of points A I} A z , as unit is meant the A s A t relative to an ordered tetrad
,
OPQR
number
(2)
where
(aiv
a iz aia ) are nonhomogeneous coordinates of A { (i =1,

,
2, 3, 4)
in a coordinate system in
(0, 1, 0),
which
0,
P, Q,
are (0, 0,
0), (1, 0, 0),
(0, 0,
if
1)
respectively.
if
-^
Two
ordered tetrads are said to be
equivalent
and only
geometry the number

tetrahedron
they have the same measure. In real affine (AÂÂ^) is called the volume of the
AÂyA^
relative to the unit tetrahedron
OPQB
and
is
denoted by v
is
(AÂÂ^.
The theory of the equivalence of point pairs, triads, tetrads, etc. the most elementary part of vector analysis and the Grassmann Ausdehnungslehre. This subject in particular, and the affine geometry
* Cf
.
M. Dehn, Mathematisclie Annalen, Vol.
LV (1902), p. 465.
of three
dimensions in general,
it
is
worthy
of a
much more
extensive
have referred only to that receiving here. part of the subject which is essential to the study of the Euclidean
is
treatment than
We
geometry of three dimensions. In the following exercises the coordinate system is understood to be that which is described in the definition of measure of ordered
tetrads above.
The
vectors OP, OQ,
OR
are taken as units of measure
for the respectively parallel
triads
OPQ, OQK,
ORP
systems of vectors. The ordered point are taken as imits of measure for the respectriads.
tively parallel
systems of ordered point
DEFINITION.
is
the projection of a set of points [X] on the re-axis meant the set of points in which this axis is met by the planes
By
and parallel to the plane x ~ ; and the through the points projection on the y- and 2-axes have analogous meanings.
By
points
meant the
the projection of a set of points [A"] on the plane x is set of points in which this plane is met by the lines on
and
parallel to the a>axis
and the projections on
the
planes y
and
have analogous meanings.
EXERCISES
1.
The measures
of ordered tetrads of
x'
points are unaltered

6 10 , & 20 , b so ,
by
trans-
formations
(3)
z'
- b n x + b 12 y 4- & 13 s + / = bai x + 22 y + b 23 z + = b 31 x + b s2 y 4- b as z +
l>
subject to the condition
A = 1,
where
''i, "
'
(*)
This group
The group
2.
et/uictffine group and also the special linear group which A 2 = 1 leaves volumes invariant. Ratios of measures of ordered tetrads of points are left invariant by
is
called the
for
the affine group.

3.
are in the
In an ordered space two ordered sets of points AS CD and A'B'C'D same sense or not according as m (A BCD') and m {A'B'C'D'} have
the same sign or not. 4. The product of

translation
if
/'
two and m' are
line reflections {IV}
at infinity
and
and {mm'} (cf. and m, are parallel.
101)
is
6. The projections of a point espectively have the measures
pair
on the
x-,
y-,
and
z-axes
f2
xv
i/2
T/J,
z2
Zj,
a;
and those
of the ordered point triad
OP^P Z on
i
the planes
0,
0, z
respectively have the measures

N
ll\ 1/2
9/~' ~2
oA
zi
zi
2
x _ -,'"
v
^2
xi
T
2
2/i
'11
v/2
These numbers
satisfy the relation
/3/x
yv
= 0. P P2 = Vect P{P 2
X
Any two
points
PiP 2
of the line
PiP2
such that Vect
deter-
^"""
~.,
y, \, p., v- These numbers are proportional to 1nr> Vol. , J) of the line. PjP^
ft,
Ex.
6,
'
= m (OPP^P^), p = m (OQP^),
the numbers analogous to
a,
(3,
n .re
y,
t-
yv +
f
'
Axi' 4-
/J>/3'
vy'.
an ordered point
triad
PPP
i
on
ire
The homogeneous
coordinates of the plane

?!
-.,
P P2 P S
1
are (w
M I} w 2 w g ), where
,
.'/l
//.,
-I'
r,,
=
I
,.
(OP 1 P2 P8 ).
and
*'o
//a
9. If
Pv P P P
2, 8
,
are four noncoplanar points
P P4
', 8
'
points oollinear with

if
P8 and 7^, then m (PxPaPaPO = m (PiPJW).

10. If
Vect (PS'P4')
= Vect (P3 P.,)
if
are two and only
P P2 P P
I;
,
3,
are four noncoplanar points and the lines
P\PZ
>
P\P%>
P^P'e have a point in
common and
Vect (PjPj)
then
Vect (P{Pfi + Vect (P'{P'^,

8
m (P^P^P^ = m (PiPP P4 ) + m (Pi'PJ'PsP*)-
*11. Study barycentric coordinates and the barycentric calculus for threedimensional space. Of. 49. 27, and references to Mobius in 51, *12. Study the measure of n-points in space, generalizing the exercises
in
49.
*13. Define two ordered tetrads and A'B'C'D' as eqxiivalent provided that (1) A', B = B', C", and the line DD' is parallel to the ( plane ABC, or (2) if there are a finite number of ordered tetrads t v
ABCD
A-
C=
if
and t n to A'B'C'D'. Develop a theory of equivalence as nearly as possible analogous to that of 48. Show that two tetrads are equivalent in this sense and only if they are equivalent according to the definition in the text. *14. An elation whose center is at infinity and whose plane of fixed points
ordinary is called a simple shear. The set of all products of simple shears is the equiaffine group. Develop the theory of the equiaffine group on this basis. 52 to space ? Is it possible to generalize
15. If a plane
is
meets the sides
A AV
.,
A-Â Z ,
.,
A nA
of a simple polygon
AÂ^y,
An
in points
B Bv
,
Bn
respectively,
AB A&
A^
(
AzBi
A n Bn A Bn A A
,
16. If a quadric surface
104, Vol. I) meets.the lines
lt
A-Â Z
respectively in the pairs of points
B C B CV
,
.,
B n Cn
respectively,
A&
following identity:
A C A.&
A& A&
A,C\ A,C\
'
A n B n A n Cn = A,Bn A Cn
of
*17. Six points of a plane
no three
which are
collinear satisfy the
m (123) m (456) - m (124) m

The
ratio of
(503)
+ m (125) m (634) - m (126) m (345) = 0.

These
any two terms in propositions are given by W. K.
this
sum is a protective invariant. Clifford in the Proceedings of the
London
Mathematical Society, Vol. II (1866), p. 3, as the foundation of the theory of two-dimensional projectivities. Develop the details of the theory outlined by
Clifford.
Cf also Mobius, Der barycentrische Calcul,

.
221.
113. The parabolic metric group. Orthogonal lines and planes. be an arbitrary but fixed polar system in the DEFINITION. Let
plane at infinity TT.

or orthogonal
lines
This polar system shall be called the absolute polar system. The conic whose points lie on their polar with respect to 2*, is, if existent, called the circle at infinity.
of all collineations leaving
its
The group
2^ invariant
is
called the
parabolic metric group and
transformations are called similarity
transformations.
Two
figures conjugate
under this group are said
to
be similar.
DEFINITION. Two ordinary planes or two ordinary lines are orthogonal or perpendicular if and only if they meet TT^ in conjugate lines or points of the absolute polar system 2*,. An ordinary line and plane are orthogonal or perpendicular if and only if they meet ir x in a
line perpendicpoint and line which are polar with regard to S m ular <-,n itsp.lf. i.ft- a line throueh a noint of the circle at, infirn'f-.v is
.
called a
Le. a
minimal or isotropic line. plane perpendicular to itself, is called plane meeting TT*, in a tangent to the circle at infinity, a minimal or isotropic plane. As the analogue of Theorems 2 and 3 we have
THEOREM
4.
The, similarity
TT
transformations which leave an ordi-
nary nonminimal plane

ordinary points of
tr.
effect in TT the transformations the of a parabolic metric group in the Euclidean plane consisting of
invariant,
Generalizing Theorem
1,
Chap. IV,
we have
of a Euclidean space the correspondevery point ence between the lines and their perpendicular planes is a polar system,
5.
tlno
THEOREM
At
nwnn'artinn o/"
"_.
All
t.l.f
lines
perpendicular to a given through vular to the given line at 0; and all the to a given plane are on the line through
r
f existent,
fn.nd
the isotropic lines through
the isotropic planes through
the
f lines.
nonminimal planes meet in a and two perpendicular nonminimal lines are parallel to a nonminimal plane. COROLLARY 2. If a plane 1 is perpendicular to a plane 2, and 2
A
nonminimal
line,
is
parallel to a plane 3, then 1 is perpendicular

to
to 3.
If a plane
1 is
perpendicular
2 is parallel to a line or plane 3, then I is perpendicular to 3. If a line 1 is perpendicular to a plane 2, and 2 is parallel to a line or plane 3, then 1 is perpendicular to 3.
line 2,
line 1 is
is
and
If a
3,
perpendicular
to
line 2,
and 2
is
parallel
to
line
then 1
perpendicular
to 3.
THEOREM 6. Two nonpar allel lines not both parallel to the same minimal plane are met by one and only one line perpendicular to
them both; this line
Proof.
is
not minimal.
Let A* and
Ba
meet 7r.
By
.
hypothesis
Ax
to the circle at infinity.
be the points in which the given lines and the line A m Bn is not tangent Let Cn be the pole of the line A n Bn with
=,,
respect to
line
E w The required common intersecting
perpendicular
is
is
the
through Ca meeting the two given lines; this line unique and not minimal.
obviously
oi the pairs 01 vertices
meet in a point
U.
of the tetrahedron at the center of the circle
he line perpendicular to any face through the three vertices in
this face passes throiigh O.
114. Orthogonal
plane reflections.
is
DEFINITION.
homology
of
period two
whose
center, P,
a point at infinity polar in the absolute
polar system to the line at infinity of its plane of fixed points, IT, is called an orthogonal reflection in a plane or an orthogonal plane
re/lection or
a symmetry with respect
to
a plane, and may be denoted
by {wP}.* The plane of fixed points is called the plane of symmetry of any two figures which correspond in the homology. Since the center and the line at infinity of the plane of fixed points
of
to 2o>,
an orthogonal reflection we have

7.
in a
plane are pole and polar with respect
THEOREM
An
orthogonal re/lection in a plane
is
a transforma-
tion of the parabolic metric group.
By
a direct generalization of
:
Theorems 3 and
4,
Chap. IV, we obtain
the following
THEOREM
the
8.
(1)
If
TT
product {pR}
to TT
I,
{TTP}
and p are two parallel nonminimal planes, is a translation parallel to any line perto
pendicular
and
TT
plane perpendicular to I passing through the mid-point of the point pair in which TT and T (TT) meet I, then
minimal
line
p. (2) If T is a translation parallel any plane perpendicular to I, and p
a non-
the
T=
and
if cr is the plane perpendicular to
TT
I
passing through the mid-point

I,
of the pair in which
and T~ I (TT) meet
T=
(3)
{irP}
{o-S}.
I
translation parallel
to
a minimal line
is
a product of four
orthogonal plane reflections.
THEOREM
where
9.
A
,
product AÂ,,.^
x
.
of orthogonal plane
reflec-
tions is expressible in the

A.{,
form
A,'(
Al_i
A^T
or
T'AÂ^
Ai,
h! 2
are orthogonal plane A' n
reflections
whose planes of
and not
* In the rest of this chapter this notation will be used in the sense here defined 101. in the >nore general sense of
fixed points all contain an arbitrary point 0, ana is left invariant by AÂ,,^ translations. In case
T ana T are A p T and T
reduce
to
the identity.
Proof.
Let
' t
(i
1, 2,
n) denote the orthogonal plane reflection
whose plane
of fixed points is the

.. t
plane through
parallel to the
.
plane of fixed points of a translation. Hence A

(5)
A Then hy = T .A and
'
Theorem
8,
AA = T T
( f,
being
An A
n_1
...A 1
=TXT
_ 1 A:...T 1 Ai.
By the
is
generalization to space of Theorem 11, Cor. 2, Chap. Ill, if 2 2T', where T' is a any affine collineation and T a translation,
TS =
J.^.
l
translation.
By
repeated application of this proposition, (5) reduces to

A
A A
_A/ A'
!
A/T J.^1.
T'A'A' ...A' IV 1} J.\^ n _l

J.
"he product
s
AÂ,^
x,
since
reflections
A|-, it is left
invariant
and T' reduce
to the identity.
three orthogonal
y
plane rejections
J.
ordinary or ideal, the pua-ne reflection whose plane of fixed

7
ine
I,
points contains
I.
Proof. One of the chief results obtained in Chap. VIII, Vol. I, can be put in the following form * If T T T are harmonic homologies g
:
x,
2,
Leaving a conic invariant
and such that their centers are
collinear,
a harmonic homology leaving the conic invariant. For by Theorem 19 of that chapter, and its corollary, the product T 2T X is
TgT^Tj
is
T T,
8
,
where
is
harmonic homology whose
center and axis are polar with respect to the conic, the axis being concurrent with those of T x T 8 and T 8 ; and from T aTj= T 8T follows
,
T T& T 1 =T 3 TO T=T. 8 2 S
8 O
JL
Now
if
A A A
x,
2,
fixed points
meet
of
in
are orthogonal plane reflections whose planes of an ordinary line I their centers are collinear.
Hence they
are
effect in TT^ three
the poles
their
axes
harmonic homologies whose centers with respect to the absolute polar

collinear.
system and whose centers are harmonic homology in the plane
Hence
and
at infinity
its axis,
AgA^ effects a m M> passes
* Cf . the fine print in
108.
1/JLU.UUgJJ.
UlIC pUlLll;
ell/
JLLJUUUlty
Ui
fc.
OUUCb
l>
clUU
7fl m
'dTG
DOUD. 11X168 OI
fixed points of
v points of the plane TT containing I and are invariant. Hence A 2 1 effects a homology having the pole of 8 with respect to S*, as center. Since this homology is of period
all
AÂ^
two in
7r, it must be an orthogonal plane In case the planes of fixed points of x
reflection.
,
A A A
2,
are parallel
we have
by Theorem 8
(1)
that
AA
2
dicular to these planes,
i.e.
a translation parallel to a line perpenparallel to a nonminimal line. Hence by

is
Theorem 8
(2)
there exists an orthogonal plane reflection,
4,
such that
or
A A A 1= A
3
2
If {\L^} and {\L 2 } are two orthogonal plane reflections, and \( is any ordinary nonminimal plane in the same pencil with X 1 and \ there exists a plane \' and L' such that 2 and points L[ z z
COROLLAKY.
,
Proof. By the theorem, if L( is the point at infinity of a line perpendicular to \[, there exists an orthogonal plane reflection {Xâ} such that
and hence
{\LZ } {\LJ =
-
{\' Z L(}
{\{L[}.
and symmetries. Congruence. We may now 57, Chap. IV: generalize directly from DEFINITION. A product of an even number of orthogonal plane
115. Displacements
product of an displacement or rigid motion. odd number of orthogonal plane reflections is called a symmetry.
reflections is called a
THEOREM
the set
11.
The
set
of
all displacements
sdf-conjugate, subgroup of
DEFINITION.
the other
the parabolic
metric group
and symmetries is a and contains
of all displacements as a self-conjugate subgroup.
Two
figures
such that one can be transformed into

are said to be congruent.
by a displacement
Two
figures
such that one can be transformed into the other by a symmetry are said to be symmetric.
THEOREM
12.
If a figure
is is
is
to
congruent
.
to is
a figure
Z,
and
F
Z)
to
a figure F^, then F^
congruent congruent
with
&,
then F^
symmetric with
&t
then
is
If F^ is congruent 3 symmetric with Fg

.
.
F to F
If Fl
symmetric with
to
Z,
and and F9
X JlOJV^.U'.l-U'M.
J.17.
J.J.vy
y
reflections
iy
JL
riant
is
a product of two orthogonal plane
whose planes
fixed points contain 0. Proof.
Consider a product of four orthogonal plane

of fixed points pass
reflections
whose planes
through
0.
Let
be the line
of intersection of X,
I
and
of
let
X be
a plane containing
and X m that of X o and X and m, where in case I = m, X

,
tt
12
'
chosen so as not to be minimal.
If
is
nonminimal, by the
corollar
Theorem 10 there
exist orthogonal plane reflections

.
{pM}, {vN
such that
and
Hence
jy {XA} = {XL} {\L }.{\L } = {vN} F = {vN} {\L} {\L} {pM} = {vN}
i 8
-
ml*
plane through
I,
{Aj^J transforms X
to the other
minims
the other plane containing I and a tangent to th circle at infinity), and {\L2 } transforms this plane back to X. In life manner the product {X4 4 } {\L S } leaves X invariant. Hence X
(i.e.
left
left
invariant by F. On the other hand the line I is obviously nc invariant by F, and therefore F does not leave all points at infii
ity invariant.
most two tangents to the circle t most two minimal planes throug O invariant Let \'2 be any' plane of the bundle containing X2 and ) which doe: not meet in a line of an invariant minimal plane of 1 By the corollary of Theorem 10 there exists a plane Xg and points 1 and L' z such that
leaves at
infinity invariant,
Hence F
and thus leaves
at
L 2 }, {\L }-{\L2 }={\' a L' s }-{\'

f
and hence such that
Now let
and
X' the plane containing
be the line of intersection of Xi and X, that of X and X I and m. If X' were minimal it would,
argued above for X, be invariant under F, whereas X^ was so chose that I cannot be in such a plane. Hence the argument in the pn
vious paragraph can be applied to the last expression obtained for
* This case obviously does not arise in the real Euclidean geometry ( 116), that this paragraph may be omitted if one is interested only in that case. It needed, however, in complex, geometry.
115]
DISPLACEMENTS
299
any case, a product of four orthogonal plane reflections whose planes of fixed points pass through reduces to a product of two such reflections. By Theorem 9 any displacement leaving
Thus, in
invariant
reflections in planes
a product of an even number, say 2 n, of orthogonal through 0. This may be reduced to a product of two orthogonal reflections in planes through 1 applicaby n
is
tions of the result proved above.
COROLLARY.
Proof.
An
orthogonal plane reflection
is
not a displacement.
be a point of the plane of fixed points of an orthogonal plane reflection A. If were a displacement it would, by the theorem, be a product of two orthogonal plane reflections containing
Let
and hence could only have a single line of fixed points.

displacement which is a product of two orthogonal plane reflections whose planes of fixed points have an ordinary line I in common is called a rotation about I, and I is called the axis of
DEFINITION.
the rotation.
isotropic or
If
the axis
is
a minimal line the rotation
is
said to be
minimal.
THEOREM 14. The product of two orthogonal reflections in perpendicular planes is a rotation of period two. It transforms every point not on its axis to a point P' such that the axis is perpendicular to
the line
PP'
to its
the points
of
at the mid-point of the pair PP'. It leaves invariant its axis and the points in which any plane perpen-
dicular
axis meets the plane at infinity.
Its axis cannot be

,
minimal
line.
Proof. Consider any plane TT perpendicular to the planes of fixed and 2 By the points of the two orthogonal plane reflections l is nonmrnimal and first corollary of Theorem 5 the axis of 2 1
AA
hence
TT
2
is
nonmrnimal.
In
TT
the transformations effected
by
and
are orthogonal line reflections in the sense of Chap. IV,
and
their product is a point

plane.
From
this the
reflection (Theorem 5, Chap. IV) in the theorem follows in an obvious way.
DEFINITION.
dicular planes
reflection or
The product
is
called
of two orthogonal reflections in perpenan involutoric rotation or an orthogonal line

If
I
a half turn.
is its
axis
and
I'
the polar with respect
300
AFFINE AND EUCLIDEAN GEOMETRY

15.
[CHAF.VII
THEOREM
each point
DEFINITION.
The product of
the
orthogonal plane
reflections in three
to
three planes is
perpendicular planes is a transformation carrying a point such tliat the point of intersection of the the mid-point of the pair PP'. transformation of
this Hort is called
point () as center. said to 'be symmetric with respect

Proof.
a point reflection or symmetry with respect to the It is not a displacement. The points P and P' are
to 0.
In the plane at infinity the three orthogonal plane reflections effect the three harmonic hoinologies whose centers and axes
.ively opposite sides of a triangle.
The
oL.ts at infinity invariant.
It also leaves
two on the
uujjuo.
jL-LoiAuc
line of intersection
of the orthogonal plane reflecj.u
10
a,
iiuinujLug^ ui
period two with
as center
and
TTM
as plane of fixed points. It is not a displacement, since by Theorem 1 3 a displacement leaving invariant would have a line of fixed points
passing through 0.
THEOREM
TT
16.
The transformations
TT
effected in
a nonminimal plane
the,
by the displacements leaving
invariant constitute
group of
displacements and symmetries of the parabolic metric group whose absolute involution is that determined by E^ on the line at infinity of IT.
Proof.
1
Let
arbitrary point of
TT,
T" F (0) = 0, and hence, by Theorem T" 1 F leaves TT invariant.

It is
be any displacement leaving TT invariant, an and T the translation carrying to F (0). Then
13,
T" 1 F is
a rotation. Moreover,
obvious from the definition of a rotation that

its
it
can leave
it is
TT
invariant only in case
axis
is
perpendicular to
IT
or in case
of
1 period two and its axis is a line of TT. If T" F falls under the first of these cases, it effects a rotation in TT according to the definition of 1 rotation in Chap. IV, and thus F effects a displacement in TT. If T" F
falls
under the second of these cases

a
it effects,
and therefore
also
effects,
symmetry
in TT according to the definition in Chap. IY.
VUSJI VOOWItt/
r = TP
As
and
T=
P'T',
where T, T' are translations and P, P' rotations leaving

Proof.
invariant.
in the proof of the theorem above, let
be the translation
carrying
to
(0).
T~ r
x
is
a rotation, P.
to F-Ô), it ing P' and hence that
Then T" F (0) = and hence, by Theorem 13, Hence F = TP. If T' is the translation carryfollows in like manner that FT'(0) is a rotation
1 1
.
F= P'T'-
COROLLARY
3.
The transformations
effected
on a nonminimal
line
ly the displacements leaving
invariant constitute the group com-
all parabolic transformations and involutions leaving the point at infinity of p invariant.
posed of
EXERCISES
1.
Two
point pairs are congruent
if
they are symmetric.
2. The set of all point reflections and translations forms a group which, unlike the analogous group in the plane ( 45), is not a subgroup of the group of displacements. The product of two point reflections is a translation, and
any translation is expressible which is arbitrary.
as a product of
two point
reflections,
one of
3. Study the theory of congruence in a minimal plane. 4. A rotation leaves no point invariant which is not on its axis. It leaves invariant all planes perpendicular to its axis and no others unless it is of
period two,
when
it is
an orthogonal
line reflection.
116. Euclidean
geometry
of three dimensions.
The
last
theorem
may
be regarded as the fundamental theorem of the parabolic metric geometry in space, for by means of it all the results of
the two-dimensional parabolic metric geometry become immediately
applicable.
Suppose now that we consider a three-space satisfying Assumptions A, E, H, C, ft (or A, E, K), i.e. a real projective space. Suppose also that So, be taken to be an elliptic polar system,* i.e. the polar system
of an imaginary ellipse ( 79). Then in any plane the parabolic metric geometry reduces to the Euclidean geometry and the displacements which leave this plane invariant are Euclidean displacements.
*
The existence and properties
of
an
without recourse to imaginaries
(in fact,
elliptic polar system on the basis A, E, P,
may
be determined
in
89.
S), as
A
sions
set of
is
composed
assumptions for the Euclidean geometry of three dime 29 and 66. We have seen of I-XVI, giveii in
29 that I-IX are satisfied by a Euclidean space of three dirnensio] a consequence of Theorem 12, and Assumptions Assumption XI XII-XVI of Theorems 11 and 16. Hence in a real three-space,
i.s
Z*,
is an elliptic polar system the parabolic metric geometry is Euclidean geometry. The general remarks in 66 are applicable to the three-dimension
case as well as to the two-dimensional one.
66 that the congruence assumptions are no longer stric It was stated in independent when a full continuity assumption is added, because by int ducing ideal elements and an arbitrary 2^ (as in the present chapter) relation of congruence may be denned for which the statements in X-X This view is not accepted are theorems which can easily be proved.
certain well-known mathematicians,
definition of the absolute involution
It
who hold somehow
that the arbitrariness in

conceals a
new
assumptioi
if
may, therefore, be well to restate the matter here.f Assumptions I-IX, XVII are categorical for the Euclidean space two sets of objects [P] and [Q] satisfy the conditions laid down
;
points in the assumptions, there is a one-to-one reciprocal corresponder between [P] and [Q] such that the subsets called lines of [P] correspo to the subsets called lines of [Q]. Thus the internal structure of
is fully determined by Assumptions I-IX, XVII. T group leaving invariant the relations described in these assumptions is t affine group, and all the theorems of the affine geometry are consequent The latter may therefore be characterized as t of these assumptions.
Euclidean space
assumptions of affine geometry. Among the theorems of the

there
is
affine
geometry
is
one which states
tl
infinity of subgroups, each one conjugate to all the rest a such that the set of theorems belonging to it constitutes the Euclide Each of these groups is capable of being called the Euclide geometry.
an
group, and there

all of
is
them. The
set of
no theorem about one of them which is not true abc theorems stating relations invariant under any c
This
i.e. the
of these groups is the Euclidean geometry.
set of
theorems
is
same whichever Euclidean group be

unique body of theorems.
selected,
Euclidean geometry
ii
Each Euclidean group has

,/hich.
a self-conjugate subgroup of displacemei defines a relation called congruence having the properties stated
* Cf the remarks
.
peklie des Sc.

t
on a paper by the writer in the Math. Ill 1, 12.
article
by Enriques, Encyc
This discussion should be read in connection with the remarks on f oundatit

;
of geometry in the introduction to Vol. I and in 13 of this volume unction with the remarks on the geometry corresponding to a group,
also in
c<
34, 39, 1
Assumptions
tions
is
X-XVL
Moreover, any relation which
satisfies these
associated with a group of displacements
which
is self-conjugate
assumpunder
a Euclidean group.
Thus Assumptions X-XVI characterize the relation of congruence as comi.e. any relation satisfying these assumptions must be that determined by one of the infinitely many groups of displacements. The set of theorems about congruence is unique and is the Euclidean geometry. The relation between the affine geometry and the Euclidean geometry is
pletely as possible,
A
<
analogous to that between the Euclidean geometry and the geometry belonging to any non-self-conjugate subgroup of a Euclidean group. Consider, for invariant. example, -the subgroup obtained by leaving a particular point relation which is left invariant by this group may be defined as follows
:
DEFINITION.
Dist (OQ).
point
There
is
in the choice of
of congruence. Moreover, the geometry of nearness is just as truly a geometry as is the Euclidean geometry.* It would be easy to put down a set of assumptions (XVIII-JV) in terms of near regarded as an undefined relation, which would state the abstract properties of this relation, just as X-XVI state the abstract
P and Q an element of arbitrary choice in this definition, an absolute involution to define the notion
nearer than a point Q if and only if Dist are equally near if Dist (OP) = Dist (OQ).
P is
(OP)
just as there is
properties of congruence.
Another non-self-conjugate subgroup of the Euclidean group which gives an interesting geometry is the group leaving invariant a line and a plane on this line. In terms of this group the notions of forward and lackward and up and down can be defined, and the geometry corresponding to this
rise to
group
is
a set of propositions
embodying the abstract theory
of this set of
relations.
It is a theorem of Euclidean geometry that the Euclidean group has as it is subgroups with the properties involved in these geometries, just a theorem of affine geometry that the affine group has Euclidean subgroups has affine the and a theorem, of projective geometry that projective group
subgroups.
Assumptions I-IX,
XVII have
a different r61e from
X-XVI
or
XVHI-JV,
assumptions is logically arbitrary. is not arbitrary; it must correspond to a properly chosen group of permutations of the objects determined by I-IX, XVII. When independence proofs are given for Assumptions X-XVI, it is done by giving new interpretations " or " line." not to " to the term "
congruence,"
point
even possible to give a psychological significance to this geometry. The normal individual has a certain place, say home, in terms of nearness to which In astronomy is the central point of home. other places are thought of here ctdvc- ova TormWipH fts npar nr tlip. r.nnt.rarv. according to their distance from the
* It
is
;
in that they determine the set of objects (points and lines, etc.) which are choice of these assumptions presupposed by all the other assumptions. The " as X-XVI The choice of such sets of "
304
The
AFEINE AND EUCLIDEAN GEOMETRY
[CHAP.VII
point of view of the writer is that if X-XVI or XVTII-.ZV are to be regarded as independent assumptions, their independence is of a lower grade than that of I-IX, XVII. They constitute a definition by postulates of a
relation (congruence or nearness)
fully determined.
among
objects (points, lines, etc.) already
Their significance is that they characterize that subset of the theorems deducible from I-IX, XVII which corresponds to any Euclidean group and which therefore is the Euclidean geometry.
EXERCISES
Develop the geometry corresponding to some non-self-conjugate subgroup of the Euclidean group. Determine a set of mutually independent assumptions characterizing this geometry. 2. The identity is the only transformation of the Euclidean group which leaves fixed two points A and B and two rays (cf definition in 16) A C and
1.
.
AD orthogonal to
3. If a
lines,
each other and to the line
AB.
and
is
b are
there
any two rays having a common origin, 0, and on different a unique orthogonal line reflection and a unique orthogonal
plane reflection transforming a into b. are any four points no three of which are collinear, there 4. If A, B, C, invariant and transforming C exists a unique rotation leaving the line
AB
into a point of the plane

5.
ABD on the
same
side of
AB
with D.
Any transformation of the Euclidean group which leaves a line pointwise invariant and preserves sense is a rotation.
6.
Any
wise invariant and alters sense
transformation of the Euclidean group which leaves a line pointis an orthogonal reflection in a plane
containing this line.

7. There is one and only one displacement which transforms three mutually orthogonal rays OA, OB, OC into three mutually orthogonal rays O'A', O'B', O'C', provided that S (OABC) =
*117. Generalization to n dimensions. The discussion of the Euclid-
ean and affine geometries in 111-116 is so arranged that it will generalize at once to any number of dimensions. It is recommended
to the reader to carry out this generalization in detail, at least in the
four-dimensional case.
The elementary theorems of alignment for four dimensions are 12, Vol. I. The definition of a Euclidean four-space is given in Riven in 28, Vol. II. The generalization of S 111 is obvious on
system
three-dimensional polar system may be defined as the polar of a proper or improper regulus (Chap. XI, Vol. I cf. also
;
100-108, Vol.
II), or it
may
be studied ab initio by generalizing
Chap. X, Vol. I. The notion of perpendicular lines, planes, and threespaces then follows at once and also the theorems generalizing those of 113. An orthogonal reflection in an S is next defined as a projec3
tive collineation of period two, leaving invariant a point
P at
infinity
and each point
a three-space whose plane at infinity is polar to in the absolute polar system. All the theorems of 114, 115 up to Theorem 13 then generalize at once. Theorems 13-15 must be modi'
of
fied,
in view of the fact that there are
dimensional displacements leaving a point invariant. holds unchanged.

Finally,
it
more than one type of fourTheorem 16
and an
elliptic
can be proved as in 116 that in case of a real space polar system the parabolic metric geometry satisfies
a set of axioms for Euclidean

set differs
geometry of four dimensions. This from the one used above, in that VIII is replaced by
VIII'. If A, B, C, are four noncoplanar points, there exists a not in the same S 3 with A, B, C, D, and such that every point point is in the same S with A, B, C, D, E, 4
The introduction of nonhomogeneous coordinates in a space of n dimensions may be made by direct generalizations of 69, Vol. I. The formulas for the affine group, the group of translations, the Euclidean group, and the group of displacements are then easily seen
to be identical
summations from
summations from
to a
with those given in the sections below, except that the or 1 to 3 must in each case be replaced by
or 1 to n.
and Euclidean groups. With respect nonhomogeneous coordinate system in which TT^ is the singular plane, the affine group is evidently the set of all projectivities of
118. Equations of the affine
the form
x'
(6)
y'
>
= a u x + a u y + a nz + a = a z + a y + a^z + a = a x + a y + a^z + a
sl
sl
10 , 20 , so ,
zz
zz
ft
where
A=
a u a i2 a a a
=f=
0,
and the variables and
coefficients
are elements of the geometric
number system.
at infinity is represented
In the system of homogeneous plane coordinates in which the plane by [1, 0, 0, 0], this group takes the form
'
In an ordered space the

all
affine
transformations for which

31.
2
group has a subgroup consisting of is positive. This group has been

for
considered in
It also has obvious subgroups consisting of all
transformations for which
A = 1 and
which
A = 1.
The equations
ad
of a translation parallel to the a>axis are evidently
' translax z, and similar expressions represent a a, y' y, tion parallel to any other axis. Hence by the corollary of Theorem 3 the equations of the group of translations are
/= + a,
03
(8)
/
z'
= + =z+
2/
&,
c.
If
z
the coordinates are so chosen that the planes x 0, y = 0, are mutually orthogonal, the equations of the circle at infinity
in terms of the corresponding
homogeneous coordinates
cap8
2
are
axl
+ Hx% +
=0,
0.
These are reducible by the transformation

(9)
#o
^<
= v/ ai
xz
= Vbxz
*
to
(10)
as?-!-!
+<=
<),
= 0.
In the real geometry a, 6, and the transformation (9)
c are positive if the polar system is elliptic ( 85), carries real points to real points. The formulas in the ones are the only present section in which irrational expressions (9) appear. Hence the rest of the discussion holds for any space satisfying
H In any such space it is easily seen that (10) reprewhose polar system may be taken as ], but it does not follow, as in the real case, that any improper conic can be reduced to this form.
Assumptions A, E, P,
.
sents a conic
The
situation here is entirely analogous to that obtaining in
62.
the planes tangent to the circle at infinity (10) satisfy the equation
(11)
u*
+ u* + u*=Q.
Any
is
plane
(12)
+ w/ + wy + u/ =
zo
the transform under a collineation of the form (6) of the plane

(13)
(u -f
Vl + a
2a
+ (Vi + a
Hence
(14)
(11)
is
wg +
V
18
u2 + a 30w 3 )
3)
+ (a uw +
1
a 2L
4- (a 18
^+
23
+ +a u
2 ss
31
^3 ) a
3)
2=0.
the transform of
+<+ <+ + + + 2 (a u a + + a aj ^tt, + 2 + 2 (a a + a + a a up, = 0.

2
22 21
12
O
22
O<+
(
(s i +
2
2
32
O<
^^
8
81
+a
a l3 a 33 ) a l2 82 +
sl
22
32
zs
sa )
In order that
(15)
(11)
and (14)
shall represent the

-f
same locus, we must have

2
82
< + < + G& = < < + < = < +
+ <,
These conditions are equivalent to the equation

>
(cf.
95, Chap. X,
/a n a 12 a ia \
(16)
Ia 21 a M a 28
\a 81 a 82 a s8 /
-U
/a n a al a 81 \ a 82 12 22 \a 18 28 a 83 /
/p
0\
=10
\0
p Ol,
p/
where p
If the
= a^ + a? +
2
iV
matrix (^ n a 22 a 33 )
=A be
I,
interpreted as the matrix of a planar

this states that the product of the
collineation, as in
95, Vol.
collineation
is
by the collineation represented by the transposed matrix the identity. Hence the product of the two matrices in the reverse order is a matrix representing the identity. This means that
<+
and
a n a 12
2 3i
+ a = a u + < + <4 = < + < + <> = na + a a + a a M a +a = a a +a a 4-a =0.

si
21
22
13
2l
23
31
12
18
22
23
32
83
Since the determinants of a matrix and of

equal,
its
transposed matrix are
we have
A = p=
2
+ ai+ <) = K+ <+ <)

8
8
.
DEFINITION.
is
1 denoted by A"
A
.
matrix such that
its
the identical matrix
(95,
Vol.
I) is
product by a given matrix called the inverse of A and

is
A
is
square matrix whose transposed matrix
equal
to its inverse is called orthogonal.
A linear
a zzy
transformation,
(17)
y'
2'
= =
V+ +
81
+a
-f-
28 2,
8S 2,
a^y
whose matrix The results

part as follows
(a n a., 2 a a3 )
at
:
is
orthogonal,
is
said to be orthogonal.
which we have arrived may now be expressed in
THEOREM
can
17.
be written
The transformations of the parabolic metric group in the form

x'
(18)
y'
==
= p (a u x + a p (a x +
zl
l2
22
+a +a
where the matrix (^ n
22
a 83 )
^s
orthogonal.
From
the
:
form
of
these
equations
we
obtain
the following
corollaries
COROLLARY
the product
1. Any transformation (18) of the Euclidean group is of an orthogonal transformation, a translation, and a
homology of the form

(19)
y'
= px, = py> z' = pz.

f
COROLLARY 2. liomology (19) is commutative with tion leaving the origin invariant.
Since an orthogonal matrix
is
any
collinea-
any matrix satisfying
(16) with
= 1, we
have
is is
orthogonal. 1 or 1.
COROLLARY 3. The product of two orthogonal transformations The determinant of an orthogonal transformation
(20)
Ay = Aa..,
is
(t
= 1,
A
i}
2, 3
j=
1, 2, 3)
where
ê determinant
of the matrix and
the cofactor
of
a.j.
The matrix of an orthogonal transformation of period two own inverse and hence its own transposed. Hence
COROLLARY 5. An orthogonal transformation and only if a~ = aj{
.
is its
is
of period two if
The double points

satisfy the equations
of
any orthogonal transformation
(17)
must
(21)
The determinant
of the coefficients of these equations is

(
A = A ~ Ai + A v + As) +
But since the transformation
determinant
of (21)
is
K+
22
- 1.
.
orthogonal,
A H = Aau Hence
the
reduces to
(1
is
Z>i
Another determinant which

orthogonal transformations
is
of
importance in the theory of
that of the equations
(22)
point satisfying these equations is transformed into its symmetpoint with respect to the origin. The orthogonal transformation therefore transforms the line joining these points into itself and
Any
ric
effects
an involution with the origin as center on this

is
line.
The
determinant of the equations (22)
which reduces
to
Let us
shall denote
least
consider an orthogonal transformation (17) which \v 1 for 2, Z> = 0, and hence there is by 2. If 2 one point which is carried by 2 into its symmetric point wit
now
A=
the origin perpendicular t respect to the origin. The plane through the line joining these points is left invariant by 2. On the oth<
hand,
(23)
!=
unless
and hence
element a u
2
of
leaves
+*-!, *!!+ no other point than the
origin invariant unle:
(23) is satisfied.
Suppose now that (23) is the main diagonal of D l is
satisfied.
cofactor of
where
i 3=
3=
Ic
3=
i.
By
(20) this reduces to
-K +
which vanishes. The cofactor
of
aM
+a +1
. B)
=
an element a~ (i
/) of
D^ is
and by
(20) this vanishes
4,+ a,,, when A =
1.
Thus we have that
A=
and (23) is satisfied, 2 has a plane of fixed points. Since transforms one point into its symmetric point with respect to the 01 gin, it must be an orthogonal plane reflection. Thus we have prove<
1
THEOREM 18. An orthogonal transformation for which A = always has an invariant plane. It either leaves no point except origin invariant or it is an orthogonal plane reflection. The latt case occurs if and only if & 4- & a 88 1. U 22
t,
By comparison with
COROLLA? Y.
Corollary 5 above
we have
1 is c transformation for which orthogonal plane re/lection if and only if # 18 =fl 21 a 23 = a 82 an d a ls = a

,
An orthogonal
A=
>
Let us
now
A = l.
In this case
consider an orthogonal transformation 2 for whi< -\= 0, and hence there is always a line of fixi
points passing through the origin.

reflection containing a line
Let
be an orthogonal
X
plai
of fixed points of 2.
orthogonal transformation for which other fixed points than the origin.
it
A=
By
Then 2A is and for which there
the last theorem, therefoi

.
is
2=
an orthogonal plane reflection A 2 A r We therefore have

19.
From 2 A
=A
follow
THEOREM
An
orthogonal transformation for which
A=l
is
COROLLARY
a symmetry.
1.
An
orthogonal transformation for which
A=
1 is
Any
transformation (18) for which p
=1
is
a product of an orthog-
onal transformation and a translation.
It is therefore either a dis-
for
placement or a symmetry. By Theorem 16, Cor. 1, a homology (19) which p 2 =f= 1 is not a displacement or a symmetry. Hence we have
COROLLARY
is the
2.
The subgroup of
(18)
for which
/>
=!
and
2
A= l
is
group of displacements. COROLLARY 3. The subgroup
the
of (18) for which p group of displacements and symmetries.
=l
and
A =1
is
The coordinate system which has been employed above

that the planes the displacement
leaves
(0,
such
x=
Q,
y=
0, 2
are mutually orthogonal. Moreover,
x/==y>
^^
z>
= x>
(1, 0, 0) to
(0,
1, 0)
(0,
0, 0)
invariant and transforms
and
1, 0)
to (0, 0, 1).
Hence the
pairs
(0, 0, 0)
(1, 0, 0),
(0, 0, 0)
(0, I, 0), and (0, 0, 0) (0, 0, 1) are congruent. these conditions are said to be rectangular.
Coordinates satisfying
EXERCISES
1.
The group
of displacements and symmetries leaves the quadratic form
u?
absolutely invariant.
2.
if
w|
4-
2 ws
Two
point pairs (a,

if
b,
and only
3.
(a
a')
(6
c)(a', V, c')
2
and
2
&')
+ (c -
(x, y, z)
c')
(x
(x
2
y', z')
a:')
(y
- y'}* +
are congruent
(z
- z')
Two
planes
^ ^ ^
+ v:x +
and only if
are orthogonal
4.
if
Three planes
^+^ ^
f
u
+ +%= v z y + vs z + v = u 1 v l + u z v 2 + u s v s = 0.
,
^=
Qj
(
(i
the coefficients being such that
+ u^ =1,
=^ = 1,
2> 2,
g) 3)
are mutually perpendicular if and only if the matrix (w 11 M 22 w 33 ) is orthogonal. 5. The three ordered triads of numbers (a^, a,-.,, i3 ), i 1, 2, 3, are direction cosines of mutually perpendicular vectors
is
if
and only
if
the matrix
n a 2 2 a 88)
orthogonal.
119. Distance,
(
area, volume,
angular measure.
The
definition
two points extends without modification The distance between a point to the three-dimensional case. and the point P in which and a nlane TT is the distance between
67) of distance between
TT
ine distance is met by the line through u perpendicular to TT. are the points between two lines l^ is Diet (P^), where P and P 2 in which the common intersecting perpendicular line meets ^ and
1
2
respectively.
If
the notion of equivalence of ordered point triads
112) be
extended by regarding two ordered triads as equivalent whenever they are congruent, it is obvious that any triad is equivalent to triads in any plane whatever and not merely, as in 112, to triads in a
system of parallel planes. Moreover, if ABC are noncollinear points such that AB is congruent to AC, the ordered triad ABC is congruent
and therefore equivalent
to the ordered triad
ACB. Hence
ABC^ BCA sc CAB ÂCB^ CBA^BAC,

i.e.
according to the extended definition, any ordered triad
is
equivalent
49) can-
to
any permutation
Since
of itself.
m(ABC} =
m(ACB), the
definition of
measure
not be extended to correspond to the new conception of equivalence. On the other hand, the notion of area ( 68) of a triangle is directly
The situation here is entirely analogous to that described 67 with regard to the measure of a vector and the distance between two points. The formal definition may be made as follows
applicable.
in
DEFINITION. Let
is
such that the
OPQ be a triangle lines OP and OQ are
(called the unit triangle}
which
OP
and
OQ
are congruent to the unit of distance.
orthogonal and the point pairs Then if A'B'C' is

to
a triangle coplanar with
OPQ
and congruent
ABC,
the positive
number
%\m(A'B'C>}\
= a(ABC}>
ABC.
where m (A'B
C')
is
the measure
49) of the ordered triad A'B'C' relacalled the area of the triangle
of
tive to the ordered triad
OPQ,
is
The
definition of the
measure
an ordered tetrad and
of the vol-
ume
of a tetrahedron
may be
taken from
the unit tetrad
OPQR is such
112, with the proviso that that the lines OP, OQ, OR are mutually
orthogonal and the point pairs OP, OQ,

distance.
OR
congruent to the unit of

be taken over
The
from
definition of the
69.
measure
of angle
may
literally
Since, however,
any symmetry in a plane can be
effected
by
a three-dimensional displacement, the indetermination in the measis
ure of an angle the measure kir
/3,
such that any angle whose measure is ft also has where k is a positive or negative integer. The
measure
=s ]3
of
an angle
< TT
or
may
=
therefore be subjected to the condition

vr/2.
7T/2
< /3
is
DEFINITION. The angular measure of a pair of intersecting lines ab the smallest value between and 2 TT, inclusive, of the measures
of the four angles
It is denoted
by m(ab).
4o. 1 6 1 formed by a ray <x l of a and a ray l of 6. If a and b do not intersect, m(ab) denotes
~b
M (a'b), where a' is a line having a point in common with b and parallel
The angular measure of two planes TT, TT' is the angular measure two lines I, I' perpendicular to TT and TT' respectively. The following statements are easily proved and will be left to the reader as exercises (cf. In the case where a and b do not inter72)
to a.
of
:
m (ab) is independent of the choice of Although in Euclidean plane geometry S m (ab) < in the three-dimensional case
sect,
the value of
a'.
TT,
If
^ and
is
^ and \
m(ab)
are the
any two lines parallel to a and b respectively, and minimal lines through the intersection of ^ and lz the smaller of the two numbers
/
2
are
^-^log
which
.(*, \\}
and
=-|logT^(^
2,
ijj,
that determination of the logarithm in each case being chosen for
S X < TT and S 6 z < TT. The numbers which we have been defining in
this section are
some
group of displacements. The algebraic formulas for these invariants and some others are stated in the exercises below. In every case the radical sign indicates a positive
of the simplest absolute invariants of the
root.
By
the angle between two vectors

<
OA
and
OB
is
meant the
measure
of
A OB.
The orthogonal projection of a set of points [P] on a plane TT is the which the lines perpendicular to TT through the points P meet TT. The orthogonal projection of a set of points [P] on a line I is the set of points in which the planes perpendicular to I through the
set of points in
points
P
z,
meet
I.
The
exercises refer to four distinct noncoplanar points P^

y,, z2 ),
= (xv yv zj,
which
are
% = (x
PZ=(XZ,
2/ 3 ,
%), P<
= (XI,
0,
y,, 24 ),
no two
is
of
collinear with the origin.

0, P, Q,
The coordinate system

(0,
rectangular, and
denote the points

ae
-Jn
0),
(1, 0, 0), (0, 1, 0), (0, 0, 1)
Taar*art.itrol-*r
1 1 9.
EXERCISES
Dist (/' 1 2 ) = "^(#1 ~ -''a) 2 + (?/i ~ 2/a) 2 "^ ( 2 i ~~ z ^f'2. The cosines of the angles between a vector OP^ and the respectively are
1.
a:-,
?/-,
and z-axes
?/
a;
yx
the vector
are
These are referred

Dist
>
to
as the
direction cosines of
OP r
If r
(/-
/J 2 ),
the direction cosines of the vector
PiP2
r 3.
The equation
The
of a plane perpendicular to the line
OP l
is
V+
4.
ci
2/i2/
+ +
zi2
= k+ ^8?, +
yz
distance from the point P^ to the plane ux
is
/3//!
Yz t
-8
on the
line C
i
Va2 + ^ +
5. If
f
Pz
\,
Qj
is
the orthogonal projection of
then
is
in case
Dist(OQ 1 ) in case Q 1 and P t are on the same Q l and P i are not on the same side of
X-^XQ ~r
siae of 0,
aid
Dist(OQ1 )
y^'i
T*
^1 ô
~~"
-^^^t
\s i
-t
X/iSt C
-i"
COS
>O /
-t
Lf*n*
6.
m(P 1 P8 P 8 P 4 ) =
and
tance between the lines
Dist(P 1 P2 ) P X P2 and
PP
Dist(P 3 P 4 ) r- sin ft where r is the disand 6 the angle between the vectors S 4
,
,
P X P2
P3 P4
7. If
denotes the measure of
^.P 1 OP2 and
I,
a vector
OK
perpendicular to the plane
OP t Pz and
(OP2 )
.
m, n the direction cosines of such that SÔP^^K)

sin (9
sin
I,
2/2
= Dist (OP X )
Dist
= Dist(OP
x)
Dist(OP 2 )
Dist (OP2 )
m,
- Dist (OPi)
8.
sin 9
n*
118, the angle
ft v J3 2 , /? s )
respect to the coordinate system employed in between two lines which meet TT*, in (0, a x a2 a3 ) and (0,
, ,
With
is
+ a!)G3+# + #)
9. If four planes a,
ft,
y, 8
meet on a
sin
line,
-D
where (ay) denotes the angular measure of the ordered pair
of planes ay.
120.
set of all points [P]
The sphere and other quadrics. DEFINITION. A sphere is the such that the point pairs OP, where is a fixed
V In case the line OP congruent to a fixed point pair OP minimal, the sphere is said to be degenerate otherwise it is nondeis called the center of the sphere. generate. The point
point, are all
is
;
By comparison with
the definition in
60
it is
clear that
any
sec-
tion of a nondegenerate sphere by a nonminimal plane is a circle. In case the circle at infinity exists, two perpendicular sections Cf and (722
of a sphere
S by nonminimal By
105, Vol.
I,
infinity three conic sections intersecting
planes constitute with the circle at one another in pairs of disis
tinct points.
there
containing them.
nonminimal plane
one and but one quadric surface TT through the center of the
two points
infinity.
sphere meets this quadric in a conic section which contains at least 2 of the circles Of and (72 and two points of the circle at
This conic
is
sphere
S which
are in
therefore a circle containing the points of the TT. Hence the sphere S is identical with the
2
set of all ordinary points of the quadric surface containing Of, <72 , and is the center of each circle in which S the circle at infinity. Since is the pole of the plane is met by a nonminimal plane through 0,
at infinity
with regard to the quadric.
Since a circle in a nonof a
minimal plane contains the ordinary points

it
follows that the quadric surface is nondegenerate, which contains two proper or improper reguli.
nondegenerate conic, i.e. is a quadric
In case the circle at infinity does not exist, improper elements may be adjoined as explained in 85, Vol. I, so that the circle at infinity exists in the resulting improper space. The argument in the paragraph above thus applies to any space whatever which Thus we have Assumptions A, E, P, Q
satisfies
THEOREM 20. nondegenerate sphere consists of the ordinary points 2 of a nondegenerate quadric surface S such that all pairs of points in 2 the plane at infinity conjugate with regard to S are conjugate with
The center of the sphere to the absolute polar system. of the plane at infinity relative to this quadric.
regard
is the
pole
Comparing the definition above with Theorem

COEOLLARY.
7,
Chap. IV,
we have
as center consists degenerate sphere with a point of all ordinary points on the cone of minimal lines through 0, except
a degenerate circle in the plane been defined in the same way that is denned above, it would have been found to consist of on only one minimal line through 0, since in the plane the group of points displacements leaves each minimal line invariant.
Had
a degenerate sphere
in a
The Euclidean classification of quadric surfaces may now be made manner entirely analogous to the Euclidean classification of conic
sections in Chap. V.
(
After completing the projective classification 103) and the affine classification ( 111, Ex. 2) and obtaining the properties of diameters and diametral planes, the principal remaining
problem
line
is that of determiniug the axes, an axis being defined as a through the center of the quadric perpendicular to its conjugate
planes.
A line
line
jpoo
2
and a plane
If If
TT
/ I
meet the plane

and TT and TT
at infinity in a point
and a
respectively.
are perpendicular,
Lx
and p x are polar
with respect to 5L.

Q'
,
are conjugate with regard to a quadric
Lx
and
p*, are polar

z
degenerate) in which Q axes is reduced to that of finding the points which have the same polar lines with respect to two conies. This problem has been treated
in
with respect to the conic (real, imaginary, or meets ir^. Hence the problem of finding the
101, Vol.
I,
for the case
where both conies
are nondegenerate.
In general the two conies have one and but one common self-polar triangle. Hence, in general, a quadric surface has three axes which are mutually orthogonal. The determination of the other cases which may
arise is a
application of
problem (Ex. 5, below) requiring a comparatively simple methods and theorems which we have already explained.
The classification of point quadrics includes that of cones and conic sections, the properties of cones and conies in three-dimensional Euclidean geometry being by no means dual to each other. In connection with this
it is of
interest to prove the following theorem,

of a conic.
is
which
embodies perhaps the oldest definition
THEOREM
circle.
21.
Any
2
nondegenerate real conic
perspective with a
Proof.
Let
(7
be a given conic and JT 2 a

2
circle in a different plane

.
having a
common tangent and point of contact with C 2 By Theorem 1 1,

I, (7
Chap. VIII, Vol.
and
are sections of the
same cone.
COROLLARY.
point.
Any
cone of lines is a projection of a circle
from a
EXERCISES
1.
The equation
of a sphere of center (a,
b,
c) in
2
rectangular coordinates
ig
(x
a)
(y
- bf +
(z
c)
k.
2. The set of points on the lines of intersection of homologous planes in the corresponding pencils, x z + V^Tar, = X(a: + xj,
X
is
~ X = ^ \X ~
l
<1
g),
a sphere.
3.
a projection of a circle from a point from which the extremities of any diameter are projected by a pair of perpendicular lines. Any conic may be regarded as the plane section of a right circular cone.
right circular
cone
is
*4. Develop the theory of stereographic projection of a sphere

(cf.
on a plane
100). *5. Classify the quadric surfaces from the point of view of Euclidean geometry. Having made the classification geometrically, find normal forms for the equations of the quadrics of the different classes and the criteria to
determine to which class a given quadric belongs.
This
of
is
analogous to the
of
work
in Chap. V.
*6. Classify the linear complexes
from the point
view
Euclidean
geometry.
*7. Starting with a definition of an inversion with respect to a sphere analogous to that of an inversion with respect to a circle ( 71), develop the theory of the inversion group of three-dimensions. This should be done both in the
real
and complex cases and the real and complex inversion spaces studied.
121. Resolution of a displacement into orthogonal line reflections.

properties of the group of displacements are closely bound up with the theorem that any displacement is a product of two orthogonal line reflections. In proving this theorem we shall place no restriction
The
on the absolute polar system except that it be nondegenerate, and We are shall base our reasoning on Assumptions A, E, Q only. therefore obliged to consider transformations which do not exist in
,
the Euclidean geometry,
namely those with minimal
lines as axes.
DEFINITION. The line at infinity polar in 2*. to the center of a, translation is called the axis of the translation. If the axis is tangent to
the circle at infinity, the translation
is
said to be isotropic or
minimal.
THEOREM 22. A product of two orthogonal line reflections whose axes I and m are parallel is a translation whose axis is the line at I and m and parinfinity of any plane perpendicular to the plane of allel to 1. Conversely, let T lie any translation and I any nonminimal
318
AFF1NE AND EUCLIDEAN GEOMETRY

L
is
[CHAP.VII
which of every pair of points, L and T (L), for the pole in 2*, of the point at infinity of
(,,
on
I,
and
if
I'
is
Proof.
If
the axes
and
of
two orthogonal
ir x
line reflections {II'}
and [mm'} are
parallel,
they meet
in a point
R. Each
of
the
whose orthogonal line reflections effects in 7r w a harmonic homology axis I is the polar of x in S,. Hence the product leaves all points the product {ml } {II'} at infinity invariant. In the plane of I and
effects a
planar translation parallel to any line perpendicular to I. Therefore the product {ml } {II'} is a translation in space parallel
1
to this line.
the line at infinity of any plane perpendicular to the plane of I and m and parallel to I. The converse follows directly in the same manner as the analogous
Its axis, therefore, is
statement in Theorem
4,
Chap. IV.
is
THEOREM
23.
Any
displacement
a product of two orthogonal
line reflections.
Proof.
In case the displacement, which
we
shall denote
by A,
is
translation the theorem reduces to

is
Theorem
22.
In any other case

1 6, Cor. 2),
a product of a rotation
and a translation (Theorem

{Sir} -T, he the identity.
i.e.
A ={.?}.
where
is
a translation which
may
Thus
effects
in the plane at infinity a product of two harmonic homologies centers and axes are R, p x and ,, r x respectively, where p x
line at infinity of
IT
whose
is
the
and r, that
of p.
Let
be an arbitrary ordinary point and Q' = A(Q).
Let
be the
line of intersection of the planes joining
to
p a and
Q' to r n
These
;
planes cannot be parallel, because cannot contain or JK M> because
p*,
is
and r x do not coincide

not on
and
I
.
p n and R x
is
not on rn
be an ordinary point of I such that neither of the lines OQ and OQ' contains or x (If the lines OQ and OQ' coincide, they coincide with 1.) Let P be the mid-point of OQ, the mid-point of
Let
OQ', and let
p and
r be the lines
PR
and
RRn
respectively.
Then
p and
r are such that there exist orthogonal line reflections
and {rrx } such that
Moreover, {pp x }
{rr^} effects
the
inverse
of
the
-
transformation
effected in the plane at infinity
by A.
invariant all points at infinity as
Hence {pp*} {rr x } well as Q, and hence
leaves
A = 1,
or
very easy to enumerate the possible types of displacedisplacement A being expressed in the form {II } (ram'}, * the following cases can arise
It is
now
ments.
/ and in intersect in an ordinary point 0. A is a rotathe product of the orthogonal reflections in the planes respectively at 0. Two subcases must be perpendicular to I and
I.
The
lines
is
tion
which
distinguished
(a)
The plane containing
and
m, is
not minimal.
I
A
is
is
a rotation
about the
(b)
common
intersecting perpendicular of
I
and m.
The plane containing
and
is
minimal.
an isotropic
of
rotation about the line joining
to the point in
which the plane
and
in touches the circle at infinity.
It
evidently effects a parabolic

its
transformation, in the pencil of planes meeting an elation in the fixed plane on the axis.
II.
axis
and
also effects
The
lines
and
are parallel.
If
we denote
their
common
point at infinity by Ii, and its polar line with respect to 2, by p* Theorem 22 states that A is a translation whose axis is the line polar
in
tt
to the point in
is
point
(a)
(6)
meets which the plane of I and the center of the translation. Two cases arise
p a The
.
latter
The axis of the translation is not tangent to the circle at infinity. The axis of the translation is tangent to the circle at infinity,
is isotropic.
and the translation

III.
(a)
The The
lines
I
and
and
lines
m do not m have a
intersect.
Again two cases
arise
common
intersecting perpendicular
line
to
a (Theorem 6) which is not minimal. Let p be the line parallel and passing through the point of intersection of I with a. Then
Thus
1
A = {M'} {pp } {pp } {mm A is the product of a rotation {II } {pp'} about a by a translation
r
1 -
}.
{PP
'
'i {r
m '}
parallel to a.
I
(b)
The
lines
and
m have
no common intersecting perpendicular.
In this case they are (Theorem 6) both parallel to the same minimal plane a. Let a*, be the line at infinity of a, and A M its point of contact with the circle at infinity. Then I and pass through points of a m
distinct
from each other and from
A*,,
and V and m' pass through
Ax
Type III ( 40, Vol. I) in the plane at infinity, with A a as its fixed point and a, as its fixed line. It also effects a parabolic transformation in the pencil of planes with aM as axis. Thus its only fixed point is A x its only fixed line and its
,
Therefore
effects a transformation of
only fixed plane DEFINITION.
77%.,.
displacement of Type Ilia,
i.e.
a product of a non-
isotropic rotation
by a translation
parallel to its axis, is called a twist or
screw motion.
The
axis of the rotation is called the axis of the twist.
THEOREM
24.
displacement which interchanges two distinct ordi-
nary points is an orthogonal line reflection. Proof. Denote the given points by A and B. The given displacement A cannot be a translation, because a translation carrying a
mid-point of the pair A C. Nor can A be a twist or a transformation of Type Illb, because either of these types effects the same transformation as a translation on a certain system of parallel planes, and hence
iso tropic rotation,
no point can be transformed involutorically. because in this case it would
And A
cannot be an
effect a parabolic trans-
formation in the planes on its axis and an elation in the one fixed plane on the axis. Hence A is a nonisotropic rotation. By reference to 1 1 5 it follows that A must be an orthogonal line reflection.
THEOREM
25.
If
A A A
1?
2,
are three orthogonal line reflections
whose axes are parallel or have a

I,
the product
to
AAA
3
parallel
the other
common intersecting perpendicular an orthogonal line reflection whose axis is three axes in the first case and is an interis
secting perpendicular
of
in the second
case.
In case the three axes are parallel, by Theorem 22, A : is 2 another a translation which is also expressible as the product of 8 by
Proof.
4,
so that
and hence
A A A =A
8 2
1
intersecting perpendicular I, the orthogonal line reflections effect involutions on I having the point at infinity of I as a common double point. Hence ( 108, Theorem 42)
the product
In case the three axes have a
common
AAA
3
effects
an involution on I whose double points and an ordinary point P. Hence, by Theorem

.
are 24,
AgAgAj^
is
by A 4
it is
/
an orthogonal line reflection A4 Since P is left invariant on the axis of A4 and this axis is perpendicular to I because
;
leaves
invariant.
EXERCISE
The product
an
isotropic rotation by a translation parallel to its axis is isotropic rotation about an axis in the same minimal plane.
of
an
122. Rotation, translation, twist. Let us now require the absolute polar system to be elliptic, as in the real Euclidean geometry. In this case there are no minimal lines, and hence the possible types of dis-
placement are reduced to
la, Ila, Ilia.
Thus we have
THEOEEM
26.
In
case the absolute, polar system is elliptic
any
dis-
With, this assumption about the absolute polar system

a particularly simple method for the combination
-
we have
which depends on Theorem 25. Suppose that we two displacements {l.2 l' z} t} {ljl[} and {lj' {l^}. Let a be a common intersecting perpendicular of ^ and lz and b of ls and l^, and let m be a common intersecting perpendicular of a and 6. Then the product A
,
displacements wish to combine
of
of
the two displacements
satisfies
the following conditions
By
the theorem just proved there exist

1
two orthogonal line
reflections
(PP'}>
{9.9.
suca
(24)
and
(25)
Hence
Another way
A = {qq'}
of phrasing this
{pp
}.
argument
is
as follows
By
(24),
(W
(25),
{l,l' z}
1
and, by Hence
= {?<?'} -{ = {mm'} {l,l[}

}
A = {qq
A
vector
is
{mm
{mm
{pp'}
= {qq }
1
{pp'}.
The analogy
striking.
of this process
with that
of the composition of vectors is very
A! where
and
denoted by two points. A displacement is denoted by are the orthogonal line reflections of which it is the x
product.
point
In order to add two vectors AB and CD we choose an arbitrary and determine points P and Q such that
Then we have
In the case
reflection
for
of two an orthogonal line 4 3 A (which is not arbitrary but is determined according to Theorem 25), which there are two others, A5 and Ac such that
,
AB = PO and CD OQ. AB + CD = PO + OQ = PQ. displacements A,^ and A A we find

A A = A A. AÂgAgAj = A AAA = A A
5
AgAj^
= AA
and
a
Hence
Similar remarks can be
made with regard
to
any group
of transformations
which are products
of pairs of involutoric transformations.
See
108 and,
to.
particularly, the series of articles
by H. Wiener which
are there referred
The resolution
of a general displacement into a
product
of
two
a given displacement as a product where P and are rotations, being of period two. The general solution of this problem may be
PA
found very simply in terms of the special one as follows Let P be any point of space, and let a be any line through such that A ={&&'} {'}
:
Let
and
on
If
let TT
be the line through perpendicular to a and intersecting b, be the plane through perpendicular to p. Then any line I
P
/ =j=
and
a,
TT
may be taken
as the axis of A.
{II } is
1
This
is
obvious
if I
I
= a.
the product {aa'}
a rotation about p, because
and a
are perpendicular to
at P.
Hence
{aa'}'{ll'}
Ais
{IV}
= {&&'}
= 'P
and
p,
a rotation about an axis through the point of intersection of &
Hence
(26)
A=PA
A ={#'}.
1
where
Moreover, if I be any line through P and not in IT, {aa } {W} is a rotation about a line q perpendicular to a and I and hence distinct from p. Since g/ is perpendicular to a and not identical with p, it
does not meet
&.
Hence
the displacement
A
is
{IV}
= {W}
{aa'}
{IV}
not a rotation.
Hence the
all lines
on
P which
where
pencil of lines on are axes of the rotations

is
P and
A
IT is
the set of
of period
two such
that
A = PA
a rotation.
culty in seeing that
This argument applies to any ordinary point P. There is no diffiany point at infinity is also the center of a flat
in (26). pencil of lines any one of which may be chosen as the axis of From this it follows by Theorem 24, Chap. XI, Vol. I, that the set of
which are axes of A's satisfying (26) form a linear complex. The argument for the case when P is at infinity is left as an exercise foi the reader (Ex. 7). By another application of Theorem 24, Chap. XI
all lines
Vol.
I, it is
satisfy
left as
(26) are the lines of another linear complex.
easy to prove that the axes of the rotations This

8).
which
is
also
an exercise (Ex.
Other instances of the resolution of a
general displacement into displacements of special types are given in Exs. 9-11. These exercises all connect closely with those given
DEFINITION.
twist
such that
F2
is
a translation
is
called
half
twist.
An
any half twist
a half twist, and orthogonal line reflection is a special case of is a product of two orthogonal line reflections whose
axes are perpendicular.
EXERCISES
1.
If
the three
common
intersecting perpendiculars of the pairs of oppo-
site
the mid-points edges of a simple hexagon are also the lines joining
of the pairs of vertices on opposite edges, they have a
common
intersecting
perpendicular. 2. If the product of three orthogonal line reflections is another line reflection, the three axes are parallel or are all met by a common perpendicular. and 3. For any three congruent figures v 2 , s there exists a figure
F F F
three lines l v
such that
(See the note by G. Darboux on p. 351 of Legons de Cine'matique, Paris, 1897, by (r. Koenigs, where the theorem is credited in part to Stephanos.)
4.
The axes
For any
of
two harmonic orthogonal
line reflections
meet and are
perpendicular.
5.
pair of orthogonal line reflections there is
a third
which
is
harmonic
6.
to both.
Under what conditions are two displacements commutative ? 7. For any displacement A there exists a linear complex C of lines such that every ordinary line of C is an axis of a rotation A of period two such that
A = PA
where
is
a rotation.
No
line not in
is
an axis
of
such a A.
is a displacement which is not of period two, the axes of the determined in Ex. 7 form a linear complex Cl which has in rotations common with C all the lines perpendicular to the axis of A.
8. If
9.
Any
displacement
can be put in the form
A = AP
is of period two. The axes of the A's satisfying this condition constitute the ordinary lines of the complex C (Ex. 7) and those of the P's the ordinary lines of (^(Ex. 8). 10. Any displacement A can be put in the form
where
and P are rotations and
(27)
A = P2
P
or of
.P t
If
where
P x and P 2
are rotations or translations.

:
is
not a rotation or trans-
lation, the axis of
P 2 can be
chosen arbitrarily.
The axes
of
the P^s
9
which
satisfy (27) are carried into the axes of the corresponding orrelation T.
Pa 's by
11.
Any
displacement
can be put in the form
(28)
A = PH
is
where P
a rotation or translation and
a half twist. The axis either of
P or of can be chosen arbitrarily. For any P and exists a rotation or translation P' and a half twist H'
satisfying (28) there
such that
A = HP'
12. Every
and
A = H'P.
expressible as a product in either order of an orthogonal reflection in a plane ir and a rotation about a line I perpendicular
symmetry
is
to
TT.
13. The mid-points of pairs of points which correspond under a symmetry are the points of the plane TT (Ex. 12) or else coincide with the point ITT. The planes perpendicular to the lines joining these pairs at their mid-points pass
through the point lir. 14. Every symmetry transformation is expressible as a product in either order of an orthogonal plane reflection and an orthogonal line reflection.
15. Determine the types under the Euclidean group.
of
symmetry transformations which are
distinct
123. Properties of displacements.
The main
properties of displace-
ments which we have found may be stated

Euclidean geometry:
displacement A has a unique axis a which is a line at infinity only
as follows for the real
s3
Any
in case
A is a translation.
i.e.
ment
is
The displacea product of two orthogonal
line reflections,
The lines ll and /2 meet a in two points and A z and are perpendicular to it. l Let the measure of the angle between ^ and lz be 6 and the distance between
A^ and
A2
be
d.
Then
A
T
is
the result-
ant of a translation
to a point
parallel to
a which
carries every point
X' such that
and a rotation P with a as axis which carries each plane TT on d to a plane TT' such that the angular measure of TT and TT' is 2 6. DEFINITION. The numbers 2 6 and 2 d respectively are called the a.nnl.c. nf rnta.tin>n, nnrl /?.ista,nr.A nf translation rp.sneotivfilv nf A
326
AFFINE AND EUCLIDEAN GEOMETBY

rotation
[CHAP.VII
The
P such
that
A=TP = PT
A
is
where
is
is
/
t
the line through
parallel to
1 2
Let
and
be
two
points of
and
TT
1 2
respectively, so chosen that the
measure
(
of 4.
0(and not
&)* Let one
of the
two
sense-classes
PÂJS^ 31) in the
Euclidean space be designated as

If
=
positive.
z,
=
,
there are two points
Dist (AjB a )
B B( 011 a such = Dist (-4^) = tan ft

sides of the plane
1
that
These points are on opposite
ABB
1
.
and hence
seen
^(A^B^B^B^^ S(A 1 B B2 BS ).
1
Let
If
which S(A i Bl B2 Bs )
is positive.
7?
3
B be that one of = 0, let B = A

s S
1
these points for

It is easily
of
that this determination of
is
the
same
for
any choice
and
subject to the conditions imposed above.

77"
Hence any displacement

and two
If
B A
A
for which 9 i=
and,
is
determines uniquely a line a

is
vectors
AÂ
which are parallel to a if a a translation and A,Bo a zero.

t
s,
i
AB
ordinary.
is ideal,
mine a unique displacement A. For let any line through A t and perpendicular
determines a unique point
Conversely, an ordinary line a and two vectors parallel to a deterA t be any point of a, and
to a.
Then the
first
vector
.
There are two lines

VJ-*
*
z,
1 2
and the second a unique point B through A v perpendicular to a and such that
m (Ma) = m (^0 = ^ where

A ^'
tan
= Dist B,B J-J

_ 1 2,
n.
Let
J?.
be an arbitrary
t/
point of l v and
measure
of
^ 4.B A B2
J?
2
,
points of
1
2
respectively,
such that
is
the
and
4-B^B^
Then
let B,,
be that one of
and .52 such that S(A 1 B1 B2 B3 ) is positive, and let 1 be the Az parallel to A^B^. The displacement determmed is
line through
is
it
A which is not a half twist determines and determined by a line a and two vectors A,A n and A B From this is plain that if it be desired to specify a displacement by means of
Hence any displacement
'
1213
parameters or coordinates,
it is
necessary to give a set of numbers
the line) and two additional
numbers which will specify the vectors AÂ 2 and A^B^. This question is considered from various points of
view in the following sections. For a treatment of the general problem of parameter representations of displacements and, indeed, of the whole theory of displacements, see
the articles by E. Study, Mathematische Annalen, Vol.
p.
(1891), 441, and Sitzungsberichte der Berliner Mathematischeu Gesellschaft, Vol. XII (1913), p. 36. The exercises in this section and the last one
XXXIX
are largely
drawn from the
first of
these articles
and from the
articles
by Wiener, referred to above.
EXERCISES
1.
Let
and
1>
be the axis of a twist, a any ray perpendicular to and intersecting I, the ray into which a is displaced. Let c be the ray with origin at the
I
mid-point of the segment joining the origin of a and b and bisecting the angle between the rays through, this point parallel to a and b respectively. (Two rays are parallel if they are on parallel lines and on the same side of the line joining
their origins.) The given twist is the product of the line reflection contains a by the line reflection whose axis contains c.
2.
whose axis
The product
of
whose angles
whose axes have a point in common and rotation are respectively double the angles between the ordered
of three rotations
pairs of planes determined by the pairs of axes in a definite order, is the identity. 3. The rotations P and P' described in Ex. 11, 122, have the same angle
of rotation, and the half twists
H and H' described in the
same
exercise have the
same
4.
distance of translation.
There There
line
I
exists
ordered pairs of
5.
is
an orthogonal line reflection interchanging two congruent points A ^B^ and A 2 B 2 if and only if A^B^ is congruent to A 2 B r
a unique orthogonal line reflection carrying a given sense-class
to a given sense-class on a line I'. The axes of the two orthogonal line reflections carrying a line I to a line I' are perpendicular to each other and
on a
to the
common
an.
intersecting perpendicular of
I
I'.
and
I'
at the mid-point of the
pair of points in
6. If
which the
triad
,
latter
meets
and
of noncollinear points A 1 B 1 C1 is congruent to the axis of the displacement carrying A v B it Ct to A 2 B z C2 respectively meets orthogonally the axis of the orthogonal line reflection which carries A t and B^ to two points A{ and B{ of the line A 2 B 2
ordered
an ordered
, ,
triad
A^B Z CZ
such that S(A{B{-) * S(A 2 B 2 ~). 7. If three noncollinear points
A v Az A8
,
are displaced into
Az A
,
At
respectively, the axis of the displacement is the common intersecting perand the line to the mid-point of : pendicular of the line joining z 3
A A
joining
8.
As
to the
Show how to
mid-point of A Z A. construct the axis of the displacement carrying an ordered
9. If a line I be displaced to a line I', the mid-points of pairs of congruent I or are identical; the planes perpendicular to points are the points of a line the lines joining the pairs of congruent points at their mid-points meet on a line I or are parallel or coincide. Under what circumstances do the different
cases arise?
10. If a plane a be displaced to a plane of, the mid-points of the pairs of congruent points are the points of a plane a or the points of a line or coincide the planes perpendicular to the lines joining the pairs of congruent points
;
at their mid-points pass
through a point
or
all
meet on a
line or coincide.
Under what circumstances do the

11. Let A be a displacement, mid-point of the pair PP', and
line
TC
different cases arise ?
Pa
PP'
if
P ^ P'.
Then
if
is
variable point of space, P' = A (P), P the the plane through P perpendicular to the not a half twist, the transformations T t such
affine collineations
that T! (P)
= P and T2
such that
T2 (P) = P' are T2 T 1 = A=T1 T2

.
and
If
A is
correlation such that P(TT)
not a rotation, the transformation P such that = P' i.e. such that
;
P (P) =
TT is
a project! ve
P =A.
S
If
is
A is
not a rotation or a half twist, the transformation
a projective correlation, and in fact is the null-system of the complex 122. These transformations also satisfy the equations referred to in Ex. 7,
N such that N (P) = TT C
T1= NT,
12.
T2 =PN,
if
NA = AN.
a is any plane, A (a) = a', and T2 (a) = a, then a bisects the pair of planes a and of, and T 1 (a) = a. 13. In the correlation N the lines I and I defined by Ex. 9 correspond.
Using the notations
of
Ex. 11,
The plane a and the point A defined 14. The linear complex C*(Ex. 7,
cides with the line
I
in Ex. 10 also correspond in
N.
122) contains every line
which coin-
determined by the same line J(Ex. 9). Hence it is the set of those lines f which are perpendicular to the lines joining corresponding
points of
15.
I
and
I',
and
it is
also the set of lines I

I
which intersect the
lines join-
ing corresponding points of
and
I'.
The The
affine collineation
(Ex. 11) carries the axis of
P (Ex.
11,
122)
to that of H'.
16.
correlation
P (Ex.
11) carries the axis of Pj (Ex. 10,

T,
122) to that
of P,. 17.
The
transformations
T- 1 T 2
,
T~ l
all
carry
C (Ex.
7,
122) into
Cx
(Ex.
8,
122).
124. Correspondence between the rotations and the points of space.

If
we
By
confine attention to the rotations leaving a point

the reasoning in
invariant,*
90
it is
clear that this
amounts
to considering the effect
the considerations of the last section simplify considerably. The points A l and A 2 may be taken as coincident with 0, and the point B S shall be denoted by R. Then every noninvolutoric rotation P
corresponds to a definite point
(orthogonal line reflection)
at infinity of
its
on
its axis.
An
involutoric rotation
axis.
be taken to correspond to the point Hence the rotations leaving invariant cor-
may
respond in a one-to-one and reciprocal

projective space
way
to the points of the real
consisting of the given Euclidean space
and
its
points at infinity.
Let OX,
Y,
OZ
as center such that
be axes of a rectangular coordinate system with S(OXYZ] is the positive sense-class. Whenever
origin, denote the
is
distinct
a,
from the
/3,
measures
of
4.ROX,
4. ROY,
&ROZ by
7 respectively.
x
Then the
tan 6 cos af
coordinates of
are
= tan cos /3, 2 = tan 6 cos 7.
Let (aQ a^ # 2
,
3)
be the homogeneous coordinates of R, so chosen
that
if
R R
is
ordinary,
~a
aQ =
'
~a
ao
a
'
~~
'
ao
and
if
is
at infinity,
'
0.
In either case we may take
= cos 0,
to
a 1 = sin
cos a.
= sin a4 n
cos
/3,
ao =
,
sin
cos
7.
According
as a product of According to the convention just introduced, the be regarded as direction cosines. Hence, by Exs.
,
Theorem 23 any rotation (<XQ a^ a 2 <x3 ) is expressible two involutoric rotations (0, X J} X2 X 3 ) and (0, / x /*2 /* 8 ).
, , ,
X's
and
7,
p's
may
and
119,
(29)
tf
= XA + X
2/
Two fundamental problems now arise (1) to express the coordinates of the point representing the resultant of two rotations in terms of the coordinates of the points representing the rotations, and in terms of the parameters (2) to write the equations of a rotation
:
The formulas
rronorcil
the solution of the

OQQO
which furnish (29) are a special case of the formulas The formulas for the first of these problems.
Vvo
mow
-Frmnrl
Tnv
an
nTVnliP.fl.Hrm
nf
tlVlfi
TnfitjQOfl
TOT
compounding
,
rotations described in
,
III.
Let the two rotations
= (at o^, a a ,) and J3 = (ft, ft, ft, ft) respectively. correspond to A Let /* /* /* be direction cosines of a line perpendicular to OA and is expressible by means of 05. Then the rotation )
, ,
o^,
8,
the formulas (29) and

(30)
(ft, ft, ft, ft)
by the following:
/*
ft=
,
which represents the product of
According to the principle explained in 122, the point (? 7^ ya 7 8 ) a lt a a 8) followed by (ft, ft, ft, ft) is (
,
,
,
\\
The
result of eliminating the X's,
/-t's,
X. X.
7B =
and
v's
from these equations
is
(32)
This
most easily verified by substituting (29), (30), and (31) in (32). 7 t 72 7 8 ) which is the product of (aQ a^ az> ag ) and for if not, there would /Q t /32 /3 a ) must be that given by (32) (/3 be some case in which (32) would not be satisfied by the values of etc. given by (29), (30), and (31). ft, 7
is
The
,
rotation (?
,
,
The formulas
(32),
which
are due to 0. Eodrigues, Journal
de
Mathematiques, Vol. V (1840), p. 380, are the same as those for 127. the multiplication of quaternions. Of.
The problem
(ao
of
,
(2) of
expressing the coefficients
of
the equations (17)
of a rotation in terms of the coordinates of the corresponding point

CK
I}
az a s ) may be solved very

,
easily "by the formulas
118, in the case of rotations

(0,
rotation corresponding to
x'
1
Xa Xg
,
period two. X 8 ) is, in fact,

of
l
and theorems The involutoric
(33)
y
z<
= (2 X? - 1) x + 2 X X y + 2 \\z, = 2 \\ x + (2 X| - 1) y + 2 X X = 2 Xâ + 2 X X y + (2 X - I)L

8
z
8 z,
This
is
easily verified, because (1) the matrix is orthogonal

is
and
,
its
determinant
+1,
(2)
the transformation leaves the point (X^ Xa X 8 )
invariant, (3) tha matrix
is symmetric and hence corresponds to a transformation of period two. To obtain the equations of the transformation corresponding to it would be sufficient to take the product of (a ctj, 2 (33) and 3)
,
,
the corresponding transformation in terms of
/A 2 , ft 3
and compare
with equations (29). The algebraic computations involved would, however, be more complicated than in the following method, which
is
based on a simple observation with regard to collineations whose equations are of the form
(34)
a^âfi + a^-a^, aj/=-a x + a$ + a& a az=ax ajjj +

a Q 2
z.
If
P(x,
(35)
y, z)
and P=(x,
y, z),
then the vector
OP
is
perpendicular
to the vector
PP, because
x(x
%}
+ y(y
y}
+ z(z
z)~Q.
The transformation
if
(34) also has the obvious property of leaving a g ). invariant all points on the line joining the origin to (a : 2
,
Conversely,
a collineation
(36)
= (x, y, z) is distinct from = (x, has the property that whenever OP is perpendicular to PP, the relation (35) requires that a ij
y, z),
a}l
whenever
all
i=t=j
and that
p= a u =
22
33
If,
,
moreover, (36) leaves

it
a a s ) invariant, points of the line joining the origin to (alt z be either of the form (34) or of the form
must
ax
(34')
<x Q
<x o
= a x - aj] + ajs, = a x + a$ - a& z = -a x + a$ + a

o
z.
It is also to
be observed that the determinant of Transformations
(34)
and
(34') is
a A, where
(37)
TViio rlnfa-nmiviei-nf.
r>a
A=a% + at + at + at.
n
xronioVi
-fn-p
TOdl
n-'fi
nnlv
if
ff.
= 0.
Now
rotation
1
consider an orthogonal transformation (17) representing a P which is not of period two. Let P be an arbitrary point,
p = P(P), and P the mid-point of P and P'. and P is given by the equations*
The
relation
between
(38)
is perpendicular to OP and (38) must have the same line invariant points as P. Hence if P is the rotation corresponding to the equations of the transformation from to must (a a , 2 3 ),
The
PP
be of the form (34) or (34'). 31, we see that Forming the determinants analogous to (19) in S(OPPR], where R = (aQ a lt a2 a s ), is positive if P is given by (34) is given by and negative if (34'). Hence (38) must be the inverse
,
,
of (34).
Solving the equations (34)
we have
a, a.
(39)
A+
l8-o
^
+<
.
5 ==
3a
Ql
y
.4
'
Since (38) and (39)
must be the same transformation, we have
(40)
a !i =2
j]
-.
These are the formulas, due to Euler, for expressing the coefficients of an orthogonal transformation in terms of the homogeneous parameters
The formulas
for the a' a in
terms of the
a^.'s
:
may
be obtained by
taking linear combinations of Equations (40)

1
i i .
**r\
wnQ*Q
4: OCn GCn
From
(41)
this it follows that

'.
:a 1M --a u :a n -a iM J]f The algebra of the last section may be put in a most compact form by means of matrix notation. This requires one or two new definitions. The sum of two matrices is defined by means of the following equation:
l
:a t :a t
=l+a +a n +aa :a n -a
ll
125.
Algebra
of matrices.
/
(42)
a ii a n
V\
AL
6
&
12
M
\a gl
22
aj+
A+
22
a 82 a B8 /
\6 81
682
.
& 88 /
U=K+& +
\a 81
ts\
a / u
+ 6u
21
a t2
22
6 81
a 82
+& +^ +6
i2
a i3
23
22 82
+& + &i8\
28
-f-
a ss
& 88 /
).
This operation obviously

laws,
satisfies
the associative and commutative
namely
where A, B, C stand
for matrices.
Multiplication of matrices has been defined in

(43)
8
95,
VoL
I, i.e.
w-(y=(a
Under
.4(5
this definition
it is
where cr
= ^ a ^w
* =1
clear that
+(7)=
f
and
Also
it
(^
+ C )JL
has already been proved that
It is
now
most
of the properties of a
"Vol. I,
easy to see that, under these definitions, matrices have noncommutative number system in the
sense of Chap. VI,
the matrices
0\
/O
000
\0
taking the roles of
/I
0\
II
of the
and
010
form
O/
\0
and 1 respectively. The matrices
fx
(0
0\ x 01
form by themselves a number system which is isomorphic with number system, of the geometry. Such a matrix may be calle
scalar and be denoted by
x.
us denote the orthogonal matrix of the equations rotation (17) by R, and let the skew symmetric matrix
let
Now
a.
be denoted by S. Then the matrix of the transformation (34) is 1 and the matrix of the transformation (38) is l (1 + 72). The compai
of coefficients of (38)
and
*
of (39)
'
amounts
to writing
1
"
3
)-
follows
a relation
if
which
is
obvi
from the point
of
view
of matrices.
For
matrix, the transposed of S is - S. posed matrices of the two given matrices product, the transposed of
be any skew symme Since the product of the tore

is
the transposed of
is
(l
is also its
)(i _#)-!,
which
inverse.
Hence, whenever
is
orthogonal.
This equation
may
be solved as follows
- (1 + j?) (1 +
which gives the formula an orthogonal matrix.

The operation
Vol.
I) in
for a
skew symmetric matrix

matrix
is is
in terms of
of taking the inverse of a
defined
(cf.
95,
from zero. In the operations above, this is a restriction on the matrix 1 + R and, by comparison with Equations (22), is seen to mean that no point
distinct
case the determinant of the matrix
must be transformed by the
rotation corresponding to
into
its
symmetric point with respect to the origin. The generalization from three-rowed to u-rowed matrices is obvious, and we thus have the skew symmetric and orthogonal matrices of
n rows connected by the

(44)
relations
J2=(l-
(45)
S=(l-
The equations between the corresponding elements in the matrices which enter in the first of these two matrix equations are the formulas given
expressing the
by Cayley (Collected Works, Cambridge, 1889, "Vol. I, p. 332), n 2 coefficients of an orthogonal transformation as
rational functions of
--
parameters.
of rotations
126. Rotations of
an imaginary sphere. The group
leaving a point invariant may be regarded as a subgroup of the collineations of a sphere having this point as center. Let us consider
the imaginary sphere

(46)
4-
4-
x2
4-
x3
102.
-1
rf>
and apply some
of the results obtained in

f
If a collineation
_1_
_l_
>
(47)
^
f
^ + ^ ^ ^J' ^ ill
one regulus on the sphere into itself, any point xv x2 x s ) satisfying the condition (46) must be carried into a point (#', x[, xz C8') satisfying the condition
carries each line of
(XQ ,
, ,
which
states that it is
on the sphere, and the condition

-4-
x.x'
x.xl 4- OUG' 4-
xxl
= 0.
tutmg
(47) in (4$)
we
nave, as in
no,
Substituting (47) in (49)
we have
The matrix
of the equations (47)
must
therefore be of the form
a.
(50)
or
/
ft
2
ft
(51} ^ '
I
\
ft "ft - P - ^8
/3
lg
ft ft
fti
-ft
P!
/3
_ jg
;
On
product
multiplying together two matrices of one of these forms, the whereas if two matrices of is seen to be of the same form
forms are multiplied together, the product does not satisfy i c Hence the matrices of the form (50) the condition c y = jt
different
,
=/
projective collineations leaving all lines of one regulus on (46) invariant, and those of the form (51) must represent the projective collineations leaving all lines of the other regulus
must represent the
invariant.
Hence, by
is
102,
the sphere invariant

(50)
any direct projective collineation leaving represented by a product of a matrix of type
by one
of type (51).
A rotation is
a direct collineation leaving invariant both the sphere
and the plane at infinity XQ = 0. A collineation (47) leaves x = invariant if and only if c01 = cQ2 = cQ3 0. But on multiplying (50) = /j/3Q a^ = pftlt and (51) it is clear that this can happen only if
,
p/32 ,
pft z , p
being any number except

is
zero.
Hence the
matrix representing a rotation
AA, where
and
A=
{(*
aa
The matrix
of the
product
+<+<
a*
+ a* - a* - <
+# a
3)
000
AA
is
2 (or^
2ffa-aa
,
a^} < + a - a, - a
2
(a^ -
2 (oâ, 2 (a2 3
+ - a aj
;
2)
2 a
which agrees with (40) of 124. Hence the parameters (a a v a2 aa ) in

,
the
Euler formulas
may be
regarded as the elements of a matrix of the form (50) which represents the projectivity effected on one of the reguli of (46) by the rotation.
If two rotations effect projectivities and B respectively on a on regulus, the product of the rotations effects the projectivity the regulus ( 102). Hence the product of two rotations whose
BA
parameters are
parameters (y
ry ry
ry ry
,
o a^ # 3 ccs ) and 7^ 72 7 3 ), where

,
(/3
/^
/32 , /33 )
respectively has the
7fy
I
'0
'3
-(V
7s
(V
i-i
O/
11
This yields the same formulas as (32) in
124.
EXERCISE
parameter representation for the sphere (46)
is
+
where
A-I/AQ
is
2
1.
/^//x,,)
The two
and
is
reguli on the sphere are the sets of lines for which respectively are constant. The transformation whose matrix
(50)
given by the projectivity

ô
( ao
ia i) ô
a ( s
ia 2> *i
M=~
127.
( as
'
aa) ô
+ ( ao ~
{a i)
Xr
product of matrices three-rowed matrices clearly apply to matrices of any number of rows. With this understanding the sum of two matrices
Quaternions.
for
The
definitions of
sum and
in
125
of the
Viflst
form (50)
is
obviously a matrix of the same form.
The same
two
hppn sppn
in flip lacf. spr>fvinn tn hft fvrnP. of thfi TVrndnP.fcs of
such matrices.
itself
Hence the
of
set of all
such matrices
is
carried into
by the operations 125. denned in
addition and multiplication of matrices
Let us introduce the notation
'100
1
010 001
0010'
0100' 0010; 0001'

00-10
-1000,
are considering
is
1000 -i
o
yO
0001 100
0-1
Then any matrix
the form
of the sort
0|'
0,
0100
we
expressible in
The matrices
i,
j,
k satisfy the following multiplication table
j
(52)
-1
,
*
^^ n
..
-j
n v
f\r
Z**
JL
j
It
-i -1
satisfy the associative
has heen seen in
125 that matrices
and
commutative laws
satisfy the
of addition, the associative
laws
of multiplication,
and the distributive laws, They obviously do not, in the present case, commutative law of multiplication. Addition is performed by the rule
(53)
(<xj.
+ aj, + aj+a. = + /3 )1
(
-f-
/ 4-
(a
and
multiplication,
(a
by the rule
o
i
(54)
fl
1+ aj, + aj +
-f
where
**!=*& +
(55)
a.B
a 8
Dnat (PO L
+ p + p j + p K)
'i,
exists
whenever the determinant
ô
-A
-ft
ftl-ft. ft ft
is different
/3 2 ,
S1
-ft
This condition
is
from
zero.
satisfied
whenever
ft, ft,
ft are -real.
,
Hence when aQ a^ a^ a 3
constitute a noncomrnutative
are real, the matrices of the
number system
form (50) in the sense of Chap. VI,
Vol.
I.
quaternions.
particularly
This number system is, in fact, the Hamiltonian system of Compare the references at the end of the next section,
p.
178 of the
article in the
Encyclopadie and the article
by Dickson
hi the
Bulletin of the American Mathematical Society.
EXERCISE
quaternions may be defined as a set of objects [j] such that (1) for every ordered pair of vectors a, & there is a q, which we shall denote by (2) for every q there is at least one pair of vectors; (3) two pairs of
system
of
(");
vectors OA,
OB
ordered triads
OAB
and OA', OB' correspond to the same q if and only if the and OA'fJ' are coplanar and directly similar in their
common plane; (4) the q's are subject to operations of addition and multiplication denned by the equations
Prove that a Rystom of q's satisfies the fundamental theorems of a number system with the exception of the commutative law of multiplication. See G. Koenigs, Lecons de Cine'matique (Paris, 1897), p. 464.
128. Quaternions
and the one-dinaensional projective group.
On
comparing (32) and (55) it is clear that there is a correspondence between quaternions, taken homogeneously, and the rotations leaving
a point invariant in
which
rotations
P P 2 respectively,
x,
group
of rotations is
two quaternions q lt q z correspond to the the product g 2 g 1 corresponds to P^. The isomorphic with the group of projective transif
formations of the circle at infinity and hence with the projective group
relation
between quaternions and the one-dimensional projectivitit ax j

x'
The simplest way to obtain a number system corresponding these transformations is to apply the operations of addition ai
multiplication as defined above to two-rowed matrices,
i.e.
7i
8
i
K
I
\/t
^
I/
"
i
i
'
V "V \ '2
c>
2'
r/ \'VC-(-O 1'2 \'l 2

'
o>
/->
'V/O
'1^2
If
we
write
_/l
1
0\
2
_/0
@\ v
1\
e3
_/0
, ,
0\
64
/O
1^0
O/'
/<%
\0 O/'
\1 O/'
,
~\0
we have
The
units eit
e
z,
\7
e
s> e^
o * = ae, + p + 76, + oe,.

i
satisfy the multiplication table
e, 2
laws
Although these matrices satisfy the associative and distributrv of addition and multiplication and the commutative law c
it is
if
addition, it is clear that they do not constitute a
because
theless,
possible to have ab
number systen
I 3= 0.
when a
and
we
write
any matrix
i
2
^
2
1c
is
j
expressible linearly in
if
1,
i,
j, Jc;
A
and
=f =
= - 1,
ji
= k,
jk
kj
tfc
=/.
Hence the system
of
two-rowed matrices
where
a, /3, 7,
8 are complex numbers,
is
equivalent to the set
o;
elements
(56)
S izaj
VqJU
AXJiJilNlUJNO
54rJ.
where
The
a, b,
multiplication table (52) of quaternions. elements (56) are quaternions, properly so called, only when c, d are real. When a, b, c, d are ordinary complex numbers, the
1,
i,
j,
k satisfy
tlie
elements (56) do not form a number system in the sense of Chap. VI, Vol. I, because there can be elements x, y both different from such
that xy
= 0.
i
It is interesting to note that 1,
W
*v
/v>
i,
j,
are the matrices
i\
/'
_V:TI/' V-i
which represent the
identity,
tv
, /yi'
wr
__ ^"
and three mutually harmonic involutions
__
/>>
tv.
,1,1 _
__
,__,_
Ml w
a?
If the projectivities are
represented on a conic, these three involutions have the vertices of a self-polar triangle as centers. The matrix represented by
its
determinant
is
The geometric
The
relation
significance of this
remark
is
obvious on comparison
with the exercise in
126.
between quaternions and the one-dimensional projec-
was discovered by B. Peirce (cf. Chap. VI by A. Cayley in Tait's Quaternions, 3d edition, Cambridge, 1890). It is an instance of a general relation, noted by H. Poincare\ between any linear associative
tive group
algebra and a corresponding linear group. On this subject see E. Study. Mathematical Papers from the Chicago Congress (New York, 1896), 12 Lie-Sheffers, der Math. "Wiss., I 4, p. 376, and Encyclopadie
Kontinuierliche Gruppen (Leipzig, 1893), Chap. XXI; and L.E.Dickson, Bulletin of the American Mathematical Society, Vol. XXII
see (1915), p. 53. On the general subject of linear associative algebra L. E. Dickson, Linear Algebras, Cambridge Tracts in Mathematics,
"No. Ifi.
1914: and
thft
fl.ri-.inl ft
bv
"R.
Studv and E. Cartan in the
W lilt; II we
unseu.
JLU
g j.^;^ uu.
a,
OÛUIJ.U.CKIJJ. uuu.oiii.u.ub.i.uu
aao LLUYV
been seen to be connected in the
closest
one-dimensional projective group. It is the correspondence between the points of space and the rotations about a point in a form which puts in evidence also the correspondence between the points of space and the one-dimensional projecThis has been studied in detail in the memoir by Stephanos tivities.
referred to in Ex. 3,
the theory of the therefore of interest to set up
way with
110. It will be merely outlined here, because simple applications of theorems which should hy The construction given below this time be familiar to the reader.
the proofs are
all
has the advantage over the one given in

general projective space.
123
of being valid
in
Let
vious
0).
$ 2 be an arbitrary sphere. (In order to connect with our prework $ 2 may be taken as the imaginary sphere 2 -f-2/ 2 -f-z2 -f-l z Let R* and R% be the two reguli on S (9 the center of S 2,
,
and
the circle at infinity.

leaving
invariant determines and
is
An arbitrary rotation P
fully
determined by a projectivity F 1} by its effect on three points P

of
of Cl,
Z)
s
and hence
If
g
is
fully
l s
determined
Rl on
P v P
(PJ,
i}
P% respectively,
and
P P of C*. m m m
1,
v l^
are the lines
z,
the lines of JS 2 on
l z
the
l
a
points
P (P z ], P (P s
correspond to P and to T (of. Ex. 2, 110). The following propositions are now easily established by reference to theorems on one-dimensional forms
meet in a point R. Let
respectively, the planes Ijm^
z>
The point
of
is
on the axis
of
P and
is

is
P v P P
2,
8.
If the line
OR
meets
S2
in
two points Q^QZ
Tfc
(Q^Q^,
OR)
the
cross ratio of F.
to points of the plane at infinity. Pairs of inverse projectivities correspond to pairs of points having as mid-point.
The involutions correspond
Harmonic projectivities (80, Vol. I) of Cl correspond to points which are conjugate with respect to S 2 The projectivities of C% harmonic to a given projectivity correspond to the points of a plane. Such a set of projectivities may be
.
called
a bundle of projectivities.
A pencil of involutions according to this definition is the same as a pencil of involutions according to the definition in 78, Vol. I. The product of the projectivities corresponding to points Jt l and Z not collinear with 0, corresponds to a point obtained by the fol3
lowing construction Let I', I" be the Hues of JRf through the points 2 in which OR 1 meets $ and let m', m" be the lines of R% through
: ,
the points in which
OR2
meets
2
.
The
line
m' and
m"
.
intersects the line
through
through 7^ meeting meeting V and I" in the
point -B 3
If
V and
,
I" coincide,
to be tangent to
$ 2 and
the line meeting them is understood a similar convention is adopted in case m'
and
If
m"
coincide.
be regarded as fixed and

relation
R2
as variable,
is
connected with
by the
where
is
invariant.
In case
a projective collineation leaving the lines V, I" point wise I' is a coDineation of the type in which all I",
points and planes on V are invariant and each plane on V is transformed by an elation whose center is the point of contact of this
plane with S If J? be regarded as fixed and J^ as variable, the transformation 3 defined by the relation
.
is
each line
line
a collineation interchanging the reguli R% and R%, of R% in the plane I of Rf into the line
and carrying and each 3 l,
of
R%
into the line
of
R%
in the plane
Om.
The propositions above In a real space we have The

ray with
are derivable
from Assumptions A, E, P.
rotations represented
by
as origin to a certain other ray
points of a line all carry a certain as origin. Conwith
versely, all rotations carrying a given ray
with
as origin to a
second ray with
as origin are represented

sufficient
The necessary and

harmonic
is
by points of a line. condition that two rotations P 1? P2 be
that there exists a ray r such that

of rotations
P x (r) is opposite to P
124
is
2
(r).
The representation
by points given in
identical
with the one given in this section, in case $
S z is
real,
In case is imaginary. the real points of space represent imaginary rotations.
If
$2
is
a ruled quadric
and C* a
real conic, the construction
above
gives a representation of the real projectivities of a one-dimensional 2 form by the points of space not on S The sets of points [D] and
.
[0] representing the direct and opposite projectivities respectively are such that any two points of the same set can be joined by a
segment consisting of points of this contains a point of $ 2 a D to an

the two sides of
set,
The
whereas any segment joining sets [D] and [0] are called
2
.
EXERCISES
Study the configuration formed by the points representing the rotationa which carry into itself (a) a regular tetrahedron; (6) a cube; (c) a regular
1.
icosahedron.
2.
(Cf. Stephanos, loc. cit., p. 348.) real quadric (ruled or not) determines
two
sets of points, its sides,
such that two points of the same side can be joined by a segment consisting entirely of points of this side and such that any segment joining two poiats
of different sides contains
ruled,
one point of the quadric.

its sides
one and only one of

the outside or
contains
all
is called
exterior,
and the other the
If the quadric is not points of a plane. This side inside or interior.
130. Parameter representation of displacements.
Simple algebraic
considerations will enable us to extend the parameter representation of rotations considered in the sections above so as to cover the case
of
displacements in general.
We
will suppose the general displace-
ment given in the form
(57)
where the matrix

aw
>
126, if (# n a 22 a 8a ) is orthogonal. According to aw = a 8o~ tne matrix of (57) is expressible in the form AA, and A being defined at the bottom of page 336.
Now
<w\
(58)
observe that
if
/2/3 2/3t
\*
\2/38
000
B
is
and C
is
any four-rowed matrix, C

Q-V>Onf.
f.llAao
f\f
f.l-IO
a matrix in which all
ol*TnOnt.G
fl-rof.
(59)
we have
\
cc.
2
a..
2 (a 2 ft& U
v
a 01 ft-
aX + V)
U
ij
O'
Hence the
coefficients of (57) are given in

,
,
terms of two sets
of
homo(40),
geneous parameters ao o^, as together with 00 =1 and
ô
/3 i; /3 2 , /3 3
by the equations
(60)
=2
2 /3 Q
provided that the a's and /8's are connected by the relation (59). Conversely, the a's and fi's are determined by the coefficients of (57) according to the equations (41) and the following
:
(61) ft
:^:/3,:A=
,(-
)+*(!,-
n)
(- aJ:
The last equations are obtained by solving

for the
It
/3's
(59)
and substituting the values
of the a's given
and (60) simultaneously by (41).

r
@j,
)8[',
remains to find the formulas for the parameters (a' a{', a", a"; = A' A, /8^, /3g') of a displacement A" which is such that A"
,
where
has the parameters
a^ a2 a^
,
Q,
ft, ft, ft)
and A' the
parameters
<
a^,
a;
ft,
/3(,
^,
ft).
given by (58).
13 i form given at the bottom of page 336 and In like manner A' can be expressed in the analogous form A'(A'-B') and A" in the form A"(A"-B"). Since the /3's do not enter into any coefficients of (57) except a jo a 2Q a go it is clear
and A are
of the
that
a'J,
a", a", a" are given by the formulas (32),

definition,
or,
in other
words,
that
A" = A'A. By
= A' A' A A - A A'AB - A'B'AA + A'B'AB.

!
In view
of (59), the
elements
of the first
Hence
all
the elements of
B'AB
are zeros.
row of Hence
AB
are all zero.
and A' are the matrices of transformations of two conjugate each transformation leaving all the lines of the other regulus invariant, they are commutative. Hence
Since
reguli,
A" (A" - B") = A'AA'A- A'AA'B - A'AA~ 1 B'AA.

But
î~ 1=: yl*
_
o
ttl
where
A*
K_
\a s
.
a* a2
-a a ~
2 *
as a~
-a\
and
B'AA= B'
+ < + a\ +
2 8 ).
Hence
(62)
A" (A" + B") =A'A (A'A-A'B - A*B>}.

it
Since
A" = A'A and 2" = 22,
follows that
(63)
Hence
(64) '
v
ffi
^=
=ai/
<x 2
-t
130]
BIQUATERNIOKS
347
Rewriting (32) in our present notation, we also have
(65) can be put into a very convenient form notation of biquaternious.* Let us define a biquaternion as any element of a number system whose elements are expressions of the form
The formulas (64) and

of the
by means
(66)
=
a's
(ff
+a +
z
l
<r
t
^)
+ e(j9 + ^ t+^y + Â),

1
where the
i,
j,
Ic
the geometric number system, are subject to the multiplication table (52), and e is subject to
and
yS's
are
numbers
of
fcherules
=0,
*,
any other element, and where the elements (66) are added and multiplied according to the usual rules for addition and multipliis
where x
cation of polynomials.
If
the product of
'
and
s',
where
= (a> + a[i + a'J + a'Jt) + e ($ + ft* + ffj
be denoted by
are given by the formulas (64) and (65). /Sg' For a more complete study of the parameter representation, of displacements, see E. Study, Geometric der Dynamen (particularly II,
the a",
21), Leipzig, 1903.
EXERCISES
parametei'S of a twist may be taken so that or 1 a v a a are the is the angle direction cosines of the axis of the twist ; <? = cot 0, where 2
1.
The
of rotation
2.
and
/3
</,
where 2 d
is
,
the distance of translation.

F,
Find the equations of Tv

123.
T2
N,
etc.
as
denned in the exercises
of
is
*3. Find a parameter representation for the displacements in a plane which analogous to the one studied above (cf. Study, Leipziger Berichte, Vol. XLI
(1889), p. 222).
GENERAL EXERCISES
Classify each theorem in this list of exercises according to the type of projectit apace in which it may 1>e valid and according to the geometry to which it belongs.
1. homology whose plane of fixed points is ideal is called a dilation c expansion. Any transformation of the Euclidean group is either a display ment or a dilation or the product of a rotation by a dilation.
2.
Any
transformation of the Euclidean group leaves at least one
lin
invariant.
'
3. Any transformation of the Euclidean group is either a displacement < a dilation or the product of a displacement by a dilation whose center is on fixed line of the displacement. 4.
Let
be a line which
let
is
invariant under a transformation

(
of tl
Euclidean group, and
k be the characteristic cross ratio
73, Vol. I)
if
the projectivity effected by 1. only if k =

5.
T on
I.
is
a displacement or
symmetry
an
Any
be expressed as a product
transformation, of the Euclidean group which alters sense ca APA, where A is a dilation or the identit;
P an
orthogonal plane reflection,
an orthogonal line reflection or
tl
identity.
6. If two triangles in different planes are perspective, and the plane of or be rotated about the axis of perspectivity, the center of perspectivity will d scribe a circle in a plane perpendicular to the axis of perspectivity (Cremon
Projective Geometry, Chap. XI). 7. The planes tangent to the circle at infinity constitute a degenerate plat quadric. With any real nondegenerate quadric this determines a range
quadrics,
i.
<
e.
a family of quadrics of the form
/(u l5 u2 u 3 )
,
4-
X(u a + uj -f
0,
where f(u v w 2 w s ) is the equation in plane coordinates This is called a confocal system of quadrics. Besides the
,
of the given quadri circle at infinity th
range contains three other degenerate quadrics, an imaginary ellipse, a re; ellipse, and a hyperbola. There is one quadric of the range tangent to ar
plane of space.
space,
There are three quadrics of the range through any point and their tangent planes at this point are mutually orthogonal. 8. Let [7] and [m] be two bundles of lines related by a protective tran formation T. There is one and, in general, only one set of three mutual]
perpendicular lines l v
lines
l
z , lz
transformed by
to three
m v m2 m
,
There are two real pencils of
lines in [Z]
mutually perpendicuh which are transf orme
Cf. Encycloped: des Sc. Math., Ill, 8, 9. 9. Let r be a collineation. of space. The planes and r- 1 (7r ) a] M r(7r ) called the vanishing planes of T. Through each point of space there is a pa: of lines each of which is transformed by T into a congruent line (i. e. pairs <
ot>
by T
into congruent pencils of [m].
What special cases arise?
carried
11.
10. A collineation F which, does not leave the plane at infinity invariant determines two systems of confocal quadrics such that the one system is 84 and the references given there. by T into the other. Cf
.
be a direct-similarity transformation of a plane, A 1 a variable point of this plane, ^4 2 = T(^t 1 ), and A s a point such that the variable triangle AÂÂS is directly similar to a fixed triangle B^B z B y Then the transformations
Let
to A & and from A z to A 3 are direct-similarity transformations. Both have the same finite fixed elements as T.* Let T be an affine transformation, A l a variable point, A^ ^Â^), and A a point such that the ratio AÂJAÂ^ is constant. The transformation P from AI to A is directly similar and has the same fixed elements as T. If
from
Al
of these transformations
12.
is
a similarity transformation, so is P. 13. If T x and T2 are amne transformations, J.

2
Az = T
*
and A 8 a point such that AjÂÂÂy formation from A Q to A 3 is amne.

(AQ),
a variable point, A 1 = Tl (A ), is a parallelogram, the trans-
On
this
and the following exercises
cf .
Encyclopadie der Math. Wiss. Ill AB 8,
pp. 914-916.
CHAPTER
VIII
131. Hyperbolic metric geometry in the plane. According to t 34 there must be a geometry con point of view explained in sponding to the projective group of a conic section. The case of
real conic in a real plane is one of
extreme interest because
of
close analogy with the Euclidean geometry, as will be seen at once DEFINITION. arbitrary but fixed conic of a plane TT is call
An
the absolute conic or the absolute.
The
interior of this conic is call
the hyperbolic, plane.

called ideal points.
Points interior to the conic are called ordina
points or hyperbolic points, and those on the conic or exterior to
it a
line consisting entirely of ideal points is call
an ideal
line, and the set of ordinary points on any other line is call an ordinary line or a hyperbolic line. The group of all projecti collineations leaving the absolute conic invariant is called the hyp^
bolic (metric)
group of
the plane,
and the corresponding geometry

IT is
called the hyperbolic plane geometry.
Let us at
first
assume only that the plane
ordered (A,
JS, S, J
On
this basis
we have
as a consequence the theorems in
'
74,
on the interior of a conic, that the points of an ordinary line satis the definition in 23 of a linear convex region. This determines t
of the terms " segment," " ray," " between," " precede," e as applied to collinear ordinary points and sets of points in the hyp' bolic plane. The ordinal properties of the hyperbolic plane may
meaning
summarized
as follows
1.
THEOREM
for
the
The hyperbolic plane Euclidean plane in 29.
satisfies
Assumptions I- VI gii
are direct consequences of t Proof. Assumptions I, II, III, proposition that the points of an ordinary line constitute a lin< convex region. Assumption VI, that the interior of a conic c(
tains at least three noncollinear points, is an obvious consequer of 74, 75.
UJL V is 1)110.1) blliCO UUiXlLS A, JJ f \J art! ij JJ jn.C3C5U.uu.jJl/lAJU. JL noncollinear and that two other points and satisfy the order relations {BCD} and {CEA}. The conclusion is that there exists a
on the line and between A and B. To prove this it is point of the projective necessary to show (1) that the point of intersection
DE
lines
DE
and
AB
is
interior
to the absolute conic
and
(2)
that
F
/
is
between
A
its
and B.
Let
be a line exterior to
points
\
the conic, and let

of intersection
s^^Vc
/
with the lines
-A^-D
A
\
O--
AB, EC, CA
respectively be
Fx
By hypothesis FIG. 77 the pair DD^ is and the pair not separated by n is separated by AC. Hence, Since F* is exterior to X is separated by AB. 26, the pair by
,
!),
Ex
ry^L.
and
75,
EC FF
EE
the conic,
F is
interior
75) and between
and B.
THEOREM 2. The hyperbolic plane On the contrary, if a is any 9.

there are infinitely
many
lines
Assumption and A any point not on a on A and coplanar with a which do

line
does not satisfy
not meet
Proof.
a.
By
75 the projective line containing a also contains an
infinity of points exterior to the absolute.
Any line
of the hyperbolic
plane contained in the projective line joining

points fails to
to
one of these
meet
a.
DEFINITION. If a projective line containing a line a of a hyperbolic plane meets the absolute conic in two points .#, Cm , and A is any ordinary point not on a, the ordinary lines contained in the projective
lines
ABX and ACn are said to be parallel to a. The segments ABX and AC^, consisting entirely of points interior to the absolute, constitute, together with A, two rays which are also said to be parallel to a.
If
the projective plane
exist for every line a,
TT be supposed real, the points and hence we have
and Cn
THEOREM 3. In the real hyperbolic plane there are two and only two lines which pass through any point A and which are parallel to a line a not on A. There are two and only two rays with A as end
parallel to
a,
352
[CHAP.VJ
This theorem of course does not require full use of continuil assumptions. It would also be valid if we assumed merely that ar Hue through an interior point of a conic meets the conic (cf. 76).
at infinity or infinite points
ultra-infinite points.
DEFINITION. The points on the absolute are sometimes called poin and the points exterior to the absohit
;
132. Orthogonal lines, displacements, and congruence. DEFINITION. Two lines (or two points) are said to be orthogonal
<
perpendicular to each other

absolute.
if
they are conjugate with respect to
tl
hnfc
Of two perpendicular points one is, of course, always ultra-infmit no fmeainn-nno c-f n^rv,~-<- ^->i(}g f or perpendicular lines. From tl
s
we deduce
at once
ndicular lines on an ordinary poh

7~
-'fjh
an ordinary point
ordinary
effects
line.
there
jiven
JLAUJJJJNJ.T.IOJN.
transformation of
TT
which
an involutic
on the absolute conic whose axis contains ordinary points is call* an orthogonal line reflection. transformation of TT which effects t
involution on the absolute conic

called a point reflection.
is
whose center
is
an ordinary point
called a displacement.
A product of two orthogonal line reflectioi A product of an odd number of orthogon

symmetry.
line reflections is called a
Two
figures
such that one
CE
be carried to the other by a displacement are said to be congruent, ar two figures such that one can be carried to the other by a symmeti
are said to be symmetric.
An
and
is
a harmonic homology
whose cent
are pole and polar with respect to the absolute conic. Sim the axis contains an interior point, the center is exterior and tl
.axis
(
involution effected on the absolute alters sense

it
74).
Conversel
follows from
alters sense is
74 that an involution on the absolute conic whk effected by a harmonic homology whose center
i.e. by an orthogonal line reflectio Since any direct projectivity is a product of two opposite involi tinns ( 74), the displacements as denned above are identical wit
exterior to the absolute conic,
On the other hand, the symmetries are the projective which carry the absolute into itself and interchange the two sense-classes on the absolute.
displacement.
collineations
From these remarks it is evident that the theory of displacements can be obtained from the theorems on projectivities of a conic in Chap. VIII, Vol. I, and in Chap. V, Vol. II. Some of the theorems may also be obtained very easily as projective generalizations of
simple Euclidean theorems. In proving these theorems we shall suppose that we are dealing with the real projective plane and not merely with an ordered plane
as in
Theorem
I.
It
would be
sufficient,
however, to assume merely
that every opposite involution is hyperbolic (i.e. that every line through an interior point of a conic meets it), for this proposition is the only
consequence of the continuity of the real plane which
we
use in our
arguments. Let us first prove that Assumption 66) of the Euclidean ( geometry holds for the hyperbolic geometry. It is to be shown
that
if
A,
B
A
are
two
on any ray
with an
to
end C there
is
a unique point
such that
AB
is
congruent
CD.
The points
absolute.
and C are the centers
of elliptic involutions
on the
It is
shown
in
76 that one such involution can be trans-
formed into any other by either a direct or an opposite involution.
Hence there is a displacement A carrying A to C. The absolute conic may be regarded as a circle C z in a Euclidean plane whose line at infinity is the pole of C with regard to the
absolute.
In this case
is
the center of the Euclidean circle, and
the hyperbolic displacements are the Euclidean rotations leaving C invariant. The required theorem now follows from the Euclidean
to proposition that there is one and only one rotation carrying The point is interior to of a ray having C as end. a point
C'
because
is.
Assumption XI,
66, holds good in the hyperbolic
geometry because
be proved for 66 for the Euclid-
the displacements form a group. Assumption XII the hyperbolic geometry by the argument used in
may
ean
case.
The same
is
true of Assumption
XIII
if
we
understand by
the mid-point of a pair AB the ordinary point which is harmonically separated by the pair AB from a point conjugate to it with respect
This theorem of course does not require full use of continuity It would also be valid if we assumed merely that any line through an interior point of a conic meets the conic (cf. 76).
assumptions.
at infinity or infinite points
ultra-infinite points.
DEFINITION. The points on the absolute are sometimes called points and the points exterior to the absolute,
;
132. Orthogonal lines, displacements, and congruence.
DEFINITION.
Two
lines (or
two
perpendicular to each other

absolute.
if
points) are said to be orthogonal or they are conjugate with respect to the
Of two perpendicular points one is, of course, always ultra-infinite, but no analogous statement holds for perpendicular lines. From the corresponding theorems on conies we deduce at once
THEOREM
4.
The pairs of perpendicular
lines
on an ordinary point
are pairs of a direct involution. Through an ordinary point there is one and but one line perpendictdar to a given ordinary line.
DEFINITION.
transformation of
TT
which
effects
an involution
on the absolute conic whose axis contains ordinary points is called an orthogonal line reflection. transformation of TT which effects an
hi volution
on the absolute conic whose center
is
an ordinary point
is
culled a point reflection.

is
called a displacement.
A product of two orthogonal line reflections A product of an odd number of orthogonal

symmetry.
line reflections is called a
Two
figures
such that one can
be carried to the other by a displacement are said to be congruent, and two figures such that one can be carried to the other by a symmetry
are said to be symmetric.
An orthogonal line reflection is a harmonic homology whose center and axis are pole and polar with respect to the absolute conic. Since the axis contains an interior point, the center is exterior and the
involution effected on the absolute alters sense
it
(
74).
Conversely,
follows from
alters
sense
is
74 that an involution on the absolute conic which effected by a harmonic homology whose center is
exterior to the absolute conic,
i.e. by an orthogonal line reflection. Since any direct projectivity is a product of two opposite involutions ( 74), the displacements as defined above are identical with
the protective collineations which transform the absolute conic into .itself with preservation of sense. In particular, a point reflection is a
132]
HYPERBOLIC GEOMETRY
On
353
the other hand, the symmetries are the protective collineations which carry the absolute into itself and interchange
displacement. the two sense-classes on the absolute.
From these remarks it is evident that the theory of displacements can be obtained from the theorems on pro jectivi ties of a conic in Chap. VIII, Vol. I, and in Chap. V, Vol. II. Some of the theorems may also be obtained very easily as protective generalizations of
simple Euclidean theorems.
In proving these theorems we shall suppose that we are dealing with the real projective plane and not merely with an ordered plane as in Theorem I. It would be sufficient, however, to assume merely
that every opposite involution
an interior point
of
is hyperbolic that every line through (i.e. a conic meets it), for this proposition is the only
consequence of the continuity of the real plane which
we use
in our
arguments. Let us first prove that Assumption 66) of the Euclidean ( geometry holds for the hyperbolic geometry. It is to be shown
that
if
A,
B
A
are
two
on any ray
with an
end
there
is
The points
absolute.
a unique point such that is congruent to CD. and C are the centers of elliptic involutions on the
AB
It is
shown
in
76 that one such involution can be trans-
formed into any other by either a direct or an opposite involution.
Hence there is a displacement A carrying A to C. The absolute conic may be regarded as a circle (7 2 in a Euclidean with regard to the plane whose line at infinity is the pole of absolute. In this case C is the center of the Euclidean circle, and
the hyperbolic displacements are the Euclidean rotations leaving C invariant. The required theorem now follows from the Euclidean
proposition that there
a point
2
is
one and only one rotation carrying
to
of
a ray having
is.
as end.
The point
I) is interior to
C'
because
Assumption XI,
66, holds good in the hyperbolic geometry because
the displacements form a group. Assumption XII the hyperbolic geometry by the argument used in
may be proved
for
66 for the Euclid-
DEFINITION.
pairs
A circle is
is
pair
If
OP 0P
where
the set [P] of all points such that the a fixed point are all congruent to a fixed
point point
the absolute be identified, as in the proof of Assumption above, 2 with its center, it is obvious that with a Euclidean circle C , and
as center are identical the circles of the hyperbolic plane having 2 Hence with the Euclidean circles interior to and concentric with <7
.
we
obtain from the properties of a pencil of concentric

(
Euclidean
circles
71)
5.
THEOEEM
DEFINITION.
circle
in the hyperbolic plane is a conic
entirely interior to the absolute.
It touches the absolute in two conjugate
imaginary points A, B, and the tangents at these points pass throug7i the center of the circle. The polar of the center passes through A and and is called the axis of the circle. All its real points are exterior to
the absolute conic.
have
134 (Theorem 7, Cor. 1) that two circles can. proved in most two real points in common. Once this is established, 66 applies without change to the the proof of Assumption XIV in
It will be
at
hyperbolic geometry.
134 as Cor. 2 of Theorem 7. proved in be proved as follows: Let A, B, C be three points in the order {ABC}, and let P and Q M be the points in which the line AB meets the absolute conic, the notation being
Assumption
Assumption
XV is
XVI may
assigned so that
in the order {IZ
of the pairs
1
we have {PÂBCQa}.
1
Let
B^ B B
Z,
&,
be points
ABB B B BB BJ3 etc.

2
lt
a,
congruent to each, Choose a scale (Chap. VI, Vol. I) in

is 1, co
such that
AB
which PnAQ* correspond to

coordinate of B.
0,
respectively,
and
let J
be the
the hypothesis about the order relations, 5 > 1. The displacement carrying is a projectivity of the line to V AB which leaves P* and Q<* respectively invariant and transforms
By
AB
BB
to B.
Hence
it
has the equation

x'
= bx
Qx
.
with respect to the scale
Pa
8
,
A,
The coordinates
of
B^ B2
The coordinate C is, by the hypothesis that {ABC}, some positive number greater than 6. There are at most, a finite number of values
-
s,
are therefore b
&
b*,
respectively.
of
c
of
.'.r.
6 (w
= l,
2,
between
b
J3
and
3>
c.
Hence there
between
are at
C.
most a
is
finite
is
i^v;
number
of the points
B lt
and
This
what
stated in
\Y.,,
.,y j;;
.;,,,.
:
y.A
v,;,^
for granted two results which will be proved in 29 and 66) 134, that all the assumptions (cf. of Euclidean plane geometry except the assumption about parallei lines are satisfied in the real hyperbolic plane, and that the parallel-line assumption is not satisfied.
We
Assumption XVI. have now seen, taking
EXERCISES
!<!.
,
,.,,,,,
1.
If
corresponding angles of two triangles are congruent, the correspondis
'
'
"'"*
J <iiJ:
ItW*
ing sides are congruent. 2. The absence of a theory of similar triangles in hyperbolic geometry <jue to what fact about the group of the geometry?
3.
The
.
perpendiculars at the mid-points of the sides of a triangle
meet in
a point (which may be ideal). *4 Classify the conic sections
from the point of view of hyperbolic geometry.
,,;,,. ;|^||fp
A.
77, Yol. I, hyperbolic displacements. According to Types any displacement has a center and an axis which it leaves invariant. If the center is interior, the axis meets the absolute in two conjugate imaginary points, and the displacement effects an elliptic transferof
133.
,L
1
,
mation on the absolute. If the center
is exterior,
the axis meets the
$ Dw*
r,
absolute in two real points, and the displacement effects a hyperbolic transformation on the absolute. If the center is on the absolute,
the axis
.
is
tangent,
and the displacement
effects
a parabolic trans-
formation on the absolute.
MS
,'
In the first case, the points into which a displacement and its powers carry a point distinct from its center are, by definition, on a circle which is transformed into itself by the given rotation. In the second case, since the displacement is a product of two orthogonal line reflections whose axes pass through the center, it is 2 obvious that the displacement leaves invariant any conic C which
touches the absolute in the two points in which it is met by the axis of the displacement. Such a conic is obtained from the absolute by
V
1
a homology whose center and axis are the center and axis of the displacement in question. From this it follows in an obvious way
that
a*
C2
is entirely interior
or entirely exterior to the absolute.
We
are interested in the case in
which C 2
is interior.
.Let
respectively.
are polar
cr,
tne points 01 contact 01 u wiui uie ausuiube jx ue JT aj and the line Since the center of the displacement 2 and Q are the ends of two segir with respect to C
,
2 T of points of C which are (in the hyperbolic plane) on opp and a point of the h) sides of the line PQ. Any line through bohc plane is perpendicular to PQ and meets cr, PQ, and r in t
points S, M,
respectively.
If
determined by another line through

the pair
M', T' are Jhe points analogc be the mid-poir 0, let
MM'. Then
the displacement which
is
the product of
as orthogonal line reflection with axis by that with OM' as azis carries
S,
OM
M,
T to
S',
M', T' respectively.
may be expressed by saying that cr is the locus of a point AS", on a given side of PQ, such that
This result
if M.' is
the foot of the perpendicular

1
from S' to PQ, S'M is congruent to 8M. For this reason cr and r are
called equidistantial curves of
PQ.
point
can be carried into a
point by a displacement leaving a given line I, not on A, invariant,

FIG. 78 and only if the two points are on the same equidistantial curve of I. The equidistantial curves some of the properties of parallel lines in the Euclidean geom
if
A
ment
(see
displacement which effects a parabolic transformation on

is
absolute
a product of
two orthogonal
line reflections
whose
intersect in the center
of the displacement.
Hence the
of
disp
leaves invariant
any conic which has contact
the third
47, Vol. I) with the absolute at 0.

case,
And by
the same reaso
as
employed in the second
a point
can be transformed
ii
point P' by a displacement which is parabolic on the absolute a fixed point at if and and P' are on a conic having coi only if
of the third order
with the absolute at
0.
conic interior to the absolute and having cor of the third order with it is called a Jioroeycle.
DEFINITION.
The
of
circles, equidistantial curves,
and horocycles are
all
path
ci
one-parameter groups of rotations.
134=.
Although the theory
Interpretation of hyperbolic geometry in the inversion plane. of conies of touching a fixed conic in
pairs
points has not been taken up explicitly in this book, we have in the inversion geometry a body of theorems from which the part of it needed for our present purpose can be obtained by the principle
of transference.
It has
been seen in
94,
Theorem
16, that
circle
the inversion group which carries a

is
any transformation
z
of
into itself effects a
protective transformation of this circle into
itself. Moreover, there one and only one direct circular transformation which effects a 2 Hence the group of direct circular transgiven projectivity on
formations leaving a
isomorphic with
circle
of the inversion plane invariant
the hyperbolic metric group,

is the
and
the
is simply geometry of this
subtjroup of the inversion group
2 The circles orthogonal to have the property that there is one and only one such circle through each pair of distinct points interior 2 to K' Since they also are transformed into themselves by the group which is here in question, it is to be expected that -they correspond
.
to the lines of the hyperbolic plane.
This
may be
proved as follows
Let the inversion plane be represented by a sphere S 2 in a Euclid2 2 be the circle in which S is met by a plane TT ean three-space. Let 2 through its center, and let us regard the points of TT interior to 1C
in
The circles of S 2 orthogonal to K z are those which S 2 is met by planes perpendicular to TT. Hence if we let 2 each point P of S 2 on one side of correspond to the point P' of TT such that the line PP' is perpendicular to TT, a correspondence F is
as a hyperbolic plane.
established between the hyperbolic plane and the points on one side 2 of a circle JK in the inversion plane in such a way that the lines
of the hyperbolic plane correspond to the circles orthogonal to
z
.
Moreover, since the direct circular transformations
of the inversion
2 plane are effected by three-dimensional collineations leaving S 2
in-
respond
plane.
variant, the direct circular transformations leaving 7f invariant corand symmetries of the hyperbolic under F to
displacements
Thus we have
6.
THEOKEM
tween
There
is
one-to-one reciprocal correspondence
be-
in 131 and the of a hyperbolic plane as defined 2 in an inversion plane (or inside a points on one side of a circle circle of the Euclidean plane) in which sets of collinear points of the
the points
2
,
and in which displacements and symmetries of the

to direct
7.
hyperbolic plane
correspond
circular transformations leaving
K'2 invariant.
THEOREM
In
the correspondence
to
the circles of the hyperbolic
plane correspond 2 on one side of 1C
circles
of the inversion plavie which are entirely
z z and let Proof. Let C be any circle entirely on one side of and 0' be the two points which are inverse with respect to both K? and <7 2 i.e. the limiting points of the pencil of circles containing C 2 and z ( 71, 96). In the Euclidean plane obtained by omitting z O' from the inversion plane, is the center of both and C 2 and
,
hence the direct circular transformations leaving

are the rotations about
and
(7
invariant
and the orthogonal
line reflections
whose
and
axes are on 0.
These correspond under

2
to the displacements
symmetries
of the hyperbolic plane
which leave
invariant.
Hence
the points of C correspond to a circle of the hyperbolic plane. Since any circle of the hyperbolic plane may be displaced into one
whose center corresponds under

shows that every
interior to
to 0, the
argument
just
made
circle of the hyperbolic
the correspondent under
plane may be obtained as of a circle of the inversion plane which is
2
.
circles
This theorem enables us to carry over a large body of theorems on from the Euclidean geometry to the hyperbolic. For example,
at
we have
once the following corollaries

1.
COROLLAKY
COROLLARY
Two
If
circles
in the hyperbolic plane can have at most
two real points in common.

2.
the line joining the centers
of two
circles in the
hyperbolic plane meets them in pairs of points which separate each other, the circles meet in two points, one on each side of the line.
The
first of
these corollaries, on comparison with
Theorem
5,
yields
the following projective theorem : Two conies interior to a real conic and touching it in pairs of conjugate imaginary points can have
at most two real points in common, imaginary points in common.
and always have two
conjugate
THEOREM
8,
In
the correspondence
equidistantial curves of the
hyperbolic plane correspond to those portions of circles intersecting K?, 2 not orthogonally, which are on one side of Two equidistantial
of
curves which are parts of one conic in the hyperbolic plane are parts circles inverse to each other with respect to K*.
Proof.
circle
K*
of
S 2 which
intersects
1C
in two points P,
without being perpendicular to
it is
a section of
S 2 by
a plane not
perpendicular to TT. The correspondence F transforms this circle into 2 a conic section C in TT which is the projection of K* from the point at infinity of a line perpendicular to TT. The and tangents to K* at
P
z
are transformed into tangents to

Q.
2
.
Hence C 2 touches
at
and
The portions
two segments
2
of
on the two
sides of
of
C 2 having
and Q as ends
2 which is in correspondence with portions of If is on the side of the hyperbolic plane by means of F. The segment of C 2 which is
2 not in correspondence with this portion of is evidently in corre2 is transformed spondence with a portion of the circle into which
on S correspond to the but only one of these
by the three-dimensional orthogonal

fixed points.
reflection,
with
TT
as plane of
which
2 This proves that the part of any circle of the inversion plane 2 is on one side of corresponds under F to an equidistantial curve I} and that that part of the circle inverse to K* with respect to
which
is
on the same side

is
curve
which
distantial curve is
of corresponds to the equidistantial part of the same conic with E^ That any equiin correspondence with a portion of some circle of
is easily
the inversion plane the last theorem.
proved by an argument like that used in
COROLLARY
sponds
to
1.
In
the correspondence
T a
circle
touching
corre-
a horocycle of
the hyperbolic plane.
Since each equidistantial curve corresponds to a portion of a circle

of the inversion plane, it follows that
two equidistantial curves can
have at most two real points in common. It must be noted that two conies containing each an equidistantial curve can have four real
points in
curves.
common,
since each conic accounts for
two
equidistantial
In like manner two horocycles can have at most two real points
in
common, and,
COROLLARY
2.
fiuriiK
still
more
loci
generally,
distantial
each of which is a circle, Jiorocycle, or r.an have at m.nst two 'nninf.a in common.
Two
EXERCISES
1.
Show that r may be extended
bolic plane correspond to z orthogonal to
imaginary
so that the ultra-infinite lines of the hypercircles of the inversion, plane which are
Study the theory of pencils of circles, equidistantial curves, and horothe cycles in the hyperbolic plane by means of the correspondence I\ (A list of theorems will be found in an article by E. Ricordi, Giornale di Matematiche, Vol. XVIII (1880), p. 255, and in Chap. XI of Non-Euclidean Geometry by
2.
J. L. Coolidge,
3.
Oxford, 1909.) Develop the theory of conies touching a fixed conic in pairs of points.
In 135. Significance and history of non-Euclidean geometry. proving the two corollaries of Theorem 7 we have completed the 66 are satisfied proof ( 132) that the congruence assumptions of
in the hyperbolic plane.
Combining
this result
with Theorems 1
and
2,
we have
9.
THEOKEM
In
the real hyperbolic
I
in
VI, VII,
X~ XVI of the
are true,
plane geometry, Assumptions assumptions for Euclidean plane geometry
29 and 66
CoKOLLARY. Assumption
plane geometry.
and Assumption XVII of 29
IX is false.
is true
in the hyperbolic
of the hyperbolic geometry therefore furnishes a proof * independence of Assumption IX as an assumption of Euclidean geometry. This assumption is equivalent to, though not identical in of the
The existence
parallel postulate
form with, Euclid's parallel postulate.! And it is the interest in the which has been the chief historical reason for the
development of the hyperbolic geometry. The question whether the postulate of Euclid was independent or not was raised very early. In fact, the arrangement of propositions in
worked on the question himself. The effort to prove the postulate as a theorem continued for centuries, and in the course of time a considerable number of theorems were shown to be independent of this assumption. Eventually the question arose, what sort of theorems could be proved by taking the contrary of
Euclid's Elements shows that he had
Euclid's assumption as a
new
assumption.
*Cf
2,
Vol.
I,
I,
and
13, Vol. II.
tCf. Vol.
p. 202,
of
Heath, The Thirteen Books of Euclid's Elements,
first
This question seems to have been taken up systematically for the time by G. Saccheri,* who obtained a large body of theorems
_
on
this basis,
least publishing,
but seems to have been restrained from drawing, or at more radical conclusions by the weight of religious
disapproval. The credit for having propounded the body of theorems based on a contradiction of the parallel postulate as a self-consistent
i.e. as a non-Euclidean geometry, belongs to Bolyait (1832) and N. I. Lobachevski (1829), although many of the ideas involved seem to have been already in the possession of
mathematical science,
J.
C. F.
Gauss.
||
It
was not, however, until
it
had been shown by
that the hyperbolic plane geometry could be regarded as the geometry of a pseudospherical surface in Euclidean space, that an
Beltrami
independence proof (cf. Introduction, Vol. I) for the parallel assumption could be said to have been given. The work of Beltrami depends
on the investigation by EiemannH

at the basis of
of the differential
geometry (1854).
geometry ideas Kieinann seems to deserve the geometry
credit for the discovery of the elliptic
(141-143
below),
though
it is
not clear that he distinguished between the two types

of the existence of a
of elliptic
geometry.**
The proof
non-Euclidean geometry was made
capable of a simpler form by the discovery of A. Cay ley ft (1859) that a metric geometry can be built up, using a conic as absolute. The
relation of Cayley's
work
to other branches of geometry and the pre-
vious studies of non-Euclidean geometry was made plain by F. Klein tt in connection with his elucidation of the rdle of groups in geometry. The representation of the hyperbolic plane by means of the interior
* Euclides ab omni naevo vindicatus, Milan, 1733. German translation in " Die Theorie der Parallellinien von Euklid bis auf Gauss," by P. Engel and P. Staeckel,
Leipzig, 1896. t English translation
t
by G. B. Halsted, under the

Texas, 1896.
title
"The
Science Absolute
of Space," 4th ed., Austin,
German
translation
by Engel, under the
lungen," Leipzig, 1898. Cf. also a translation "The Theory of Parallels," Austin, Texas, 1892. Werke, Vol. VIII, pp. 157-268. Saggio di interpretazione della geometria non-euclidea, Giornale di Matema||
"Zwei geometrische Abhandby Halsted of another work entitled

title
tiche, Vol.
IT
VI
(1868), p. 284.
Clifford's
.
W. K. Clifford, in Nature, Vol. VIII (1873), and in "Mathematical Papers" (London, 1882), p. 55. ** Cf F. Klein, Autographierte Vorlesungen iiber nicht-euklidische Geometrie,
English translation by
Vol. I (GOttingen, 1892), p. 287. t-i- Collected Works, Vol. II (Cambridge, 1889), p. 583.
Other texts in English are
J. L.
Coolidge, Non-Euclidean
Geometry,
Oxford, 1909; Maiming, Non-Euclidean Geometry, Boston, 1901; D. M. Y. Sommerville, The Elements of Non-Euclidean Geometry,
London, 1914; H. S. Carslaw, The Elements of Non-Euclidean Plane Geometry and Trigonometry, London, 1916. Besides these we may mention D. M. Y. Sommerville' s Bibliography of Non-Euclidean
Geometry, London, 1911. There are numerous other geometries closely related to the nonEuclidean geometries touched on in. this chapter. Of particular
interest
are
the geometries
associated with
Hermitian forms in-
vestigated
(1904),
p.
by G. Fubini (Atti del Eeale Istituto Veneto, Vol. LXIII 501) and E. Study, t and the geometry of the Physical
Theory
of Relativity.
136. Angular measure.
The measure
of angles
may
be defined
defi-
precisely as in the Euclidean geometry,
and we carry over the

2
nitions
69 without modification. If we represent the absolute and an arbitrary point by a Euclidean circle C" and
and theorems
of
its
center, the Euclidean rotations about
are identical with the
hyperbolic rotations about 0, and hence the two angular measures as determined by the method of 69 are identical. By 72, if a and b
are
two
lines intersecting in 0,
and 6
is
the measure of the smallest
angle 4.
AOB
for
which
is
a point of a and
a point of
b,
(1)
=lines
where \ and
are the
2
minimal
through
0.
Since
and
are
the tangents to C through 0, it follows that (1) may be taken as the formula for the measure of any ordered pair of lines a, "b in the
* Atti della R. Accademia del Lincei, Ser. 3, Vol. II (1877-1878), p. 31. t Acta Mathematica, Vol. I (1882), p. 8, and Bulletin de la Soeie'te' mathematique de France, Vol. (1887), p. 203. J Matliematisclie Annalen, Vol. (1905), p. 321. Of. F. Klein, JahresberichtderDeutschenMathematiker-Vereini,suncT. Vol. XIX 48 above, {1910), p. 281, and the article by Wilson and Lewis referred to in
XV
LX
hyperbolic plane if i l and is are understood to be the tangents to the absolute through the point of intersection of a and I. If the hyperbolic plane is represented as in 134 by the interior
identical with the Euclidean
the angular measure of any two hyperbolic lines is measure of the angle ( 93) between the 2 two circles orthogonal to (7 which represent them. This has just
of a circle
,
C2
been seen
center of
circles
where the two circles are lines through the In the general case a point A of intersection of the two 2 2 orthogonal to C may be transformed to the center of C by a
for the case
.
C2
direct circular transformation A.
The transformation
as a direct
circular transformation leaves Euclidean angular

(
measure invariant
93),
and
as a displacement of the hyperbolic plane leaves hyper-
bolic angular
measure invariant.
Z
Since the two measures are identical
at the center of
<7
they must also be identical at A.

this
As an
application of
result
we may prove
a triangle
the following
remarkable theorem:
THEOREM
Proof.
10.
The sum of
the angles of
is less
than
TT.
point A the hyperbolic lines
Let the triangle be ABC, and let the absolute and the 2 be represented by a Euclidean circle C and its center. Then
AB and AC
2
are represented
by Euclidean
lines
through the center of C" , and the hyperbolic 2 line BC is represented by a circle through
and C orthogonal to C (fig. 79). The hyperbolic measures of the angles
at A,
B, and
respectively are equal to the Euclidean
measures of 4.JBAC and two angles formed by AB and AC with the tangents to 2 at B and C
FicT~79
respectively.
three angles is of the Euclidean easily seen to be less than that- of the angles
of these
(rectilineal) triangle
The sum
ABC. Hence it is less than TT. The theorem that the sum of the angles of a triangle
is TT
may
be substituted for Assumption IX as an assumption of Euclidean geometry;* the proposition just proved can be taken as the correthat the
and the proposition sponding assumption of hyperbolic geometry; sum of the angles of a triangle is greater than TT can be
*
taken as an assumption for elliptic geometry.
On
will also
the history of this theorem cf Bonola, loc. cit., Chap. II. be found useful in connection with the exercises.
.
This reference
EXERCISES
*
of
1.
Prove from Assumptions I- VI,
X-XVT
that
if
the
sum
of the angles
one triangle is greater than, equal to, or less than TT, the corresponding statement also holds for all other triangles. *2. Prove from Assumptions I-VI, X-XVI that the sum of the angles of
a triangle
is less
than or equal to
TT.
137. Distance.
Since the conic section
is
a self-dual figure,
it is
to
be expected that the formula for the measure of point-pairs is analofactor gous to (1). As a matter of fact, we shall only modify the
i/2.
If
and
in
which the
line
B are two ordinary points, let A x B x be the points AB meets the absolute, the notation being assigned
,
so that the points are in the order

is
{A,ABBm }. Then T^(AB,A n B x
and hence log R (AB, A^B*) has a real value. define the distance between A and B by means of the equation
positive
(
24),
We
(2)
Diet (AB)
is
= 7 log ft (AB,
XJ?),
real determination of the
where 7
an arbitrary constant and the
logarithm is taken. It is seen at once that

Dist (45)
= Dist (.R4),
t
because
$ (AB, A m B )= B (BA Bx A m ),
a)
and that
if
A, B,
are collinear points in the order
{ABC},
Dist (AB)
because
K(AB, A m Bx
it is
+ Dist (BC) = Dist (A C), &(BC, A X BX )=^(AC, A X B X
).
evident from the properties of the collineations Moreover, transforming a conic into itself that a necessary and sufficient condition for the congruence of
two point-pairs AB,

ft
CD
is
&(AB, A x B m ) =
where
(CD,
,
0,-D,,),
A x B^
,
are chosen as above
and Cn
Dx are chosen
analogously.
of
Hence
and
a necessary
is
and
sufficient condition for the
congruence
AB
CD
Dist(^J?)= Dist(CZ>).
distance function denned above
(
Hence the
is
fully analogous to that
used in Euclidean geometry
67).
The constant 7 may be determined
by choosing a fixed point-pair

(3)
OP as the unit of distance. We then have
-=
logB(0P, 0..).
138. Algebraic formulas for distance and angle.

the symmetric bilinear form
J
**{*<$-)
Let us consider
=:
)
"'oo
'
^'Ol^'O'^'l
'
^02^1)^2
-\_1_
a x x
n
-\'
a x
x*
'
-\_1_
a x
x^
'
_1_
and the covariant form

F(u,
u')=A 00 u
u'
+A
ol
u Q u'
85. With respect to homogeneous the equation of a point conic, and F(u, w)=0 of the line conic composed of the tangents to f(X, X] = 0. Let us
where the
Ay's are
denned as in
is
coordinates, f(X,
X) =
take this conic as the absolute and derive the formulas for the
measure
Let
of distance
Q,
and
of angle.
Y=(y
y l} y2 ) and
^=(
them
points of the line joining
\, are
,
2)
be two distinct pouits.
The
\Y + pZ = (Xy + /!,
and the points in which
by the values
this line
\y
fi#v
tyz + /w 2 ),
are determined
meets f(X, X~) =
of A,//* satisfying the equation

2
=/(AX+ A*Z, XF+ A*Z) = X /(F,

These values are
F)
+2
Z)-f(Y, Y}f(Z, Z]
A*,
/(^ ^)
,
Let us denote the two points of the absolute corresponding to (\ 1
and (X2
/* 2 )
by 71 and 72
respectively.
Then
J/,).
Dist(FZ)=7lpgR(F^
Since
(X, p) is (1, 0) for
F and
(0, 1)
for Z,
we have
65,
VoL I)
Hence
(4)
,
Z)
log &
>
(F, ^)
T
-/(F,
j? /-rr
By
precisely the
for the
measure
of a pair of lines
v) =
'
same reasoning applied to the dual u = (u u^ u 2 ), v = (v

,
, ,
case
,
we have
a ).
v :,
(5) '
v
m (u,
V
log &
v)-F(u, u)F(v,
v)
"
__li ~
2
F(u F (u (
,
>
g
d,
Denoting Dist(F, Z) by
(b)
e
we
obtain
a
f(Y,
-
z-)+v/
(r.
z)-/(r. r)/(s, z)
and hence
(?)
Codl
2T
_ 2
and
In like manner,
/ON iyi
e
if
= m (uv),
V, V)
,.,
^K
i
,/)
-^(^
iy
/}0)
F(u, u}F(v, v)
F(u, u)F(v,
1
v)
v)
jp
^, u}F(v,
For a further discussion
of these
formulas see Clebsch-Lindemann,

of
Vorlesungen iiber Geometrie, Vol. II, Part III, Leipzig, 1891. *139. Differential of arc. The homogeneous coordinates
points not
/i
all
on the
absolute,
2^
f(X
X\~^~-
may
be subjected to the relation
(13)
f(X,X)~C,
is
where
a constant.
Since f(X,
X]
is
quadratic, this determines
two
sets
of coordinates (#.. x..
xn \
for each point of the hyperbolic
plane instead of an infinity of sets as in unrestricted homogeneous

coordinates.*
Some
definite determination of the values of

is
each of the homogeneous
manifestly necessary in order to apply the processes of differential calculus to formulas in homogeneous coordinates. The particular relation
coordinates
f(X, X)
C has the advantage, point not on the absolute.

Suppose
among
others, of not being singular for
any
now
condition that XQ x lf x2 are functions of a parameter

,
that (XQ xl} xn ] describes a locus determined by the t. Then, in the

,
familiar notation,!
ds__
dt ~AJ=O
Pist(X,X+AX)
A
2
7 sinh -- Dist (X,
X+ AX)
= L
= Z i2
A-O A*
by
(8).
2
2
)/ (X,
X+ AX) -/(X, X)/(X+ AX, X+ AX) x+ AX)

;
f ,
Since
,
f(Y+ Y'
2
Z) =f(Y, Z) +f(Y Z), this reduces to
ds\
I
== -"
47
/(X,X) -f/(X, AX)) -/(X,X)(/(X,X) +2/(X,AX)+/(AX, AX V) f(X,X) (f(X, X) + 2/(X, AX) +/(AX, AX))
2
AX\
Y
2/(X,
./AX AX\
j
=
A ( =o
'
y2
/(^^)(/(^^)+ ^X
N
AX)+/(AX, AX))
f(X, X)
* If x t x 2 ) are interpreted as rectangular coordinates in a Euclidean space (x C is the equation of a quadric surface, and we have of three dimensions, f(X, X] a correspondence in which each point of the hyperbolic plane corresponds to a pair of points of the quadric surface. By properly choosing f(X, X), this correspondence can be reduced to that given in 134 between the hyperbolic plane and the
, ,
surface of a sphere. t are applying theorems of calculus here on the emnloved algebraic theorems in other parts of the work.
We
same
basis that
we have
in
which
...
dX
dt
represents *
\dt
dx2 dx -r
*
dx\
>
-r^
In
, .. ,.. ... ,, differential notation this

.
dt
dtj
formula
(14)
is
&- 1 v
d6 z
/2 (X dX}
>
~{
z
(X
'
X ^( dX dX
>
f (X,X]
By
angle,
duality
we have
a corresponding formula for the differential of

'
(15)
= F Û
du }~ F ( u u
'
">
F du
(
>
du
">
F*(u, u)
These formulas are independent of .the particular determination of our coordinates by means of the relation (13). If we differentiate (13)
we
obtain
so that for this particular determination of coordinates
(16) ^ '
Let us
now
choose the homogeneous coordinate system so that
and choose
(17)
If
C=
47
so that, for points not on the absolute,
xl
is
%l
4 72 # 2
4 72
real
and not zero, we are dealing with hyperbolic geometry, and

ds
z
(18)
=f(dX, dX}
= dx? + dx%
If
4 ffdx~Q.
v
we
substitute
'-^
>
in the value for ds* given in (18),

(19)
we
obtain
[Regarding % and v as parameters of a surface in a Euclidean space, (19) gives the linear element of the surface (cf. Eisenhart, Differential Geometry,
is
30).
This
is
a surface for which, in the usual notation of
differential geometry,
E= G and ^=0.
l/47
it
2
The curvature
of this surface
loc. cit.,
constant and equal to Vol. II, p. 525). From this
(cf.
Clebsch-Lindemami,
follows that the
hyperbolic plane
geometry in the neighborhood of any point is equivalent to the geometry on a portion of a surface of constant negative curvature. If we substitute u = # /# and v = in (18), we obtain a 2
_. ~
(4
- v*)du + 2 uvdudv + (4 72 2
This
is
cited above.
the form of linear element used by Beltrami in the paper This form is such that geodesies are given by linear
equations in u and v. Hence geodesies of the surface correspond to lines of the hyperbolic plane. It is to be noted that the curvature of a surface, while often defined
in terms of a Euclidean space in which the surface is supposed to be situated, is a function of E, F, and G and therefore an internal property
of the surface,
i.e.
a property stated in terms of curves (u
=e
and
c)
in the surface
and entirely independent
of its being situated
in a space.
Another remark which may save misunderstanding by a beginner that the geometries corresponding to real values of 7 are identical. The choice of 7 amounts to a determination of the unit of length, as was shown in 137.
is
EXERCISES
Express the differential of angle in terms of (x xv zc ) and their derivatives (cf. Clebsch-Lindemann, loc. cit., Vol. II, p. 477). *2. Develop the theory of areas in the hyperbolic plane. For a treatment
1.
,
by differential geometry cf. Clebsch-Lindemann, loc. cit., p. 489. For a development by elementary geometry of a theory of areas of polygons which is equally available in hyperbolic, parabolic, and elliptic geometry, see A. Finzel, Mathematische Annalen, Vol.
LXXII
(1912), p. 262.
140. Hyperbolic geometry of three dimensions.

of three dimensions is the interior
(cf.
A hyperbolic
space
129) of a nonruled quadric surface, called the absolute quadric, and the hyperbolic geometry of three dimensions is the set of theorems stating properties of
2,
Ex.
this space
which are not disturbed by the projective collineations
leav-
ing the quadric invariant. The definitions of the terms "displacement," " " congruent," perpendicular," etc. are obtained by direct generalization of the definition in chapters on Euclidean geometry.
132 and the corresponding definitions in. the They will be taken for granted in
3TO
XON-EUCLIDEAN GEOMETBIES
[CHAP.VIH
The fundamental theorems on congruence may be obtained from the observations (1) that any displacement of space leaving a plane invariant effects in this plane a displacement or a symmetry in the sense of 132, and (2) that no two displacements of space leaving a
plane invariant effect the same displacement or symmetry in this
plane.
From
this
11.
we
In
infer,
by reference
to
135,
THEOREM
the
real
Assumptions I-XV1 of tion IX, which is false.
three-dimensional hyperbolic geometry 29 and 66 are all true except Assump-
100 there is a simple isomorphism between the displacements hyperbolic space and the direct circular transformations of the inversion plane. Hence the theorems of inversion geometry or of the
By
of a
theory of projectivities of complex one-dimensional forms can all be translated into theorems of hyperbolic geometry. The reader who carries this out in detail will find that many of the theorems of
Chap.
VI assume
if
very interesting forms
when
carried over into
the
In particular,
the absolute,
of
it
as a line reflection (
line reflection, or half turn, is defined 101) whose directrices are polar with respect to follows at once that every displacement is a product
an orthogonal
line reflections. With this basis the theory of displacements is very similar to the corresponding theory in Euclidean geometry, but many of the proofs are simpler.
two orthogonal
The formulas for distance and angle are identical with those of 139 if f(X, X') 138, and the differential formulas with those of
be understood to be a bilinear form in (x
,
xv xz) #8 ) and
(x' , x[, x(, a^).
EXERCISES
and only if their three ordinary directrices have a common intersecting perpendicular line. 2. Tf a simple hexagon be inscribed in the absolute, the common inter1.
The product
of three half turns is a half turn if
secting perpendicular lines of pairs of opposite edges are

.
met by a common
intersecting perpendicular line (cf 108). 3. Determine the protectively distinct types of displacements.
*5. Classify the quadric surfaces from the point of view of hyperbolic
geometry. *6. Given the existence of a hyperbolic space, define a set of ideal points euch that the extended space is projective. Cf R. Bonola, Giornale di Mate.
matiche, Vol.
American
(1900), p. 105, and F. W. Owens, Transactions of the Mathematical Society, Vol. XI (1910), p. 140.
XXXVIII
* 7. Obtain theorems 122, 123 with analogous to those in the exercises of regard to the hyperbolic displacements. *8. Study the theory of volumes in hyperbolic geometry by methods of
differential geometry.
141. Elliptic plane geometry. Definition. The geometry corresponding to the group of projective collineations in a real* projective z invariant is called the plane TT which leave an imaginary ellipse
two-dimensional
aginary conic
7<7
elliptic
2
is
geometry or elliptic plane geometry. The imcalled the absolute conic or the absolute. The
projective plane TT is sometimes referred to as the elliptic plane. The order relations in this geometry are of course identical with
those of the projective plane (Chap. II). The congruence relations are defined as in 132, with suitable modifications corresponding to
the fact that JS
allel to
is imaginary. Some of the theorems which run parthe corresponding theorems of hyperbolic geometry are put in the following list of exercises.
2
down The formula
may
for the measure of angle used in hyperbolic geometrybe taken over without change, i.e.
= m (ijj = -
log
B (y
s,
2 ),
where ^ and
and \ and iz are tangents to z the absolute in the same flat pencil with l t and l f The formula for distance may also be taken from hyperbolic geometry
l
are intersecting lines
= Dist (PQ) = 7 log E (PQ, H <?.
).
In order that this shall give a real value for the distance between two real points, 7 must be a pure imaginary. So we write
Tci
7
*TMs geometry
Ji
can in large part be developed on the basis of Assumptions A, E, S, P alone, the imaginary conic being replaced by the corresponding elliptic 89. As a polar system, the existence and properties of which are studied in matter of fact there is considerable interest attached to the elliptic geometry in a modular plane, but the point of view which we are taking in this chapter puts
and in order
choose
Ic
to
have formulas in the simplest possible form, we
may
= I,
so that
= ~i log
is
The discussion
if
in
138
f(X, X'} be taken
to be a bilinear
applicable at once to elliptic geometry form in three variables such
that /(X,
X) =
is
the equation of the absolute of elliptic geometry.
Thus we have
<
20)
(21)
..
cos- =
(22)
EXERCISES
1.
2.
The The
principle of duality holds good in the elliptic geometry. elliptic geometry is identical with the set of theorems about
the
geometry of the plane at infinity in three-dimensional Euclidean geometry. 3. The pairs of perpendicular lines at any point are pairs of an elliptic
involution.
4.
The
lines perpendicular to a line
to the absolute.
I all meet in the pole of I with respect Through any point except the pole of I there is one and but
one line perpendicular to I. 5. Defining a ray as a segment whose ends are conjugate with respect to the absolute, prove that Assumption X, 66, holds in the single elliptic geometry if the restrictions be added that A and B are on the same ray.
for
Assumptions XI and XIII of 66 hold for single elliptic geometry. How may Assumptions XII, XIV, and XV be modified so as to be valid single elliptic geometry ? A 8. circle is a conic toiiching the absolute in two conjugate imaginary
6.
7.
points.
9.
circle is the locus of a point at a fixed distance
from a
fixed line.
10. If
A, B,
C are
three collinear points,
Dist (A,
5) + Dist (BC) + DiBt(C4)
=
ir.
TT.
In other words, the total length of a line is IT. 11. The sum of the angles of a triangle is less than
and
defining as collinear any set of elliptic points
on a
circle
C2
142. Elliptic geometry of three dimensions. The three-dimensional elliptic geometry is the set of theorems about a three-dimensional
projective space which state properties undisturbed by the projective collineations leaving invariant an arbitrary but fixed projective polar
polar plane.
system, called the absolute polar system, in which no point is on its It is a direct generalization of the elliptic geometry of
the plane and may be based on a similar set of assumptions. In a real space this polar system is that of an imaginary quadric (called the absolute quadric) with respect to which each real point
has a
real polar plane,
and the equation

8
aJ 2 1 a a
of the absolute quadric
may
be taken to be
+aj +aj .+ a,'=0.
displacement is denned as a direct* projective collineation 32) which leaves the absolute polar system invariant ; a symmetry is defined as a nondirect projective collineation leaving the
(cf.
absolute polar system invariant.
The
definitions of congruence, per-
pendicularity, distance, etc. follow the pattern of the hyperbolic and parabolic geometries, and the same method may be used, as in those
geometries, to extend the theorems on congruence from the plane

to space.
It can easily be proved by means of the theorems on the quadric in Chap. VI that any displacement is a product of two line reflections whose axes are polar with regard to the absolute. From this proposi-
tion a series of theorems on displacements can be derived, just as in
the parabolic and hyperbolic geometries. Through a given point not on a given line
to
I
there
is
no
line parallel
in the sense in which the term
geometry.
of
There
is,
used in parabolic or hyperbolic however, a generalization of the Euclidean notion

is
parallelism to
elliptic
many
more
of the properties of
three-dimensional space which preserves Euclidean parallelism and is, if possible,
interesting.
* Without appealing to order relations, the direct collineations may be characterized as those which do not interchange the reguli on the absolute quadric.
and through these points there are two lines p v p z of one regulus and two lines q v q z of the other regulus. The lines p l} p 2 are conjugate imaginary lines of the second kind ( 109), and I is one line of an.
elliptic
congruence of which
p p2
lt
are directrices.
.
similar
remark
Any line of the applies to the conjugate imaginary lines q 1} q z or q q z as directrices is called elliptic congruences having p l} p z lt
a Clifford parallel* of
Clifford parallels to
parallel to itself.
I
I or a paratactic^ of I. through any point not on
Thus
I,
there are
I
two
and
is
a Clifford
The two
Clifford parallels to
any
:
line
may
lines
be distinguished as follows
Let
E^ and R% be
through any point not on it the two reguli on
the absolute.
Two real lines I, m meeting two conjugate imaginaryEI are right-handed Clifford parallels, or paratactics and two real lines I', m' meeting two conjugate imaginary lines q lt q z of
p pz
lt
of
R$
lels
are left-handed Clifford parallels, or paratactics.
and left-handed Clifford paralbe drawn entirely in terms of real elements by means of the notion of sense-class ( 32), and thus connected with the intuidistinction between right-handed
The
may
tive distinction
between right and left. This matter will he taken up again in the next chapter. In the meantime it may be remarked that the definition in terms of the two reguli on the absolute is indeall
pendent of
A, E,
question of order relations and
is
based on Assumptions
P alone. From the definition

m,
is
Clifford parallel to
if
is a right-handed if a right-handed Clifford parallel to I that also a right-handed Clifford parallel to n, I is a right-handed
it
follows immediately that

is
Clifford parallel to n.
In general, two lines have one and only one
common
intersecting perpendicular; but if they are right-handed Clifford parallels, there is a regulus of common intersecting perpen-
diculars,
and the latter are all left-handed Clifford parallels. The product of two orthogonal line reflections whose axes are
congruence
of Clifford parallels
is called
Clifford parallels leaves each line of the
perpendicular to the axes invariant, and

* Cf
.
a translation.
Preliminary Sketch of Biquaternions, Mathematical Papers (London, 1882), p. 181, and Klein, Autographierte Vorlesungen liber nicht-euklidisohe Geometrie, Vol. II (Gottingen, 1892), p. 245. t E. Study, Jahresbericht der Deutschen Matbematikervereinigung, Vol. XI
Clifford,
(1903), p. 319.
translation
is
right-handed or left-handed according as the congruence
of its invariant lines is right-handed or left-handed.
Any
displace-
ment can be expressed

For a discussion
as a product of
of Clifford
two translations. parallels and related questions
see
Appendix II of the book by Bonola referred to above, F. Klein, Mathematische Annalen, Vol. XXXVII (1890), p. 544, and the other
references given above in this section.
143. Double elliptic geometry. The geometry corresponding to 2 the group of projective collineations transforming a sphere in a Euclidean three-space into itself is called spherical or double elliptic plane geometry. The sphere S is called the double elliptic plane. The
circles in
which
/S
is
met by planes through

if
its
center axe called
lines,
conjugate under the group of direct projective collineations transforming the sphere into itself.
and two
figures are said to be congruent
The plane which

described.
is
called elliptic in
141
is
sometimes called
plane here Since the plane at infinity TT* of a Euclidean space is a 3 single elliptic plane, and since each line through the center of S 2 meets S in two points and TT in one point, there is a correspondence
single elliptic to distinguish it
elliptic
from the double
between a single elliptic plane and a double each point of the first corresponds to a pair
elliptic plane, in
which
of points of the latter.
of this correspondence any result of either geometry can be carried over into the other geometry. These remarks can all be generalized to ^-dimensions. For a set
By means
of assumptions for
J.
double elliptic geometry as a separate science, see E. Kline, Annals of Mathematics, 2d Ser., Vol. XIX (1916), p. 31. 144. Euclidean geometry as a limiting case of non-Euclidean. In
the two-dimensional case

lute
we have seen
that the equation of the abso-
may
be taken as
or in line coordinates, as
g
/O K\
\
I
I nt 2
n 1 2\
^^
f\
4 72
of hyperbolic
The formulas
and
(25)
of elliptic
geometry
is
arise if
is real
and not
^
zero,
geometry
if
imaginary.
If
we
set c
~ 0,
may
be regarded as the equation of the circle at infinity of the
geometry
c
ill
UU
lUIUl U&CU Ul
4.
J.VXUICUVCA,
J.A
VYG
oou
in the formulas of
138 and 141, we obtain
j and
- log -J 2 wî +
i 1
^-
(^v 2
- ujfj
72. which, agree with the formulas of Euclidean geometry given in in the formula for the differential of In like manner, if we set c
distance in
139,
we
obtain ds*
these remarks to three or

If c
= duz -\- dv n dimensions is of
The generalization of course obvious.
value,
is
changes by continuous variation from a positive to a negative must pass through zero. Since the corresponding geometry elliptic while c is positive, parabolic when c is zero, and hyperbolic
it
c is negative, the parabolic geometry is often spoken of as a uniting case both of elliptic and of hyperbolic geometry. This point of view is reenforced by observing that the formula (10) makes the measure of a fixed angle a continuous function of c, so that
while
for a small variation of c the value given
by
(10) for 6 suffers a
like remark can be correspondingly small variation. the distance between a fixed pair of points.
made about
This has the consequence that for a given figure consisting of a number of points and lines, and for a given number e, a num8 and B, the disber 8 can be found such that if c varies between
finite
tance of point-pairs and the angular measure of line-pairs of
F do not
c
vary more than
e.
Nevertheless, in this interval of variation of
the
geometry according to which the distances and angles are measured changes from elliptic through parabolic to hyperbolic.
For example,
if
F were
a triangle, and the

to
sum
TT
of the angles
were
e,
found by physical measurement
be between
+e
and
TT
the
geometry according to which the measurements were made might be either parabolic, hyperbolic, or elliptic. Further refinements of
experimental methods might decrease e, but according to current physical doctrine could not reduce it to zero. Hence, while experiment might conceivably prove that the geometry at the bottom of
the system of measurements was elliptic or hyperbolic,
it
could not
prove
it
to be parabolic.
For the details
of
showing that the Euclidean formula for distance
is
a limiting case of the non-Euclidean formula, see Clebsch-Lkidemann,

loc. cit.,
Vol. II,
p.
530.
of elliptic displacements. Suppose the coordinate system so chosen that the equation of the absolute is
145.
Parameter representation
The projective collineations which leave the lines of a regulus on the absolute invariant have been proved to have matrices of the form 126. Let JSjf be the regulus on the absolute left in(50) or (51) in
variant
by those of type
2 by the transformations of type (50), and -K 2 that (51). The transformations of type
left invariant
(50)
are the
systems of right-handed Clifford parallels invariant, and those of type (51) the translations leaving systems of
left-handed Clifford parallels invariant. Since any transformation leaving the quadric invariant
of
is
translations leaving
a product
lines of
one leaving the lines

invariant,
of It* invariant
is
by one leaving the
R$
any displacement
a product of a transformation of
type (50) by one of type (51). Denoting (50) by the matrix A of any displacement can be written
and (51) by B,
If
A' = B'A'
is
the matrix of a second displacement, and B' and A'
are of the types (50)
and
A'
.
(51) respectively,
.
(27)
A = B'A'BA = B'B
all
A'A,
2
because any displacement leaving

rrmf.aH\-o
irrif.Ti
lines of -R
nil
invariant
7? 2
is
com-
onir
rliaT-vlar>OTYiont.
loQvinrr
lines nf
invarinnh
#2 ^8
is
^20^0 ~f~
==
a 2i a a
'l
'
^22^2
~^~
a 2S X8>
^80^0
~f~
~^~
^82*2 ~f~ ^88^8
given parametrically in terms of two sets of homogeneous parameters a o> a i> a 2> a a anc A>> @2 $8 ky means of the formulas obtained
'-
>
by equating a^ to the corresponding element

Equation (26). The formulas for the parameters ments are determined by (27), for if
of the
of the last
matrix in
product of two displace-
A" = "4" = A' A,
and
A = A'A,
n
and hence
A" = A
for the
A + A'/s, + /sô - /m,

as for the multi/3's
The formulas
plication of quaternions,
's are, by 127, the same and the formulas for the
are given
by
the following quaternion formula:
Now let \
(31)
and X2 be two symbols denned by the multiplication table
and the conditions

If
\q = q\ \z q
lf
g\z> where
is
any quaternion.
we
(32)
write
the
tlie
cr'"s and /9"'s are given in terms of the equations (29) and (30).
or's, /3's,
a''s,
and
/3"s
by
The number system, whose elements are X^-h \(?.,, where q l and q are quaternions, is one of the systems of biquaternions referred to in
the footnote of
as follows
:
130.
It is often given a
form which may be derived
Let
The
rule for multiplying biquaternions,
K (7 + 7^ + 7 y +
2 x
(S
+ B^
+ 7i + T ;/ +
;
gives the following equations
= 7o 7 ~
/
7 3'=
(36)
,'
_i_
,'X
/'^
J- iw'^ 4- S'rv J- X'rv
_ X'v.
-4-
parameters for the
elliptic displacements,
bmce the as and ps are
separate sets of homogeneous variables, they the relation

(37)
may
2 2
be subjected to
2 . 8
+
of (35)
+ ccl + a* = /3 + {3? + /3 + /3
2
By means
(38)
the relation (37) becomes
7<A+
new parameters
yi
i+
yA+%A=displacement (28) in terms
The formulas
of
for the coefficients of a
the
are found
by substituting
in the formulas for a- in terms of the a's
and
r (
/3's.
In other words, the

i
matrix
of the
displacement corresponding to
,y 1 ,'Y,
,'y a
S
',
Q,
^, ^2 , 8 S ) is
?!
%,
^8
7
7.
^v
~7
7
*v
72
-7n
two displacements are (36).
'2
'1
and the formulas
for the composition of
EXERCISE
The
elliptic
geneous variables. the formula

of
displacements are orthogonal transformations in four homoWork out the parameter representation determined by
.a
r>
= (1
,-.
o)
o,
,,
(1
ON
o)
125.
146. Parameter representation of hyperbolic displacements. Let the equation of the absolute be taken in the form
If /A is real, the corresponding is a geometry is elliptic; and if pure imaginary, the corresponding geometry is hyperbolic. No generality is lost by_taking /n = l (as in the section above) for the elliptic
/u,
case and p 1 in the hyperbolic case. For the sake of the limiting process referred to at the end of the section, we shall, however, carry out the discussion for an arbitrary p.
=V
By
precisely the reasoning used in
126
it is
seen that
any
colline-
ation leaving one regulus on the absolute invariant has the matrix
/
(ta z
p.
A=
and any
collineation leaving the other regulus invariant has the matrix
-ft
^1032
ft ft
"ft
ft
^ft-ft
ift
Hence any displacement has
ft
ft
-ft
a matrix
BA. In other words,
if
(40)
As we have already seen in the elliptic case, matrices analogous to A and H,

B'A'
it
A' and
are
.BA^B'B-A'A.
Hence the product
of
two displacements
BA
and B'A'
is
a displace-
ment^"^" such
and
that
A" = A A B" = B'B.

,. 1
J
...
On
is
multiplying out the two matrix products A' A and B'B, it evident that the elements of A" and B" are given by the
formulas (29) and (30) found above for the elliptic case. These formulas are associated with the biquaternions determined by the
table (31).
The remark must now be made that
if
fi
=V
_
.
1,
?
the parameter
representation above does not give real values of a,. for real values of the a's and /3's. Suppose, however, that we transform the biquaternions as follows <4< :
Then
e
t
and
obey the multiplication table
and
we have
or
t
+ ^/ + ajc) 4+ V + 7*7 +
..~
T2
S
.
-f-
where
= (^0+^0), .-CfcA), 7~
2
2
S
5
~
2
The rule
for multiplying biquaternions
according to (41), gives the following equations
7?= 7?=
7^7,
7^7 3 + 7^7,- 7^7i+ 7i7
- 7(73+ 7^7o + 7a'7 + /**$.+ f*'{8J.-i1
For
(U
for /j?=l they reduce to (36).
these equations reduce to (64) and (65) of 130, and For 1 they give the standard
^=
formulas for combining hyperbolic displacements. Thus there are three essentially distinct systems of biquaternions, determined respec-
= 1, /i = 1, /A 0. The first corresponds /* the second to the hyperbolic, and the third to the parabolic geometry. The geometry in each case is determined by an
tively by the conditions
a 2
to
the
elliptic,
absolute whose equation in point coordinates

coordinates,
(44)
is (39),
and in plane
/*X
-I-
<+<+<=
0.
Since the same geometry corresponds to any two real values of there must be a simple isomorphism between any two systems
2
;
p, of
and a like statebiquaternions corresponding to positive values of ju. ment holds with regard to the systems of biquaternions corresponding 2 The biquaternions for which //. to negative values of /u, may
.
be regarded as a limiting case between those for which jj? is positive 2 and those for which /a is negative, just as the parabolic geometry is
regarded as a limiting case between the hyperbolic and elliptic ( 144). In these remarks it is understood that the coefficients 7() , 7l 7a 7g 8 8^ 8 S are always real. From the geometrical discussion above 2 &
,
if
io
o1ar,n
<-T-inf
if
fl-iaca
nAnflfim'on f a
wro-pa
fob-On
aa
nnninlo-r
fclioi
biquaternions for which
^=1
would be isomorphic with those lor
which
2
/Lt
1.
e=l
The multiplication, table (41), in case ^ = 1, is satisfied if we take 1. Hence the biquaternions with real coefficients, and e
=V
are equivalent, in case
/x
1,
to the quaternions
with ordinary
complex
coefficients,
The biquaternions
coefficients,
for
which
\i
= 0, when
taken with
complex
number system of sixteen units with real coefficients. This is the number system ( 130) which is needed to study the displacements in the complex Euclidean
may
be regarded
as a
geometry, and
it
may be
regarded as containing the other
systems
of real biquaternions.
CHAPTER IX
147. Plan
of
the
chapter.
The theorems
and
of
definitions
of
Chapter II are for the most part special cases
more general
concepts of Analysis Situs. The present chapter develops these ideas further, so that the two chapters together lay the foundation
for
the
class
of
the application of
geometry
theorems which are particularly of use in to analysis, and vice versa.
In most of the chapter attention is confined to theorems which can be proved without the use of the continuity assumptions (C, E). Many of the theorems are proved on the basis of A, E, S alone and
others on the basis of A, E, S, P.
first sections ( 148-153) of this chapter we prove some the general theorems about convex regions. These are followed ( 154-157) by the definitions of some very general concepts, such as curve, region, continuous group, etc. It will not be necessary (or
In the
of
eral theory to
possible in the remaining pages) to develop the corresponding genany considerable extent. Nevertheless, these general
notions underlie
may
In
and give unity to the rest of the chapter, which in fact be regarded as a study of certain continuous families of
by special methods. 158-181 the theory
of sense-classes is
figures
developed in consid-
erable detail for the various cases considered in earlier chapters and
for
other cases, the
eleprincipal idea involved being that of an
182-199), we prove the fundamentary transformation. Finally ( mental theorems on the regions determined in a plane by polygons
and
by polyhedra. THEOREM 1. If I is a line coplanar with a triangular region R and containing a point of R, the points of R on I
148.
in space
Convex regions.
a segment.
constitute
line coplanar with a triangle and not containing more than one vertex meets the sides of the triangle in at least two and at most three Doints. These points, by S 22, are the ends of two or
Proof.
three segments. By Theorem 20, Chap. II, the points of any one of these segments are in the same one of the triangular regions deter-
mined by
the triangle, and two points in different segments are
in
different triangular regions.
COROLLAKY. The points common

containing one of
its
to
a tetrahedral region and a line
points constitute a linear segment.
line not on one of the planes of a tetrahedron meets Proof. these planes in at least two and at most four points. The rest of the argument is the same as for the theorem above, replacing
Theorem 20, Chap. II, by Theorem 21 of the same chapter. Convex regions on a line have been denned and studied in
DEFINITION.
23.
set of points in a
plane
is
said to be a
two-dimen-
sional (or planar) convex region if and only if it satisfies the following conditions: (1) Any two points of the set are joined by an
interval consisting entirely of points of the set, (2) every point of the set is interior to a triangular region containing no point not in the set, and (3) there is at least one line coplanar with and not containing
any point
of the
set.
A triangular
region, a Euclidean plane,
and the
interior of a
conic
are examples of planar convex regions.
THEOREM 2. If I is a line coplanar with a two-dimensional conveoo region R and containing a point of R, the points of R on I constitute a linear convex region.
23. Proof. The definition of a linear convex region is given in That the points of R on I satisfy (1) of that definition follows directly from (1) of the definition of a planar convex region. To prove (2) that any point P of R on Z is interior to a segment of points of R on I,
we observe
that by (2) of the definition of a planar convex region interior to a triangular region consisting entirely of points of R and that by Theorem 1 the points common to I and this triangular region are a linear segment. Condition (3) of the definition of a
is
linear
convex region is satisfied by the points of R on I because I contains one point of the line coplanar with R and not containing
of R.
any point
set of points in space is said to be a three-dimensional (or spatial) convex region if and only if it satisfies the following conditions (1) Any two points of the set are joined by an interval
:
DEFINITION.
consisting entirely or points 01 tne set, (z) every point 01 tne set
interior to a tetrahedral region containing
is
no points not in the set, and (3) there is at least one plane containing no point of the set. A tetrahedral region, a Euclidean space, and a hyperbolic space are
examples of three-dimensional convex regions.
THEOREM
3.
If a
line
convex region R, the points of R on
contains a point of a three-dimensional I constitute a linear convex region.
The proof of this theorem follows the same lines as that of Theorem 2, the corollary of Theorem 1 being used instead of Theorem 1 in showing that the points of I in R satisfy Condition (2) of the
definition of linear
convex region.
In consequence of Theorems 2 and 3 the definitions (between, 23 are applicable to colprecede, ray, sense, etc.) and theorems of
linear sets of points in two-
and three-dimensional convex
regions.
In the rest of this chapter the segment AB where A and are in a given convex region R always means the segment AB of points of R.
are three noncollinmr points of a convex THEOREM 4. If a point of R a point of R in the order {BCD}, and region R, in the order {CEA}, there exists a point of R in the orders {AFB}
ABC
and {DEF}.
Proof.
Let
F
(fig.
be denned as the point of intersection

77, p. 351).
of the lines
DE
and
and
AB
By
(3) of the definition of a two- or
three-dimensional convex region there
is a line I* coplanar with A, B, containing no point of R. Hence L does not meet any of the segments AB, JSC, CA. Hence (Theorem 19, Chap. II) the line which meets the segment CA and does not meet B C must meet
C and
DE
AB. Hence {AFB}. The line L does not meet any of the segments FB, BD, DF, and the line A C meets the segment BD and does not meet the segment
BF. Hence
AC meets
the segment
DF. Hence {DEF}.
THEOREM 5. A three-dimensional convex region R satisfies As9. sumptions I-VIII of the set given for a Euclidean space in
Proof.
Theorem
Assumptions I, II, III, V, VIII are direct consequences of and the theorems of 23. Assumptions VI and VII are
consequences of Condition (2) of the definition of a three-dimensional convex region. Assumption IV is a consequence of Theorem 4.
Ul
.i.SbU.JLU.|Jl/iUUtJ
XJLJU
UJL
jj
AH/.
kjj.ij.oo
uuujj.
ULLC -u uÂAu.c;a,ja.
ai_n_l
UIJLO
hyperbolic spaces satisfy these assumptions, this method of convex regions is of considerable interest from the point of
treating view of
foundations of geometry (cf. references in 29). The methods required to prove the theorems on this basis are but little different from those
used in the next
section.
COROLLARY. In a real projective space a convex region also satisfies
Assumption
XVII of
29.
EXERCISES
to a set of convex regions which are all contained in a single convex region is, if existent, a convex region. (In otlxer a of convex regions contained in a convex of set words, the logical product region is a convex region.)
1.
The
set of all points
common
2.
Prove on the basis
of
points are all
P v P2
Pn
finite in
I.
Assumptions I- VIII of number, there is a line
29 that for any set

I
of
JP
such that
P v P z>
. ,
on the same
side of
*3.
set of points in a projective space
such that any two points of
the
set are joined
set
and
by one and only one segment consisting entirely of points of the such that every point of the set is interior to at least one tetrahedral
1
region consisting entirely of points of the set, is a convex region. *4. Study the set of assumptions for projective geometry consisting and the assumption that in the projective space there is a set A,
of of
points satisfying the Assumptions I-VTII,
XVII
for a
convex region.
6.
149. Further theorems on convex regions.
THEOREM
If A, J3,
noncollinear points of a convex region R, they are t7ie vertices of one and only one triangular region consisting entirely of
are
three
points of R.
This triangular region consists of all points on the
segments joining
Proof.
to
the points
of the segment BC.
By Theorem
4 a
line joining
to a point of the
segment
meets a segment joining A to any point A l of the segment j?<7; and by the same theorem any point of the segment A A is joined to B by a line meeting the segment CA. Hence the set of points
CA
[P] on the segments joining A to the points of the segment identical with the set of points of intersection of lines joining to points of the segment BC with lines to of the joining
is
BO
A
points
segment CA.
By
section of lines joining
similar reasoning [P] is the set of points of interto points of the with lines segment
BC
ii.CjU-j.uiN
is
joining
to points
of the
segment AB.
The
points [P] form a
triangular region because they are all the points not separated from a particular by any pair of the three lines AB, BC, CA.
The other three triangular regions having A, B, C as vertices contain points of the line which by (3) of the definition of a convex
is coplanar with and contains no point of R. Hence [P] is the only triangular region satisfying the conditions of the theorem.
region
ABC
In the
rest of this section the triangular region
three noncolliuear points A, B,
of a
determined by convex region R according to
Theorem
shall be called the triangular region
ABC.
It is also
called the interior of the triangle
ABC.
are four noncoplanar points of a convex and only one tetrahedral region consisting entirely of points of R. This tetrahedral region consists of
COROLLARY.
tliey
If
ABCD
region R,
are the vertices of one
the segments
of points of R joining
to
points of the triangular
region
BCD.
Let [a] be the set of segments joining A to points of the BCD and [P] the set of all points on the segments
Proof.
triangular region
to a point of the triis also on a segment joining []. Any angular region ACD, as is seen by applying the theorem above to the figure obtained by taking a section of the tetrahedron ABCD
by the plane
ABP. In
like
manner any
is
on a segment joining
a segment join-
to a point of the triangular region
DAB, and on
ABC.
ing
The same argument shows that any point

line joining
of intersection of a
BCD
is
with a
line
joining
CAD
hi the set
this it
[P] and that every
is
a point of this description.
From
follows that [P] contains all points not separated from a particular by the faces of the tetrahedron ABCD. Hence by Theorem 21,
Chap.
II,
[P]
is
a tetrahedral region.
as vertices and distinct tetrahedral region having from [P] contains points not in R, because it either contains points
Any
ABCD
on the segments complementary to [a] or on the hues joining A to the noinfcs of the trianjmlar regions different from BCD in the
THEOREM
region.
1.
If a plane
TT
convex region R, the points of
contains a point of a three-dimensional R on TT constitute, a planar convex
Proof.
The
points of
R on
TT
satisfy Conditions (1)
and
(3)
of
the
definition
of a planar
convex region because R
satisfies
Con-
ditions (1)
region.
tion
line
and (3) of the definition of a three-dimensional convex To prove that the points of R on TT satisfy (2) of the definiof a planar convex region, let P be a point of R on TT and I a on P and TT. By Theorem 3 there are two points A, A l of R
such that the segment
on
APA
TT
is
composed
entirely of points of R.
I.
Let a be a line on A^ and
but distinct from
By
the same
reasoning as before there are two points B, C of R on a such that the segment BA^C is composed entirely of points of R. By Theorem 6 the triangular region having A, B, C as vertices and containing
contains no points not in R. Hence the points of R on Condition (2) of the definition of a planar convex region.
TT
satisfy
THEOREM 8. If I is any line coplanar with and containing a point of a planar convex region R, the points of R not on I constitute two convex regions such that the segment joining any point of one to any point of the other meets the linear convex region which
I
has in common with

Proof.
R.
By
definition there is a line
coplanar with R and con-
taining no point of R.
of
By Theorem
18, Cor. 1, Chap. II, all points
I or m fall into two classes [0] and [P~\ such two points 0, P of different classes are separated by I and in and (2) two points of the same class are not separated by I and m. The region R contains points of both of these classes. For let I be any point of R on By Theorem 2 any line through J coplanar with R and distinct from I contains a segment of points of R of
the plane not on
that (1)
I.
\vhich
is
one point.
If
and
are
in the order {AIB},
and
are separated
exist
two points of this segment by I and m and also
are points of R.
[0']
all
Hence there
and [P
],
points of Since any
subsets of [0] and [P] respectively, not on I.
two mutually exclusive classes which contain
0'
segment O'P' contains a point

a point of
I.
and any P' are separated by I and m and no of m, every segment O'P' contains
CONVEX KEGKXNS
f
391
Since two points of the same class ([O ] or [P'}} are not separated by I and m, and since the segment joining them does not contain a point of m, it does not contain a point of I.
It
remains to show that any point of either of the
classes, say
p point coplanar Let 0[ and 0! be two points of R on p in the order {0(0' 0'^ and 2 such that the segment 0[0' Let q be z does not contain a point of L
any
line distinct
[(/], is interior to a triangular region consisting entirely of points of this class. Let he any line on a 0' and with R.
from
p,
coplanar
with R and on
be two points
0[, of
and let 0(, 0( R on q in the
order {0' 0' 0' and such that the 8 2 4}
segment 0[0( does not contain a

is
point of /. By Theorem 6 there a unique triangular region with

0(>
Ox>
^4
as
vertices
consisting
only
all
of points of
R and containing
segment 0(02
.
points
/
of
the
Since
does not meet any of the

,
it cansegments 0[0' 2 Z 0[, 0[0[, 0' not meet any segments joining 0( to a point of the segment 0(0+
FIG. 80
(Theorem
4).
Hence the
consists entirely of points of [0']. triangular region 0(0' a O[
COROLLARY 1. If TT is any plane containing a point of a threedimensional convex region R, the points of R not on TT constitute two three-dimensional convex regions such that the segment joining any point of one to any point of the other meets the planar convex
region which
Proof.
TT
has in common with

is
R.
The proof
space,
6.
above
to
using
a strict generalization of that of the theorem the corollary of Theorem 6 instead of
Theorem
COROLLARY 2. For a given line I (or plane TT) and a given convex in region R, there is only one pair of regions of the sort described Theorem 8 (or Cor. 1\
called the
two
sides of the plane relative to the
DEFINITION.
Two
sets of points [P], [Q] in a
convex region. convex region or
in a protective plane or space are said to be separated by a set [$] if every segment of the convex region or of the projective plane or
space which joins a
to a
contains an S.
EXERCISE
convex region but intersecting in a to a point Q in the outside the region. Construct the line joining region by means of linear constructions involving only points and lines in the Cf. Ex. Vol. I. 20, 4, region.
Given two
lines containing points of a
point
150.
Boundary
of
a convex region.
boundary point
containing
of all
of a set of points
contains a point
of [P]
P
is
DEFINITION. is a point [P] if every tetrahedral region and a point not in [P]. The set
boundary points
9.
called the
boundary
of [P].
THEOREM
I
are on are on
Proof.
I.
All boundary points of a set of points on a line All boundary points of a set of points on a plane
a point not on a line
IT
TT.
If
is
I
I,
faces contains
and none
of
whose
faces contains
any tetrahedron one of whose Q will determine
a tetrahedral region ( 26) which contains Q and does not contain any point of 1. Hence Q is not a boundary point of any set of points on I. A like argument proves the second statement in the theorem.
COROLLARY
line
I
1.
A
A
boundary point
B B
of a
I
set
of points [P] on a
is
any point such that any segment of

a point not in
2.
containing
B contains
P and
[P].
COROLLARY
plane
TT is
boundary point
of a
set
of points [P] on a
any point such that any triangular region of TT containing B contains a P and a point not in [P]. THEOREM 10. Let a- be the convex region common to a line I and a planar convex region R and let R and R 2 be the convex regions formed by the points of R which are not on a-. The boundaries of R I and of R contain cr and all boundary voints of <r. Each boundary
;
i5oj
CONVEX REGIONS
of
393
,
and each boundary point of R or of R boundary point of R I or of R Z which is not on I is a boundary

point
is
point of
Proof.
R.
If
is
any point
/,
of a,
and
a line on
Q coplanar with
R
I
and distinct from
points both of
any segment of R : and of R 2 Since any

.
containing
contains
triangular region contain-
ing
Q
of
contains a segment of
.
both of R 1 and of R 2
Hence Q
is
containing Q, it contains points a boundary point both of R l
and
"
is a boundary point of cr, any triangular region contains a point Q of <r, and hence, by the argument Hence JB is a just given, contains points both of R : and of R 2
.
R2
If
containing
boundary point both of R., and of R 2 Let A be a boundary point of R. Any triangular region T containing A contains at least one point not in R 1 or R 2 namely, A
.
itself.
Since
is
a boundary point of
R,,
contains at least one
point of R, which may be in R t or in R.. or in <r. In the latter case T contains points of R and R both, by the paragraph above. I g Hence in every case T contains points of R I or R 2 If every trian.
gular region containing boundary point of both

triangular region
contains points of R i and of R 2 A is a R t and R 2 If this does not happen, some

,
.
R2
ing
(say
containing A contains points of one of R X and of the other. Any triangular region T containthen contains points of R I because by an easy construction
TQ
R X ) and not
both
a triangular region T' containing A and contained in since T' contains A, it contains points of R, which because they are in T must be points of R^ Hence A is a
we
obtain
T and TQ and
;
boundary point of R r Let C be a boundary point

gular region
points of
is
of
R X which
is
not on
I.
Any
trian-
Rr
containing C contains points of R, because it contains It also contains points not in R r One of these points
,
T consists entirely of points of R I( R 2 and /. the latter case should arise, since C is not on I a triangular region T' could be constructed containing C, interior to T, and not
not in R unless
If
a
containing any point of /. T' then would contain points of both R I and R 2 and hence would contain a segment joining a point of R to a. Twirif. nf R TirT-iinVi aanrmanf. Vnr TTionronn R wrmld P.nnhnill n
COROLLAEY. Let cr be the convex region common to a plane TT and a three-dimensional convex region R, and let R l and R 2 be the convex R which are not on TT. The boundaries regions formed ~by the points of
of R and R contain cr and all boundary points of a. Each 'boundary R or of R 2 and each boundary point point of R is a boundary point of X of R or of R 2 which is not on TT is a boundary point of R.
,
It is to be
noted that
we have
Of.
always has a "boundary.
Ex.
7,
not proved that a convex region below.
EXERCISES
are two points of the boundary of a convex region R, one of the segments joining them consists entirely of points of R or entirely of points of the boundary of R.
1. If
A and B
2.
line has
no points, one point, two points, or one interval in

set, its
common
with the boundary of a convex region. a Tf a segment consists of boundary points of a given
T
ends are also

or its
points.
ng the notation
,-e
of
Theorem
10,
no point
of
not in
cr
bound-
a boundary point of R. Hence if P is a point of a two- or threeand dimensional convex region R, and B a boundary point of R, the points
B are
IT
joined by a segment consisting entirely of points of R. 5. Using the notation of the corollary of Theorem 10, no point of
TT
not in
boundary can be a boundary point of R. 6. Using the notation of Theorem 10, if R and its boundary are contained in another convex region R', then no point of the boundary of R x not on or or its boundary can be on the boundary of R 2 7. Give an example of a space containing a convex region which has no
or its
.
boundary.
8. ray whose origin is in the interior of a triangle meets the boundary of this triangular region in one and only one point. be an arbitrary point of a Euclidean plane, and R an arbitrary *9. Let
convex region containing and having a boundary which is met in two Let any set of points into points by every line which contains a point of R which the boundary of R can be transformed by a homothetic transformation ( 47) be called a circle. Let the point to which is transformed by the
.
homothetic transformation which carries the boundary of R into any circle be called the center of this circle. Let two point-pairs and A'B' be said
AB
to be congruent
through and passing through B'. The geometry based on these definitions is analogous to the Euclidean plane geometry. Develop its main theorems. Cf. the' memoir of H. Minkowski by D. Hilbert, Mathematische Annalen, Vol. LXVUI
).
if and only if there is a circle with A as center and passing which can be carried by a translation into one with A' as center
445.
151. Triangular regions.
The theorems
of the last sections can
be used to complete the discussion of the regions determined by a

triangle.
We
shall continue to use the notation of
26 and
shall
denote the sides

of
is
the
the
AB, BC, CA by c, a, and b respectively. The points plane which are not on a form a convex region, of which a boundary. By Theorem 8 the points not on a or b fall into
of
two convex regions,

and
each of which a and
b together (by
Theorem
10)
constitute the boundary.
The
line c
meets a and
b in the points
respectively and hence has the segment
in
common
with
one of the regions and 7 in common with the other. By Theorem 8 the region containing 7 is separated into two convex regions, each having 7 on its boundary, and the other into two, each having 7 on
its
boundary. Thus the three lines a, b, c determine four planar convex regions which are identical with the four triangular regions
Theorem 20, Chap. II. Since the lines enter symmetrically, each the segments a, fi, 7, "a, /?, 7 is on the boundary of two and only two of the triangular regions.
of of
The three
vertices A, B,
are on the boundaries of
all
four
tri-
angular regions, because every point of the plane can be joined to these three points by segments not meeting the lines a, I, c. No point
not on
regions, because
can be a boundary point of any of the triangular such a point is an interior point of one. of them. Since any line which meets one of the four planar convex regions
a, b, or c
meets
in a segment the ends of which are the only points of on the boundary, the three segments which bound one of the four
it
triangular regions cannot be
met by the same

with
line.
The boundaries
the
of the four regions therefore consist respectively (cf. fig. 16) of
vertices of the triangle, together

a, /3,
a, /3,
7
7 7 7
for for for for
Eegion
Eegion
I,
II,
a, a,
/3, /3,
Eegion III, Eegion IV.

2,
In addition to what has already been stated in Theorem

discussion above gives us the following information
:
the
THEOKEM 11. triangular region is bounded by the three vertices of the triangle, together with three segments joining them which cannot
If AOB and APC are two noncollinear segments they may be and G denoted by a and ft. The two segments whose ends are may be denoted by 7 and 7, 7 being the one met by the line OP. As we have just seen, a, ft, and 7, together with the vertices of the
triangle, are the
boundary
of
a convex region, and there
is
one
and
/3
of whose boundary only one of the four convex regions
a and
form part. Hence THEOREM 12. For any two noncollinear segments a, ft having a common end there is a unique triangular region and a unique seg-
ment 7 such that

the
a,
ft,
and
7, together
with the ends of a
and
ft,
form
boundary of the triangular region. COROLLARY 1. On any point coplanar with but not in a given
is
triangular region T, there points not in T.
at least one line composed entirely
of
COROLLARY 2. The triangular region determined according to Theorem 12 by two noncollinear segments CB'A and CA'B consists
of the points of intersection of the lines joining B to the points of the segment with the lines joining A to the points of the second segment.
first
The complete
set of relations
among the
points, segments,
}
and
tri-
angular regions determined by three noncollinear parts A, be indicated by the following tables,
B C may
EL:
where in the
first
table a "I" or a
"0"
is
placed in the *th
row and
yth column according as the point whose name appears at the beginning of the ith row is or is not an end of the segment whose
name
as the
is
table a " 1" or a
appears at the top of the /th column and where in the second "0" is placed in the ith row and/th column according
;
segment whose name appears at the beginning of the *th row or is not a part of the boundary of the triangular region whose
EXERCISES
1.
The
lines polar
18, Vol. I)
points of one of the four triangular regions one of the four sets of lines determined
with respect to a triangle determined by
ABC
to the
ABC
constitute
by ABC, according
to the dual of
Chap. 1 1. The points on these lines constitute the set of all points coplanar with but not on the given triangular region or its boundary.
'20,
Theorem
2.
Divide,
the lines of the plane of a complete quadrangle into classes
according as the point pairs in which, they meet the pairs of opposite sides separate one another or not. Apply the results to the problem: When, can a
real conic
be drawn through four given points and tangent to a given line
Dualize.
152.
The tetrahedron. The discussion

Let us use the notation
of
in
to space.
26.
The
151 generalizes at once points not on <xl con-
convex region of which a l is the boundary. By Theorem 8, the points not on # t and cc2 constitute two convex regions, of each of which, by Theorem 10, a and a form the boundary. z 1
stitute a
Cor. 1,
regions
The plane ag has points in each of the three-dimensional convex bounded by a l and a2 and hence by Theorem 7 has a planar
convex region in common with each of them. By Theorem 8, Cor. 1, each of these planar convex regions separates the spatial convex region in which it lies into two spatial convex regions, of each of which
it forms part of the boundary. Thus the points al} az a3 form four spatial convex regions. Since any plane not on A 4 meets az and a s in a triangle, it meets each of these a^,
(Theorem 10, Cor.)

,
not on
four spatial
planes
a l}
convex regions in a triangular region. Thus, since the az a s enter symmetrically, we have
,
THEOREM
13.
DEFINITION.
in three lines,
regions.
and each pair of

a^,
Three planes a^ a z a^ meet by pairs these lines bounds two planar convex
,
The points not on
regions (called trihedral regions) each

three
az and as form four spatial convex bounded by the three lines and
,
of the planar convex regions.
are fully represented by the matrices
denoted by A, B, C, the
the
among these regions 151 if the three lines are planar convex regions by a, /?, 7, a, ft, 7, and
relations
The
of
three-dimensional regions by
of the four spatial
I, II, III,
IV.
,
convex regions determined by a^ #2 ag is met by ai in a triangular region and separated by it into two convex regions each of which is partially bounded by the triangular
Each
of
Theorem 21, Chap. II. Since the planes a^ a ? 8 a.t enter symeach of which is on metrically, there are sixteen triangular regions
,
,
the boundary of two and only two three-dimensional regions; and, moreover, each tetrahedral region has one and only one triangular
on its boundary. region from each of the four planes can be joined to any of the Since any point not on a lt a z 3 cc^ a segment not containing any point of points A I} A 2 A s A^ by a a , a.A or a 4 the points A lt A z A S A^ are on the boundary of all
, , ,
each segment eight tetrahedral regions; and by similar reasoning which bounds a triangular region also bounds each of the tetrahedral
regions bounded by the triangular region.
THEOREM
four
14.
vertices, together
The boundary of a tetrahedral region consists of its with four triangular regions and the six seg-
ments bounding
vertices.
the
four triangular regions and bounded by the four
COROLLARY. Three noncoplanar segments having a common end are on the boundary of one and only one of the tetrahedral regions having their ends as vertices.
regions,
The complete set of relations among the points, segments, triangular and tetrahedral regions determined by A A 2> A.Jt _4 4 may be
I}
indicated
151. by three matrices analogous to those employed in That the points A and Aj are ends of the segments a- arid cr y is indicated in the first matrix, a "1" in the 1'th row and yth column
{
{j
signifying that the point whose name appears at the beginning of the 2th row is an end of the segment whose name appears at the
" top of the yth column, and a
"
signifying that
it is
not.
The four triangular regions in the plane a .(i' = l, 2, 3, 4) determined by the lines in which the other three planes meet at may be denoted by riv Ti2 r i8 rf4 Applying the results of
t , ,
.
is
or
is
is
nut
at
on the boundary
of
the triangular
region
whose
name
the top of the yth column.
H,,:
struct a
"
T8 and conLet us denote the eight tetrahedral regions by T 1? matrix analogous to the preceding ones, in which a "1" or
,
a appears in the i'tli row and jtli column according as the triangular region whose name is at the beginning of the -ith row is or is not on the boundary of the tetrahedral region whose name is at the top of the /th column. By definition there is a plane TT which
all the six segments a-., and none of the segments &~. There one and only one tetrahedral region not met by TT. Let us assign the notation so that this region is called T^ As TT cannot meet the
"
meets
is
segments and triangular regions on the boundaries of T x these segments must be the six segments cr~ and these triangular regions must,
,
be those bounded by ~a The latter can be found by means of the ij matrix This determines the first column of the matrix to be 2
.
constructed.
The other columns
are found
by considering succes-
sively the planes of the
seven other classes of planes described in
26.
Thus, for example,
is
2
the region on whose boundary are the

o-
segments
%, %, <T M
ff
23 ,
<T
M,
EXERCISE
A to planes polar ( constithe points of one of the tetrahedral regions determined by tute one of the four sets of planes determined by 26. according to The points on these planes constitute the set of all points not on the given
The
18, Vol. I) with respect to a tetrahedron
BCD ABCD
ABCD
tetrahedral region
or its
boundary.
* 153. Generalization to n dimensions.
The
generalization to
the point pair, triangle, and tetrahedron is the 1 points no n of This is any set of n (n -f- T)-point in n-space. which are in the same (n l)-space, together with the lines,
n dimensions
of
planes, 3-spaces, etc. which they determine by pairs, triads, tetrads, etc. By a direct generalization of 26 one proves that the points
not on the
1 spaces of
an (n
+ IVpoint
fall
into 2 71 mutually
153, 154]
CURVES
,
401
that any two points of the same set R^ are joined by a segment of points of the set and that any segment joining two points not in the same set contains at least one point on an (n
*
exclusive sets
R^ such
3)-space
i
of
the
(n -f l)-point.
Any
one
of
the
sets
RI>
'>
R-j"
willed a simplex or n-dimensional segment.
Thus the simplex is a generalization of the linear segment, trianand tetrahedral region. By replacing triangular and tetrahedral regions by simplexes throughout 148-152 we obtain
gular region,
immediately the theory of w-diruensional convex regions. A like process applied to 154-157, below, gives the theory of w-dimensional
connected
tinuous
sets,
regions, continuous families of sets of points, conof
families
transformations, continuous
groups,
etc,
"We
leave both series of generalizations to the reader.
154. Curves. DEFINITION. Let [T] be the set of all points on an interval set of points [P] is called a conT^T^ of a line 1. tinuous curve or, more simply, a curve, if it is in such a correspondence T with [T] that
one and only one P such that F(T); one T such that P = T(T); R con(3) for every T, say T', and for every tetrahedral region taining F(T'), there is a segment <r of I containing T' and such
(1)
for every
T there
is
P=
(2)
for every
there
is at least
that for every
Tin
<r,
F(2 )
is
in R.
if
curve
if
is
said to be closed
F(ro = F(r
)
i ).
It is said to be
simple
can be chosen so as
to satisfy (1), (2), (3)
and so that
if
T'^T", T(T')3= T(T") unless

pair
the pair
T'T"
is
identical with the
T Tf
is
The point T(T)

curve
is
said to describe the curve as
varies.
The
said to join the points
F(r and F(I

o)
1 ).
In view of the definition of the geometric number system in Chap. VI, Vol. I, and the theorems in Chap. I, Vol. II, this definition could also be stated in the following form Let be the set of numbers such that OSiSl. A set (/:) that of points [P] is called a curve if it is in such a correspondence F with [f]
:
(3) for every t there is least one t such that
one and only one
P= F
is
(C),
t,
P = F (V),
is
and
(3) for every
there is at (2) for every say t', and for every tetra-
hedral region
R containing
B,
F (Y)
there
a number S
>
such that
if
C-8<* <t'+
F(Y)
in R.
The
most obvious examples
of
simple closed
curves
are
the
conic. The proof that these are simple projective line and the point closed curves will be given for the planar case, and may be extended
at once to the
three-dimensional case by substituting tetrahedral
regions for triangular ones.
THEOREM
Proof.
lt
15.
closed curve. projective line is a simple
P P P P
,
of points on a projective line and let be four particular values of [P] in the order {P^ 2 P;} Let T , TI} T^, Ts T4 be five collinear points in the order r and let [T] be the set of all points of the interval l\l\T^. If is let T(T) be the point to which is on the interval
Let [P] be the set

s
2,
{T^T^T^
T
T^T^,
carried
by a
projective
,
T,
val
T T^ into P P v P T T T^, let T(T) be

I}
correspondence* which takes the points respectively; and if T is on the inter-
projectivity
respectively.
which
carries
the point to which the points T2 TS

,
T
,
is
carried
z,
T^ into
P P P Q
8
,
by a
The correspondence F is denned so that there is one and only rf one point P=T(T) for each T; and also so that T(T')^T(T ) t
unless
T'=T", or T' = T^ and T"= T4 or T'= T^ and T"= Tf Thus [P] satisfies conditions (1) and (2) of the definition of a curve and the condition that a curve be simple. Let R be any triangular region containing a point P'=T(T ).
,
By Theorem
be the ends
1 there is a
segment
R; let
of the projective line
[P] con-
taining P' and contained in

of this
P"=T(T") and P'"=T(T'")
This segment is the image either of f" or of the points T not between the points T between T" and T T" and T'". Hence if or be any segment of the line T^T^ con1 taining T and not containing T" or T'", every point T on a- is such
segment.
that
T(T)
is
in R.
Hence [P]
satisfies
Condition
(3) of
the defini-
tion of a curve.
THEOREM
Proof.
16.
A
is
point conic
is
is
a simple closed curve.
The
proof
precisely the
same
as that of
except that [P]
lemma
is
the set of points on a conic, used instead of Theorem 1.

it
Theorem 15 and the following
*This does not use Assumption P, because
requires only the existence of a
Proof.
obvious.
If
C2
is
entirely in
R the conclusion
of the
theorem
is
If not, let
be a point of
Cz
not in
R.
By
75 the points
of the line
PQ
interior to
as ends.
of this
Let JK be a point
C 2 constitute T
a segment having
and Q
also
segment which is on the segment con-
taining
(Theorem
1),
which the line
PQ has in
(fig.
common with R
Let
81).
be the
common
1
point of the tangents at
and Q and let T and T" be points of R in the
order
{TT'PT"}.
Let S'
and S" be the points in
which QT' and
QT" meet
{TS'KS"}be points
interior lines to
TR]
Let
so that
Sl and Sa
to
1).
]/
C z and
,
FlG
in the
81
interior
R,
order
in
(Theorem
The
QS^ and
QS2 meet TP
two points T^ and
11 these points are T^ respectively in the order {TT'T^PT^T }. Since on the segment T'PT" they are in R. Since Q is on the conic C*
the lines
QT^ and
QT
meet C 2 in two points
J^
and
respectively.
2
,
and T^ (a point of a tangent) exterior to <7 we have the order {QS&T^. But 8^ and Tx are in R and Q is not in R. is in R. Hence by Theorem 1, P^ is in R. In like manner P 2
Since
is
S1
interior
The segment
of the conic
C7
is
now
easily seen to consist
entirely of points of R.
For
if
T and S
the points in which

2
QP
any point of this segment, and meets PT and ET respectively,

is
PP%P A PTT& | BSS&.

T is on the segment T^P'T^ and S is on the segment SjRSf Hence J and S are interior to R, and S interior to C*. Since T is 2 exterior to C it follows that /S and T separate P and Q. Therefore,
Hence
7
as
is
not in R,
is
in R.
404
THEOREMS ON SENSE AND SEPAEATION

EXERCISE
[CHAP. IX
The boundary
155.
of a triangular region is a simple closed curve.
Connected
if
sets, regions, etc.

if
set of points is said to of the set are joined
be
a
is
connected
and only
any two points
by
curve consisting entirely of points of the set. sometimes called a continuous family of points.
connected set
In a space satis-
or A, E, J) a connected fying Assumptions A, E, H, C (or A, E, connected set in a plane such that set is also called a continuum.
every point of the set is in a triangular region containing no points connected set of points not in the set is called a planar region.
in space
such that every point of the set is in a tetrahedral region containing no points not in the set is called a three-dimensional
region.
one-to-one transformation
carrying a set of points [X~\ into a

if and only if for every X> containing F(X'), there is a and such that for every in R,
set of points
[Y]
is
said to be continuous
say X', and every tetrahedral region

tetrahedral region T(X) is in T.
R containing X'
If a linear interval joining two points A, is subjected to a continuous one-to-one reciprocal transformation, it goes into a curve joining the transforms of A and B ( 154). The set of points on the curve, excluding the transforms of A and B, is called a 1-cell.
If a triangular region and its boundary are subjected to a continuous one-to-one reciprocal transformation, the set of points into which the triangular region goes is called a simply connected element
of surface, or a 2- cell.
If a tetrahedral region and its boundary are subjected to a continuous one-to-one reciprocal transformation, the set of points into which the boundary goes is called a simply connected surface, or simple
surface,
is
and the set of points into which the tetrahedral region goes called a simply connected three-dimensional region, or a 3-cell.
EXERCISES
1.
A region
contains no point of
its
boundary.
3. Given any set of regions all contained in a convex region. The set of points in triangular regions whose vertices are in the given regions is a convex region. This region is contained in every convex region containing the
all
given set of regions (J.

4.
W.
Alexander).
on segments joining pairs of points of an arbitrary region R contained in a convex region constitutes a convex region R'. The
set of all points
The
region R'
5.
is
contained in every convex region containing

(
R.
The boundary
the set of all points in the region not in the region.

6.
150) of a region in a plane (space) separates (149) from the set of all points of the plane (space)
continuous one-to-one reciprocal transformation of space transforms
any region into a region.
156. Continuous families of sets of points.
The
notion of con-
tinuous curve has the following direct generalization:
DEFINITION.
}f
Let [T] be the set
of all points
on an interval
TQ T^
a line L
set of sets of coplanar points [S] is called a con-
tinuous one-parameter
family of
sets
of points
if
it
is
in such a
3orrespondence
(1)
with
T that
is
for every
T there
(2)
for every set
S there
one and only one set S such that S = T(T) is at least one T such that S = F(T);
(3)
for every T, say T',
and
for every triangular region

is
includ-
ing a point of
the set T(T'), there

if
a segment
or
of
containing T'
md
such that
T is
in
o-
at least one point of the set
F(T)
is
in R.
The
ihe
definition of a continuous one-parameter family of
sets of
by replacing the -triangular region R in statements above by a tetrahedral region. If the sets S are taken to be lines, planes, conies, quadrics, etc., :his gives the definition of one-parameter continuous .families of
points in space is obtained
lines,
planes, conies, quadrics, etc., respectively.
Of. Exs. 1~5, below.
DEFINITION.
connected set of sets of
pamts
or a continuous
family of sets of points is a set of sets of points [S] such that any }wo sets B I} S 2 are members of a continuous one-parameter family
3f
sets of [S].
For example, the discussions given below in terms

nected set of sets of points.
Of. also
of
transformations establish in each case that a sense-class
elementary is a con-
Exs. 6~7, below.

extended in an obvious way
The
so as
definition of a continuous family
may be
to include sets
whose elements are
of points, sets of points, sets of sets
EXERCISES
Defining an envelope of lines as the plane dual of a curve, prove that an envelope is a continuous one-parameter family of lines. 2. The space dual of a curve is a continuous one-parameter family of planes.
1.
3.
Pencils of lines
and planes
are continuous one-parameter families.
4.
5.
A line conic or a regulus is a continuous one-parameter family of lines. A pencil of point conies is a continuous one-parameter family of curves.
The The
set of all lines in a plane or space or in a linear
is
6.
congruence or a
linear
7.
complex
a connected set of sets of points. set of all planes in space or of all planes tangent to a quadric ia
connected
set of sets of points.
157. Continuous families of transformations.

of all points
on an interval
TT
of a line
I.
transformations of a set of points [P]. If (1) responds one and only one transformation Ii T> and (2) for every the set of points [n r (JP)] is a curve for which the defining point
Let [T] be the set Let [II r] be a sefc of to every T there cor-
154) may be taken to be correspondence F (in the notation of 7 the correspondence between I and II r (P), then [II r] is said to be a continuous one-parameter family of transformations. The curves
[IIy(P)] are called the path curves of [II r].
The term
"
continuous one-parameter family of transformations

of sets of points
I).
"
may
also be applied to a set of transformations [TL T] of a sefc jS of
points
and
(e.g.
S may
and
defined in
13, Vol.
In
this case (1)

:
be a set of figures as (2) must be satisfied,
following condition S, (3) For every set of points one-parameter continuous family of sets of points for the F which defining correspondence (in the notation of 156) may
and
also the
[II r (S)] is a
be taken to be the correspondence between and r (S). If the set of correspondences [II r] is both a group and a continuous one-parameter family of transformations, it is called a one-parameter
continuous group. set of transformations [II] of a set of points and of sets of points, such that any two transformations of [II] are members of a con-
tinuous one-parameter family of transformations of [II], is called a continuous family of transformations. If [II] is also a group, it is called a continuous group.
If [Ily] is a continuous one-parameter family of one-to-one recip-
rocal transformations of a figure F,
and
if
II
TQ
is
the identity, then J?
of intermediate positions
set of points and all the transformations of the family [TI T] are continuous, the deformation is said to be a continuous deformation.
is is
UT
[H T (F}]. called a deformation; if
Any
a
one of the transformations
158. Affine
theorems on sense.
about
main
propositions
sense-classes
Let us recapitulate some of the in Euclidean spaces by
enumerating the one-dimensional propositions of which they are

generalizations.
The group
of
all
projectivities
x'=ax-\-'b on a Euclidean line
has a subgroup of direct projectivities for which a 0. This subgroup is self-conjugate, because if a transformation of the group be
>
denoted by 2, and any other transformation x
= ax + ft
by T, then
a(i,-U aj
\a:
transformation
in
which the
coefficient of
is
positive.
From
18
0,
the fact that the subgroup is self-con jugate, it follows as in that the same subgroup is defined by the condition a
>
no
matter
infinity.
how
the scale
is
chosen, so long
as
P&
is
the point at
These statements are generalized to the plane in

31.
30 and
any dimensionality in in replacing a by the determinant

to spaces of
The
generalization consists
for
the two-dimensional case,
and by the corresponding %-rowed

case.
all
determinant in the w-dimensional
A
into
sense-class
B(AB)
is
the set of
ordered pairs of points
which a pair
(
jectivities
23).
can be carried by direct proThis proposition is generalized to the plane in

of distinct points
30 and to w-space in
31.
particular arbitrarily chosen sense-class shall be called positive
and the other sense-class shall be called negative. This statement reads the same for any number of dimensions. In the three-dimensional case the positive sense-class is also called right-handed
and
the negative
i-
162). left-handed (see the fine print in In the one-dimensional case a nonhomogeneous coordinate sys-
sense-class
n -~-.
,'
11,1
'
'.*
'(!
C/r>r>\
~;4
T-~
V-T-,/1
4-iTiY-.
î-iv-irmcjirvnol
case
nonhomogeneous coordinate
is
system
is
called
positive
(0, 1).
if
S(OXY)
the
is
positive
when
= (0,
0),
X=(l,
if
0),
and
Y=
In
three-dimensional case a nouliomogeneous

called positive
coordinate
is
system
or
right-handed
S(OXYZ)
and
positive
0, 1).
when
.4'jB'
(0, 0, 0),
lliJ
X=
(1, 0, 0),
Y= (0,
if
1, 0),
Z= (0,
On
Euclidean line two ordered pairs of points
AB
and
are in the
same sense-class
1 1
and only
if
a
I
and
have the same

of A,
tive,
-B, A',
sign, a,
b,
B' respectively.
is
being the uonhomogeneous coordinates Hence, if the coordinate system, is posia) is
S(AB)
positive or negative according as (b

criteria
for
positive or
negative.
30, 31.
Similar
It
is
the plane
and space are given in

if
follows
immediately that
the coordinate
system
in the plane
positive,
S(ABC)
-,
is
positive or negative
according
as the determinant
is
positive
or
negative, where
(a lt
2 ),
(b^
&
2),
O
is
(c^,
2 ).
the coordinate system in space is positive, or negative according as the determinant

If
S(ABCD)
positive
a2 a a
is
positive or negative,
c
3 ),
where
A(a
B
is
1,
az
a 8 ),
JS
= (b
b
lf
g ),
C=
(c v c 2 ,
D=(d^
d2 d s ).
,
In the one-dimensional case
on one or the other of the rays
having A as origin according as S(AB) is positive or negative. In a Euclidean plane C is on one side of the line AB or the other
according as
space
S (ABC)
is
positive or negative
(30). In
Euclidean
is
on one side or the other of the plane

is
ABC
according as
S (ABCD)
The
positive or negative.
/
projectivities
= ax
-V-
b of
the Euclidean line 'are in one-to(a, b) of
one reciprocal correspondence with the points

plane.
the Euclidean
The
direct projectivities
correspond to the points on one

tf\ao
r\-
f.Ti <
other side.
projectivities
tivities is
From
this it readily follows that the set of all direct

set of all projec-
forms a continuous group, whereas the a group which is not continuous.
In like
manner the transformations
2l
22 y
a 20
can be set in correspondence with the points of a six-dimensional Euclidean space, the direct and opposite collineations respectively corresponding to
points of
two regions separated by the locus

1122
12
a 21
'
Similarly, the direct
dimensions
dimensions.
may
In
collineations in a Euclidean space of three be represented by points of two regions in a space of twelve
and opposite
all
three cases the set of
all direct collineations
forms a con-
tinuous group, but the set of all collineations does not. Another way of coming at the same result is this Let the ordered pairs of points of a Euclidean line be represented by the points (a, b) of a Euclid:
AB
ean plane, a being the nonhomogeneous coordinate of A, and b that of B. Under this convention the points representing pairs of the positive sense-class are
on one side of the line

sense-class
tivities
and those representing
pairs of the negative

affine projec-
on the other
side of this line.
The one-dimensional
pairs to
tivities
are in one-to-one reciprocal correspondence with the ordered point which they carry a fixed ordered point pair PQ. The direct projec-
the line b
thus correspond to point pairs represented by points on one side of a and the opposite projectivities to point pairs represented
side.
by points on the other

159.
Elementary transformations on a Euclidean line. DEFINITION. Given an ordered pair of points AB of a Euclidean line, the operation of replacing one of the points
by a second point not separated
from
it
by the other point
is
called an elementary transformation
of the pair
Thus
that
AB* AB may be
(cf.
transformed into AB'
other words
23)
In if {ABB } or {AB'B}. can be transformed to any point B' such

into no other.
S(AB}=S(AB'}, and
is
Hence
it
follows that
if
AB
transformable to A'B' by any sequence of elementary trans-
S(AB} S(A'B'}. Conversely, if S(AB) S(A'B } } it is easy to see, as follows, that a can be transformed by sequence of elementary transformations
formations,
AB
The transformations which we have considered heretofore have usually been
to A'B',
From
that there are two points
the theorems on linear order in Chap. II it follows A" and B" satisfying the order relations
{ABA"B"}
and
{A'B'A"B"}.
goes to
By
elementary transformations
AB
AB"; AB" to
A""; A"B"
to A'B"; and
A'B"
to
A'B
1
.
Hence we have
THEOREM 17. On a JEuclidean line the set of all ordered pairs of points into which an ordered pair of distinct points A B can be
transformed by elementary transformations
is
the sense-class
S(AB}.
An
to
elementary transformation
157).
If
may be
regarded as a special type of con-
tinuous deformation (
AB', the point
B may be
AB is carried by an elementary transformation thought of as moved ( 157) along the segment
to B', and since this segment does not contain A, the motion is BE' from such that the pair of distinct points never degenerates into a coincident pair. Thus we may say that a sense-class consists of all pairs obtainable from a
by deformations in which no pair ever degenerates. the ordered point pairs are represented by points in a Euclidean. plane, as explained at the end of the last section, an elementary transformation corresponds to moving a point (a, b) parallel to the a-axis or the &-axis
fixed pair
When
in such a
way
as not to intersect the line a
b.
DEFINITION.
AB
is
elementary transformation of a pair of points said to be restricted with respect to a set of points if and
An
[P]
only if it carries one of the pair, say B, into a point B' such that the segment BB' does not contain any one of the points P. (Any
one of these points may, however, be an end of the segment BB'.} It is evident that any elementary transformation can be effected
as a resultant of a
sequence
of
elementary transformations which
are restricted with respect to an arbitrary finite set of points.
Hence
Theorem 17 has the following

COROLLARY.
I.
corollary:
Let be any finite set of points on a a} >, n v ordered pairs of points are in the same sense-class if and only if one canle carried into the other ly a sequence of elementary
line
P P
Two
transformations restricted with respect
to
P P
I}
z,
Pn
concept of a restricted elementary transformation is intimately connected with the idea of a " small motion." In the metric geometry the points
The
Pif P2
Pn
can be chosen so as to be in the order
{P v
Pn
and so
that the segments -FVPi+i are arbitrarily small. Any elementary transformation of a pair of points on the interval Pf/'f+i will be effected by a small
DEFINITION.
a
Elementary transformations in the Euclidean plane and space. Given an ordered set of three noncollinear points in Euclidean plane, an elementary transformation is the operation
160.
of replacing
one
of
them by
a point
which
is
joined to
it
by a
segment not meeting the line on the other two.

As in tho one-dimensional case, an elementary transformation, may be regarded as effected by a continuous deformation of a point triad. path is specified along which a point may be moved without allowing the triad to
degenerate into a collinear one.
Let A, B,
be three noncollinear points and
let C'
and B' be
points of the segments
AB
and
84,
CA
p.
respectively.
tary transformations
this to C'BJi';
(cf.
fig.
423)
ABC
Then by elemengoes to C'BC; and
and
this to
it
BCA. In
to
like
manner
C'CB'] and this to BCB'\ and this to can be shown that ABC can be carried
Hence any even permutation of three noncollinear points can be effected by

by a sequence of elementary transformations.
elementary transformations.
CAB
an elementary transformation leaves the 30, Hence, by Theorem 26, no odd permutation can be effected by elementary transformations.
27,
30,
By Theorem
sense of an ordered triad invariant.
If
A', B',
C'
are
any three noncollinear

of
points,
ABC
can be
carried into
tions.
A'B'C' by elementary transformaFor since at most one side of the triangle A'B'C' is parallel
some permutation
this line
to the line
AB,
meets two
of the sides in points
which we
denote by A" and B". By one-dimensional elementary transformations on the line AB, the ordered pair AB can be carried
may
These one-dimensional elementary either to A"B" or to B"A". transformations determine a sequence of two-dimensional elemenABC to tary transformations leaving C invariant and carrying
or to B"A"C. The point C can be carried by an obvious not elementary transformation to a point C" such that A"C" is be carried to parallel to any side of A'B'C', and then A"C" can
A"B"C
two
sides
of of
r the points, say A'"C ", in which the line A"C" meets the the triangle A'B'C'. The points A'"B"C'" are on the
sides of the triangle A'B'C', and the one-dimensional elementary transformations on the sides which carry them into the vertices determine two-dimensional elementary transformations which carry
412
Since
[CHAP. IX
f
ABC
all
cannot be carried into A'B'C

of
if
S(ABC)^ S(A
BC
f
))
and since
even permutations
tary transformations, it by a sequence of elementary transformations
A'B'C' can be effected by elemenfollows that ABC can be carried into A'B'C'
if
S(ABC)==S (A'B'C').
Hence we have
THEOREM
18.
In a Euclidean plane S (ABC) = S'(A'B'C') if and
only if there exists a finite set of elementary transformations carrying the noncollinear points A, B, C into the points A', B', C' respectively.
DEFINITION. Given an ordered set of four noncoplanar points, an elementary transformation is the operation of replacing one of them by another point which is joined to it by a segment containing no point of the plane on the other three.
Let be four noncoplanar points. Holding fixed, may be subjected to precisely the sequence of elementary transformations into BCA. This given above in the planar case for carrying
ABCD
ABC
ABC
effects the
permutation
(A
B C D\
(BOA vr
the symbol for each point being written above that for the point into which it is transformed. In like manner we obtain the permutations
(ABC D\ (ABC D\ /ABCD} D C B D C D

\B
and
it is
A)
\C
A]
\A
B)'
easily verifiable that
product of
any even permutation of ABCD is a these permutations. Hence any even permutation of a
set of four points
may
be effected by elementary transformations.
By Theorem
(
23,
b
2>
a i>
2>
a s>>
fii>
27, an elementary transformation of four points &)> (<v c 2 , c a ), (dlt d2 , da ) leaves the sign of
&
invariant, and hence leaves their sense-class invariant. Hence ( 31) no odd permutations of four noncoplanar points can be effected by
e
lementary transformations.
An
Ordered
tfifvrarl
ABUT)
nf nnnr>.rmlnnfir rwvints nan
ho
no^riori
160,161]
ELEMENTARY TRANSFORMATIONS
two points A"B"of the planes
this
of
413
carried into
the tetrahedron A'B'C'D'.
By repeating
case it
be carried to points
it is easily can also proved that C and on these planes. By the two-dimensional follows that the ordered tetT&&A"I>"C"D" of points on the
argument
C"D"
A'B'C'D' can be carried into some permuABCD cannot be carried into A'B'C'D', if S(ABCD)^ S (A'B'C'D'} it follows by the last paragraph but one that it can be carried into A'B'C'D' if S (All CD) ^S (A'B'C'D }.
planes of the tetrahedron
tation of its vertices.
Since
Thus we have
THEOREM
the
19.
In a Euclidean space
S(ABCD}S (A'B'C'D
if and
only if there exists a finite set of elementary transformations carrying
noncoplanar points A, B, C,Dinto
the points A',
1
,
C',D' respectively.
re-
The theorems and definitions
of the last
two sections can be
garded as based on any one of the sets of assumptions A, E, H, C, or A, E, or A, E, P, S. Assumption P is used wherever coordinates
the argument without the aid on A, E, S alone (cf. Ex. 2, 161). 161. Sense in a convex region. DEFINITION. Given a set of three noncollinear points of a planar convex region R, the operation of R on the same replacing any one of them by any other point of
are
employed, but
it is
possible to
make
of coordinates
and thus to base
it
side
of
the
line
joining
the other two
is
called
an elementary
transformation.
The
set of all ordered triads obtainable
by
finite
sequences
of
elementary
transformations
is
from one noncollinear and

is
ordered triad of points
ABC
denoted
by S(ABC}.
This definition
is
in
agreement with the propositions about sense
given for the special case of a Euclidean plane. Moreover, if R is any convex region, and is any line coplanar with R but containing no
point of R,
respect to
two
if
triads of points of
if
R are in the same sense with
they are in the same sense with respect to the Euclidean plane containing R and having L as singular line. Hence the theorems of 160 may be taken over at once to convex
and only
regions in general.
This result
may
be stated as follows
there are
THEOREM
two senses.
20.
In a planar convex region Sense is preserved l>y even and
two and only
altered
l)y
odd pertnu-
DEFINITION. Given a set of four uoncoplanar points of a threedimensional convex region R, the operation of replacing any one of them by any point of R on the same side of the plane of the other
three
is
called an elementary transformation.
The
set of all
ordered
tetrads obtainable
sequences of elementary transformations I) is called a from one noncoplanar ordered tetrad of points
by
finite
ABC
sense-class
denoted by S(ABCD). The theories of sense in a three-dimensional convex region and in a three-dimensional Euclidean space are related in just the same
is
and
way as the corresponding planar theories. Hence we have THEOREM 21. In a three-dimensional convex region there are two even and altered l>y odd and only two senses. Sense is preserved D and E are on opposite permutations of four points. Two points
liij
sides
of a plane
ABC
if and only if
S(ABCD)
S(ABCE).
EXERCISES
The whole theory of order relations can be developed by defining senseclass on. a line by means of elementary transformations instead of as in Chap. IF. *2. Develop the theory of order in two- and three-dimensional convex regions, defining sense-class in terms of elementary transformations and using Assumptions A, E, S or Assumptions I-VIII of 29 (cf. Theorem 5, 14-8)
1.
as basis.
is said to 3. An elementary transformation of a triad of points Pn if it carries a point be restricted with respect tr> a set of points v P2 of the triad, say C, into a point C" such that the segment CC' does not
ABC
contain any point collinear with two of the points P v P 2 Two n ordered triads of points are in the same sense-class if and only if there is a sequence of restricted elementary transformations carrying the one triad,
, , .
into the other.

4.
Generalize the notion of restricted elementary transformation to space.
162. Euclidean theorems on sense.
the point at infinity of a Euclidean line invariant
The involutions which leave may be called

reflections is a parabolic
point
reflections.
The product
of
two point
projectivity leaving the point at infinity invariant,
and may be called
a translation.
(1)
point reflection has an equation of the form

?'
+ &,
form
and a translation has one
of the
The point reflections interchange the two sense-classes of the Euclidean line, and the translations leave them invariant.
tions
or,
In generalizing these propositions to the plane, the point reflecmay be replaced by the orthogonal line reflections (Chap. IV)
indeed, by the set of all symmetries, and the one-dimensional by the set of all displacements in the plane. Since an orthogonal line reflection in the plane interchanges the two sensetranslations
classes,
leaves each of
is
any symmetry interchanges them, but any displacement them invariant. The generalization to three-dimensions
similar.
The equations
of)
of a displacement in two or three (or any number dimensions are a direct generalization of the one-dimensional
equations, namely,
(3) a/ t
= JV,^ +*<>
j=i
(*
= !'
' >
'
*'
w)
where the matrix (a^ is orthogonal and the determinant a^\ is 4-1. The equations of a symmetry satisfy the same condition except that the determinant a 1 instead of 4- 1. is y
It is worthy of comment that the distinction between displacements and symmetries holds in the complex space just as well as in the real, whereas the distinction between direct and opposite
collineatious holds only in the real space. Algebraically, this is because the distinction of sense depends merely on the sign of the
determinant
symmetries and ag =
of
is
whereas the distinction between displacements and between collineations satisfying the condition |a y =4-1 1. In the representative spaces of six and twelve dimena^.
,
sions referred to in
158, |a tf
1
|
and
\ay
=-1
are the equations
nonintersecting
loci.
From the point of view of Euclidean geometry, as has been said above, the two sense-classes are indistinguishable.* In the applications of geometry, however, a number of extra-geometrical elements enter which make the two
* This does not contradict the existence of a geometry in -which one sense-class is unlike specified absolutely in the assumptions. The group of such a geometry the Euclidean group in that it does not include symmetries though it does include of the displacements. Its relation to the Euclidean geometry is similar to that geometries mentioned in the fine print in 116. Those geometries, however, correspond to groups which are not self-conjugate under the Euclidean group, whereas this one corresponds to a self-conjugate subgroup. On the foundations of geometry in terms of sense-relations taken either absolutely or relatively, see the article by
is
416

identifies the abstract Euclidean space
[CHAP. IX
sense-classes play essentially different roles.
who
Thus any normal human being with the space in which he views
himself and other material objects may as single out one of the sense-classes
follows: Let
him hold
in such a
way
his right hand that the index finger is
in line with his arm, his middle finger

at right angles to his index finger, and his thumb at right angles to the two fingers (fig. 82). Let a point in his palm
be denoted by 0, and the tips of his thumb, index finger, and middle finger by X, Y, Z respectively. The sense-class
S (OXYZ)
or positive,
negative.
shall be called right-handed and the other left-handed or

is
cause
of
This designation the mechanical
unique beof
structure
the body.
A
is
nonhomogeneous coordinate system

right-handed or positive
if
called
if
and
FIG. 82
only
S (OXYZ)
0, 0),
0, 1).
is
0,
positive
when
1, 0),
=
and
(0,
X = (1,
0)
Y= (0,
Z = (0,
The
reader will find
it
sense-class is called positive to identify it
convenient whenever an arbitrary with the intuitively .right-handed
sense-class.*
163. Positive and negative displacements.

if.
On
a Euclidean line,
such, that
a translation carries one point

positive, it carries
to a point
S(AB)
is
is
any point
to a point
B
if
such that S(AB]

it
positive.
is
Such
a translation is called positive.
Any
other translation
carries
called negative
is
and has the property that
to D,
S(CD)
negative.
Any
i.e.
translation carries positive translations into

if
positive translations;
translation,
is
T'
is
a positive translation
and
any
1
TT'T~ l
is
is
a positive translation.
is
A translation
x'= x +
positive or negative according as &

siich that
positive or negative, provided

positive.
that the scale
S(PQP1 )
is
The inverse
of a
positive translation is negative.
The distinction between positive and negative translations is quite distinct from that between direct and opposite projectivities, for all translations
are direct.
carrying a point
be positive
negative.
if
to a point
is
not collinear with
and
if
is
said to
is
S(OAB)
is
positive
and
if
to be negative
S(OAB]
It
easily proved
that
S(OAB)
is
positive
for one
value of
rotation
A
is
it is
positive for all values of A.
The inverse
of a positive
negative.
Any
displacement transforms a positive rotation
into a positive rotation.
rotation
is
a product of
two orthogonal
line reflections {IL*}
and {mJ/oo} such that the lines I and m intersect in 0. Hence the ordered pairs of lines which intersect and are not perpendicular fall
into
two
classes,
which we
shall call positive
and
negative respec-
tively,
according as the rotations which they determine are positive
or negative.
In a three-dimensional Euclidean space let A be a point not on be the which is not a half-twist, let foot of a perpendicular from A on the axis of the twist, and let A' and
the axis of a given twist
0'
The twist
be the points to which A and O respectively are carried by the twist. is said to be positive or right-handed if S(OAO'A') is posi-
tive or right-handed
is
and to be negative
or left-handed
if
S(OAO'A'}
A, so
negative.
It is easily seen that
S(OAO'A')
is
is
the same for
all choices of
that the definition just made inverse of the twist carrying

to
independent
of the choice of A.
The
and
to 0'
and A'
carries 0'
and A'
and A, and thus
is
positive
Since S(0'A'OA)
tive.
= S(OAO'A'},
if and only if S(O'A'OA) is positive. the inverse of a positive twist is posi-
Any
direct similarity transformation carries a positive twist
into a positive twist.

the choice of the right-handed sense-class described in the fine print 162, the definition here given is such that a right-handed twist is the displacement suffered by a commercial right-handed screw driven a short
With
in
distance into a piece of wood.
Since a twist
is
a product
of
two orthogona
line
reflections,
follows that the pairs of ordinary lines Im which are not parallel, intersecting, or perpendicular fall into two classes, {IL} is positive or negative. according as the twist {mm*}
{#}
{mm*},
it
We
shall call the line pairs of these
two
classes positive and right-handed
or negative,
and left-handed respectively. is positive, the ordered pair mJ is positive if Im is posipositive twist tive. Hence a pair of lines is right-handed or left-handed without
direct similarity transforregard to the order of its members. Any mation carries a right-handed pair of lines into a right-handed pair and a left-handed pair into a left-handed pair.
Since the inverse
of a
EXERCISES
1.
The
collineations
which are commutative with a positive displacement
are all direct. (or with a negative displacement) 2. By the definition in 69, 0<4.AOB<tr or
Tr<4.AOB<2tr according as
S(OAB)
3.
is
that the points 0, positive or negative, provided
P P^
,
are so
chosen that
S(OP P^)
is positive.
By the
definition in
72,
0<m(l^z )<-
or
- <m (7^) <TT according as the
an. ordered line pair l^l^ in a plane as being either the operation of replacing ^ or / 2 by a line parallel to or 72 say l v by a line through the point itself, or the operation of replacing ^ which, is not separated from l v by 12 and the line through. l^l z perpendicuIJ.2
,
ordered line pair IJ2 is positive or negative. 4. Let us define an elementary transformation of
lar to y Two ordered pairs of nonparallel and nonperpendicular lines are if and only if they are both in equivalent under elementary transformations the positive or both in the negative class.
l
5. Let us define an elementary transformation of a pair of nonparallel and in space as the operation of replacing one of the 2 nonperpendicular lines lines, say llt by a line intersecting Zj and not sejjarated from / x by the plane of intersection the perpendicxilar to ^ and the plane through point through
this point
and
by a sequence
The pair l^2 can be transformed into a pair of lines m^m^ z of elementary transformations if and only if both pairs are
.
right-handed or both pairs are left-handed.
164. Sense-classes in
projective
spaces.
It
has heen seen
in
18 and 32) that the distinction between direct and opposite collineations can be drawn in any projective space of an odd number of dimensions which is real or, more generally, which
Chap. II
(cf.
satisfies
A, E,
S.
\a..
This depends
(i,j
0, 1,
(
,
32) on the fact that the sign of

n) cannot be changed
if
a determinant
plying every element by the same factor
is
odd, and
by multican be
1 if n is even. changed by multiplying every element by In a real projective space of odd dimensionality the direct collineations form a self-con jugate subgroup of the projective
group and
thus give
rise
to
the definitions of sense-class in
19
and
32.
164,165]
SENSE IN PllOJJiUTlVE SPACES
419
frame of reference as in the Euclidean cases, and the criteria for sense in terms of 24 products of determinants are given in and 32. If one forms the analogous determinant products for the
of the
projective spaces of
the product
point
even dimensionality,
verifies
it is
found that the sign
of
may be changed by
which
by
1,
multiplying the coordinates of one in a second way that there is only one
sense-class in a projective space of
an even number of dimensions.
The projectivity
may be represented by means

jective
of a point (a , n
a ln
2l ,
a 22
in.
a prodirect
space of three dimensions.
The points representing
projectivities are
on one
side of the ruled quadric
= 0,
and those representing opposite
with that considered in
129.
projectivities
on the other
side,
This representation of projectivities
by points is in fact identical It can be generalized to any num158.
ber of dimensions just as are the analogous representations in

It readily
follows that the group of all projective collineations in a real space of n dimensions is continuous if n is even, and not continuous if n is odd. If n is odd the group of direct collineations
is
continuous.
In the following
sections
sense-classes of projective spaces
165-167) we shall discuss the by means of elementary transfor-
mations, the latter term being used as before to designate a particular type of continuous deformation. After this ( 169-181) similar
considerations will be applied to other figures. 165. Elementary transformations on a projective line. DEFINITION.
Given a
set of
mation
point
is
!
three collmear points A, B, C, an elementary transforthe operation of replacing any one of them, say A, by another or C. such that there is a segment AA' not containing
THEOREM
line
the
22. Two ordered triads of points on a real projective have the same sense if and only if one is transformable into other l>y a finite number of elementary transformations.
the other by a finite number of elementary transformations. Tlie converse statement, namely, that a triad A, B, C can be transformed
in. the by elementary transformations into any other triad A'B'C' same sense-class, follows at once if we establish (1) that ABC can
and
be transformed by elementary transformations into BCA and ordered triad of points A, B, C can be transformed (2) that any
CAB
by elementary transformations into one of the six ordered triads formed by any three points A', B', C'. D be a point in the order {ABCD}. Then by elementary (1) Let
transformations
then into
we BCD, and
more
(2)
ABC
can change ABC into ABD, then into A CD, then into BCA. By repeating these steps once can be transformed into CAB.
If A! does not coincide with one of the points A, B, C, it is on one of the three mutually exclusive segments ( 22) of which they are the ends and by (1) the points ABC may be transformed
;
so that the ends
of
this
segment are
1
and
C.
Hence we have
{ABA'C},a.nd by elementary transformations AB C goes successively into AA'C, BA'C, BA'A, BA'C. If A does coincide with one of the points A, B, C the triad ABC may be transformed according to
s
(1)
so that A'
= A.
transformed into A',

in
In like manner the three points A' BO can be B C in some order, and then A'B' C into A'B' C'
1 ,
some order. The proof given

DEFINITION.
for this
S.
theorem holds good without change on the basis

15.
of
Assumptions A, E,
Cf.
An
I
-
ABC
points
of a line
elementary transformation of a triad of points is said to be restricted with respect to a set of

it n if
P v P
2 2
',
P
P n
carries
a point C' such that

points
C and
(C or
C' are
C'
one point of the triad, say C, into not separated by any pair of the
coincide with any of the points
Pv P ,'
may
-TP
P\
-Siv
It is obvious that
any elementary transformation whatever

of restricted
is
the
resultant of a finite
elementary transformations. Hence Theorem 22 has the following immediate corollary

:
number
COROLLARY. Let
P vP
z,
P n
be
I
Two
one
ordered triads of points of
any finite set of points on a line I. have the same sense if and only if
a finite number of elementary
to
is
transformable into
the other 'by
transformations restricted with respect
Pv P%,
>,
P n
Jo
le
3
elementary transformations restricted with respect to

arbitrarily small.
Pv P
z,
Pn can
made
166.
iven a set of four points in a projective plane,

uear, aii
:
Elementary transformations in a projective plane. DEFINITION. no three being colelementary transformation is the operation of replacing one of the same plane joined to the point replaced
of
them by a point
a points.
segment not meeting any side of the triangle
the other
iree
THEOREM
23.
tnyles in the
If ABCD and A'B'C'D' are any two complete guadsame projective plane, there exists a Jinite set of elemen-
ry transformations changing the points A, B, C,

cpectively.
D into A', B
1
,
C',
Proof. It can be
onal case, just as in the proof of
shown by means of the Theorem
result for the one-dimen18, first that the ordered
trad
ABCD
can be carried by elementary transformations into ao
FIG. 83
dered tetrad
B'C'D' and then that
A"B"C"D" of points on the sides of the A"B"C"D" can be carried by
quadrangle elementary
msformations into some permutation of A'JB'C'D'.
To complete the
proof
of
tion of the vertices
it is necessary to show that any permutaa complete quadrangle can be effected by
elementary transformations. Given a complete quadrangle intersection of the lines A 2 A S and

points in the order
tion of
AÂ A^
Z
let
A^,
Let
2,
s,
and
6
let
C and
l
be the point of Z> be two t
{A^B^DÂJ.
let
A C
2
with AyD^ and
Ca
X>
B C
3,
be the point of intersec<7 f B 6 be the points 4

,
D
:
defined by the following perspectivities

s
(fig.
83)
By Theorem 7, Chap. II, it follows that no two of the pairs of points AÂ^, A 2 A & AÂ & A A Z and A A are separated by the lines joining
,
the other three of the points A lt A , A & A^, A 6 Hence there exist z elementary transformations changing each of the following sets of
,
.
four points into the one written below it:
Hence the permutation
n=
1
can be effected by elementary transformations.

notation in 1^
it is
By
changing the
clear that
.can be effected by elementary transformations. Hence the product 2 (i.e. the resultant of II applied twice and followed by II ), 2
nn
which
is
(AAA
be effected by elementary transformations.
can
also
Hence any
of
vertices can
elementary transformations, and hence any permutation of the be effected by means of elementary transformations.
167.
Elementary transformations in a projective space. DEFINITION.
Given a set of five points in a projective space, no four of the points being coplanar, an elementary transformation is the operation of
replacing any one of them by a point joined to it by a segment not meeting any plane on three of the other four. It follows man 27 that the determinant product (25) of 32 is
unaltered in sign
the points
by any sequence of elementary transformations of whose coordinates are the columns of (21) in 32. Hence
a sequence of elementary transformations cannot carry an ordered pentad of points from one sense-class into the other. Hence the odd permutations of the vertices of a complete five-
point cannot be effected
by elementary transformations.
Let
C,
That the
even permutations can be thus effected
may
and
be
seen
as
follows:
the
vertices be
let
denoted by A, B,
D,
the line
DE
meet the plane
ABC in a point F. This point is not on a side of the triangle ABC. Let
A
1
lines
be the point of intersection of the FA and BC, B' that of and
FB
A r &**'
CA, and C' that of
FC
to Bi x>
_,
Ji
and AB.
Let
FIG 84
A^ be a point in the order {BAÂ'C} B l the point in which the line (fig. 84) and
FA^ meets AC,
so that
{ABjB'C}. Let B z be a point in the order {AB^B^B'C}. We now can transform ABODE by elementary transformations
successively
into
AA^CDE, AA^B^DE, BA^B^DE, BCB^DE,
BCADE. Thus
the even permutation
(ABGDE\ \BCADE)
can be effected
by elementary transformations.
is
It is easily verifiable
that any even permutation

this type.
a product of even permutations of
It can
that any five points
be proved by the same methods as in Theorems 18 and 19 no four of which are coplanar can be carried
details of this proof are left as

is
an exercise
to the reader.
When
1
this
combined with the paragraph above, we obtain
THEOREM
if
24.
In a
real protective space,
S(ABCDE}-=S(A'B'C D'E'}
and only
The
if there exists a sequence of elementary transformations

C,
carrying the points A, B,
D,
E into
A',
1 ,
6", D',
respectively.
proof just outlined for this

S, P.
of
Assumptions A, E,
theorem holds good on the basis Assumption P comes in because of the

This, however, can be avoided
is
;
use of a coordinate system.
and the
construction of a proof on the basis of A, E, S alone to the reader as an interesting exercise.
recommended
*168. Sense in overlapping convex regions. The discussion of sense in convex regions by means of elementary transformations in 159-161) is essentially the same for any number of (as made
dimensions.
point in
Now
if
two regions
of the
same dimensionality have a
common, they have at least one convex region of that dimensionality in common. Assigning a positive sense in this region determines a positive sense in each of the given regions. Thus if we have
a set of convex regions including all points of a space,, we should have, on assigning a positive sense to a tetrad of points in one region, a positive sense determined for any tetrad of points in any of the
Since, however, it is in general possible to pass from one region to another by means of different sets of intermediate regions, the possibility arises that this determination of sense may not be
regions.
unique.
tive
In other words,
senses.
it is
logically possible that a given tetrad
in a given region might, according to this definition, have both posi-
and negative
The determination of sense by this method is unique in projective spaces of odd dimensionality and is not unique in projective spaces
shall prove this for the two- and threeof even dimensionality. dimensional cases, but since it reduces merely to a question of even and odd permutations the generalization is obvious.
We
25. Tliere. exists a unique determination of sense for three-dimensional convex regions in a real projective three-space, "but not for all two-dimensional convex regions in a real projective
all
THEOREM
plane.
Proof.
it
Consider
intn fmip
first
the plane and in
it
a triangle
WP.
slifl.ll
ABC
decom-
fvria.Ticrnlar rficrirms
wTiiV.li
rlfinnt.fi
hv the
168,169]
OVERLAPPING CONVEX KEG-IONS

26, Chap. II.
425
I,
notation of
is
Any
one of these regions, say Region

I'
(e. g.,
contained in a
infinity
convex
region, say
a Euclidean plane with
line at
of
the triangular region.
not meeting Region I), which contains the boundary So the determination of sense for Region I
extends to all the points of its
boundary and
also to a portion of
Region II
Let the sense of
ABC
with respect to Region I be
positive.
The
any
segment
aries of
one of the segments
AB (fig.
16), is
common
I'.
to the
bound-
I and II to
I'.
and hence
I'
is
point
common
II.
and
II,
contained in Region and C' are on opposite

29,
If C' is
sides of the line

is
AB
in
in Region
Hence, according to
is
S(BAC'}
positive
Region
Hence S(BAC)
positive with respect to Region II.
Regions II and
a repetition
IV
have in
common
a segment BC, and thus by

is positive
of -this
argument S(CAB}
with respect to
Region IV.
Region I, But by hypothesis
The latter region has a segment AC in common with and hence S(ACB] is positive with respect to Region I.
S(ABC)
is
Hence there
is
not a unique
is
I. positive with respect to Region determination of sense in a real
projective plane.
To show that there
a unique determination of sense for a real
projective three-space, let a given sense-class
S(ABCDE]
(cf.
164)
be designated as right-handed,
class
is
and in any convex region
let
a sense-
interior to
S(A'B'C'D') be right-handed if S(OA'B'C'D') is positive, where the tetrahedron A'B'C'D'. This convention satisfies
the requirements laid
down above
for overlapping convex regions
and,
by
167,
is
unique
for the projective three-space.

is,
Any
two-dimensional region whatever

points in
by
definition
set of all
an
infinite set of triangular regions, i.e. in
(155), the an
infinite set of
convex regions. In
like
manner, any three-dimensional
region is the set of all points in a set of three-dimensional regions.
The method given above

sense-class in all
said to
or is
may
be applied to determine the positive

R,
convex regions in a given region
and R
may
be
be two-sided or one-sided according as this determination is not unique. Another, slightly different, method of treating this
This proposition is valid whether the pencils are considered in a protective or in a Euclidean space.* DKKIXITION. In a plane any point associated with one of the senseclasses
among
the lines on this point
is
called
an oriented point, and
a line associated with one of the sense-classes

called an oriented line.
among
its
points
is
Two
oriented points are said to be similarly

I if
oriented with, respect
to
a line
their sense-classes are perspective

/.
with the same sense-class
in the points of
By
Ex.
1,
26,
if
two
oriented points are similarly oriented with respect to a line I, they are if and only if / and do similarly oriented with respect to a line not separate the t\vo points.
.By
30 a direct collineation
of a
is
Euclidean plane transforms any
oriented point into one which,
similarly oriented with respect to
the line at
infinity.
Hence the
oriented points fall into two classes
such that any two oriented points of the same class are equivalent under direct collineations and that the two classes are interchanged
by any nondirect collineation. No such statement as this can be made about the oriented lines in a Euclidean plane, because any oriented line can be carried by a direct collineation to any other oriented line. This is obvious because (1) an affine collineatiou exists carrying an arbitrary line to
any other
line
and
(2)
changed by a harmonic homology whose center

of the line.
It is a corollary of
the two sense-classes on any line are interis the point at infinity
the last paragraph that any oriented line of a
projective plane can be carried into any other oriented line of the projective plane by a direct collineation.
By
duality the same propo-
sition holds for oriented points in a projective plane.
The oriented points determined by associating the points of a segment 7 with sense-classes in the flat pencils of which they are centers fall into two sets, all points of either set being similarly oriented with respect to any line not meeting 7. These two sets shall be
called segments of oriented points and may be denoted by 7 (+) and ~ If A and B are the ends of 7, the two oriented 7 points determined by A and B and oriented similarly to 7 (+) with respect to a
(
*In general, the geometry of a Euclidean space or, indeed, n dimensions involves the study of the projective geometry of n
of
any space
of
I dimensions, in
169]
ORIENTED POINTS
I
427
not meeting 7 or either of its ends are called the ends of 7 (+) and may be denoted by A and J5 (+) The other two oriented points ~ determined by A. and B are the ends of 7 ) and may be denoted by
line
( yi ->
and
/*<->.
In terms of these definitions

of similarly
it is
clear that each of the
two
classes
oriented points determined by a Euclidean plane satisfies a set of order relations such that it may be regarded as a Euclidean
plane.
situation in the projective plane is entirely different. Let us consider a projective line, and let 7 and S be two complementary ~} ( (+) (-) (+) J5 c+), segments whose ends are A and B. Let
first
The
be defined as above, and let S (+)
and
of
B ( ~^
be the two segments
oriented points determined by & and oriented similarly to
"} ( ' respectively with not meeting respect to a line
A (+ and^
o or either of its ends.
Since
A (+)
and
J? <+) are similarly oriented

I,
with respect to
are separated
by
and A and B and m, A
and
B^
86
are similarly oriented
with respect to
S (+) are
Aw
m
J?
and
(cf.
,
Ex.
1,
26,
and
fig.
85).
and the ends
of 8 ( -> are
A^
Hence the ends of and J5 (+) Hence

.
the oriented points
and segments are arranged as follows:
the symbols for segments and their ends being written adjacent. Let A^ I} A 2 be four points in the order {A^B^B^ on a proz
jective line or
on a
conic.
They separate the

,
,
line
21) into four

:
mutually exclusive segments 7^ S x 7 2 S2 arranged as follows
Llie
Letting
symbols for segments and their ends being written adjacent. A correspond to A (+ \ to 7<+) etc., it is obvious that there
1
an oriented point be moved along a projective line in such all oriented points of any segment described are similarly oriented with respect to a line not meeting the segment, the oriented point must describe the line twice before returning to its first position. A motion,
Thus,
if
way that
of this sort will
obviously carry any oriented point of the projective plane point. Thus the oriented points either of a prothe jective line or of a projective plane constitute a continuous family in 156. sense of
into
any other oriented
Let
let
TT
us suppose
denote the projective plane under consideration here and it contained in a projective space S, and let S' be a
Euclidean space obtained by removing from S a plane different from 2 Let S be a sphere of S' tangent to TT TT which contains the line AB. 2 of S and let P be the other point at a point let the center be 2 v
in
meets the sphere. Let P (+) and P } be the two oriented points of tr determined by r A correspondence T between the points of the sphere S 2 and the
which the
line
OP
oriented points of the projective plane following rule Let PI correspond to

:
TT
may now be
,
set
up by the
)
;
(+)
and
to
P~
(
if
any point
points in
of IT not
on the line
line
at infinity, denote
by X^ and
X X the
is
2
which the
so that each of the angles
angle
(i.e.
so that the
OX meets the sphere, assigning the notation 4P OZ and 4-P OX is less than a right points X are all on the same side as P of the
1
l
plane through
side of this plane)
parallel to
;
IT,
and the points
the oriented point of TT deterand denote by mined by and joined to P(+) by a segment of oriented points con~~> the other taining no point of the line at infinity AB, and by (+ oriented point determined by X. Let to \ and l correspond 2
Xw
are on the other
X
(
to
the line at infinity AB, and F one of the oriented points determined by it, F () is an end of a
(
X ~\
o-
If
is
any point
of
>
(+) whose other end is P (+) and of a segwhose other end is P ~\ The line Y meets the sphere in two points one of which, Y is an end both of a segment of points l corresponding to tr and of a segment of points
segment
<r (+)
ment
~)
of points
(
of points
X~
i}
() This construccorresponding to cr^. Let Yt correspond to F tion evidently makes the oriented point other than T () which is
.
determined by
correspond to the point other than
in
which
meets the sphere.
The correspondence T
is
one-to-one and reciprocal and makes each
on
S2
inversion plane, this result
In view of the correspondence between the sphere and the may be stated in the following form
:
a one-to-one reciprocal correspondence preserving order-relations between the oriented points of a real protective plane and the points of a real inversion plane.
26.
THEOREM
There
is
of oriented points in this section does not generalize directly dimensions, because there is only one sense-class in a projective plane and, therefore, also only one in a bundle of lines. The discussion of sense in terms of the set of all lines through a point is .therefore possible along these
to three
The treatment
an even number of dimensions. uniform for spaces of any number of dimensions can, however, be made in terms of rays. An outline of the theory of pencils and bundles of rays which may be used for this purpose is given in the next three sections, and an outline of one way of generalizing the contents of the present
lines only in spaces of
discussion
which
is
section is given in
173.
is
of the theory of oriented points in the plane the theory of doubly oriented lines in three dimensions which is given in 180, below.
Another type of generalization
*170. Pencils of rays.

a linear convex region
The term "ray"*

to
is
defined in
23
for
and extended
28 will
The
definition of angle in
148. any convex region in be carried over to any convex
region.
DEFINITION. The set of all rays with a common origin in a planar convex region is called a pencil of rays. The common origin is called the center of the pencil.
The order
as those
relations in a pencil of rays are essentially the
among the
points of a projective line.
This can be
same shown
by setting up a correspondence between the rays through the center 69. of a circle and the points in which they meet the circle, as in It can also be done on the basis of Assumptions A, E, P alone by proving Theorems 27-33, below. The proofs of the theorems are not
given, because
they are not very different from those of other theo-
rems in
this chapter.
third
way
of deriving these relations is
indicated in
Theorems
34, 35,
and a fourth in Theorems 37-41.
THEOREM 27. If a, 5, c are three rays of a pencil, and if any segment joining a point of a to a point ofc contains a point of I, then every segment joining a point of a to a point of c contains a point of 5.
* In some books the term " ray "
is
used as
synonymous with
"
projective line,"
DEFINITION.
between a arid
If a,
c if
It,
c are
and only
if
three rays of a pencil, b is said to be a aiid c are not collinear and (2) (1)
any segment joining a point
of
a to a point of
contains a point of b.
If b is any ray between two rays a and c, any other ray between a and c is either between a and b or between b and c. No ray is both between a and b and between b and c. Any ray 'between a and b is between a and c.
THEOREM
28.
THEOREM 29. There is a, one-to-one reciprocal correspondence prea segment of a line serving all order relations between the points of and the rays between two rays of a pencil.
THEOREM 30. // three rays a, b, c of a pencil are such that no two of them are collinear and no one of them is between the other two, then any other ray of this pencil is between a and b or between
b
and
or between
and
a.
DEFINITION.
Given a
set of three distinct rays a,

is
c'
b, c
of
a pencil,
of replac-
by an elementary transformation ing one of them, say c, by a ray

that neither a nor b
is
meant the operation

not collinear with
c'.
and such
sequences denoted
between
and
The
class consisting of all

finite
ordered triads into which abc
is
transformable by
of elementary transformations is called a sense-class
and
is
by
S(abc).
elementary transformation of abc into abc' is said to be restricted with respect to a set of rays a v a z an of the pencil if none of , a n is between c and c'. the rays a lf a 2>
,
An
pencil.
an be an arbitrary THEOREM 31. Let a a 2 Two ordered triads of rays of the pencil
1?
,
,
set
of rays of
a,
are in the
sense-class if
and only if one can

a2
32.
be
sequence of elementary transformations

respect to a^,
,
transformed into the other which are restricted with,
same by a
an
Let a i} a 2 a s be three distinct rays of a pencil such that no one of the three is between the other two. There exists a oneto-one reciprocal correspondence T between the rays of the pencil and
,
THEOREM
the points
of a projective
line
such that
is restricted
to
formation of the rays which

there
line
each elementary transwith respect to a , 2 8

,
corresponds
an elementary transformation on
respect
to
the projective
which
is
<*_.
restricted with
the points
corresponding
to a,,
a.
follows
spondents of a lt a z>
of
Let three arbitrary collinear points A^, A.,, A & be the corrett let F be a projectivity which g respectively; I
carries the lines wliich contain the rays
between
a^
and a to the
the segment complementary to A AyA z and carries the line containing a, to A for the rays between a and a let T be J 1 1
points
', '
the correspondence in
to
and 2 corresponds a^ the point to is carried by F let t F 2 be the projectivity wliich carries the lines which contain the rays between a z and a 8 to the points of the segment complementary to
; ;
which each ray between which the line containing it
12
AyAÂ & and carries the line containing 2 to A z for the rays between and a 3 let F be the correspondence in which each ray corre2
sponds to the point to which the line containing it is carried by F 2 let F g be a projectivity wliich carries the Hues which contain
;
the rays
between a s and a l
let
to the points of the
segment complemen;
AyAyA^ and between 8 and a^

tary to
A 6 for the rays be the correspondence in which each ray corresponds to the point to which the line containing it is carried
carries the line containing a a to
tyr
the
There is a one-to-one reciprocal correspondence between points of a protective line and the rays of a pencil such that two ordered triads of rays of the pencil are in the same sense-class if and only if the corresponding triads of points are in the same sense-
COROLLAEY.
class
on the
line.
THEOREM
33.
If
a,
l>,
are three rays of a pencil
and
a',
V,
c'
are
the respectively opposite rays,
S(abc)=
S(a'b'c').
pencil,
'
If a and b are any two noncollmear rays of a by an elementary transformation of the ordered pair ab is meant the operation of replacing one of them, say b, by another of the pencil, such that no ray of the line containing a is ray, b
DEFINITION.
between
(i.e.
and
o'
or coincident with
b'.
The
set of all ordered pairs
angles) into
which an ordered pair
of rays ab can be carried

is
and
by sequences of elementary transformations is denoted by S (ab).
THEOREM
A, B,
then
C,
34.
If
if
is
the
center
b,
c,
of a pencil of rays
and
are points of rays a,
respectively of the pencil,
S(ab)=S(cd)
and only if
S(OAB)= 8 (OCD).
THEOREM
pencil,
35.
If a and
S (ab)i=S (ba).
the pencil is either

a,
b are any two noncollinear rays of a Every ordered pair of noncollinear rays in in S (ab) or in 8 (ba). If a' is the ray opposite to
S (ab)
^ S (a'b).
36.
c
THEOREM
a pencil and
then
If and
1
a, b
c'
and
t7ie
a', b
are two ordered pairs of rays

to
of
are
if
is
rays opposite
only if
between a
a and
a' respectively,
1
S (ab)
S (a'b )
c
and
S (abc)
S (a'b'c
and
b
).
The same
c'
conclusion holds if
any ray
and
any ray
between a'
and
37.
4-
b'.
THEOREM
of an angle
DEFINITION.
The points not on
the sides or vertex
ab
f<M
v^to two classes having the sides
and
vertex as
class to
boundary and such that any segment joining a point of one
a point of the other contains a point of the sides or the vertex. If the angle is a straight angle, both of these classes of points are convex
regions.
If
of
not, one
and only
one of them
is
convex and
is
called the
interior
the angle ; the other is called the exterior
of
the angle.
4 A OB,
THEOREM 38. If A' is any point of the side OA of an angle and B' is any point of the side OB, then S (OAB) = S (OA'B ).
1
If Cis any point interior to the angle, S(OAB) = S(OAC) = S(OCB), and any point satisfying these conditions is interior to the angle.
THEOREM 39. Any ray having the vertex of an angle as origin, and not itself a side of the angle, is entirely in one or the other of the two classes of points described in Theorem 37. If it is in the interior
contains one and only one point on each segment joining a point of one side of the angle to a, point on the other side.
it
DEFINITION.
Two
rays
a, & of
a pencil are said to be separated
by two other rays h, k of the same pencil (or by the angle hk) if and only if a is in one and b in the other of the classes of
hk. set of points determined according to Theorem 37 by rays having a common origin are said to be in the order {a 1 a3 8 a 4 a ) if no two of the rays are separated by any of the angles
re
4 a^,
4.
a zas
, ,
4 ^- A> 4 ana r
set
THEOREM
40.
A
Any
and
also in the orders {a a z s
an _ l an} are of rays in the order {a 1 a z a a a n a^ and {an an _ : aâ^}.

a, b
COROLLARY.
the orders {ab}
two rays
{ba}.
Any
three
having a common origin are in rays a, b, c having a common
oriain are in the orders labc\. fbca\. fcab\. facb\. ibac\. lcba\.
THEOREM
41.
mon
origin
may
n}-
be assigned
To any finite number n ^ of rays having a coma notation so that they are in the order
{<WV a
*171. Pencils of segments and directions.
The
notion of a ray
belongs essentially with that of a convex region, but the theorems of the last section may easily be put into a form which is not
limited to
convex
regions.
The proofs
are all omitted for the
same
reasons as in the section above.
DEFINITION.
lying in the
set of all
is
segments having a common end and
same plane
called a pencil of segments.
The common
end
is
(ailed the center of the pencil.
Two
segments or intervals
directed at
having a
of
common end A
is
are said to
be similarly
AH either
entirely contained in the other. The set of all segments similarly directed at a given point with a given segment is called a direction-class or, more simply, a direction. The set of all directions
them
of the
tions.
segments of a pencil at its center is called a pencil of direcThe directions of two collinear segments having a common
end
and not similarly directed are said

if
to be opposite,
and the
two segments are said to be oppositely directed at A.
Thus
segments
ABCD are four collinear points in the order {A3 CD} the ABC and ABD are similarly directed, while ABG^xA ADC
At a given point on a given line there are obviously two and only two directions, and these are opposite to each other. Two noncollinear segments with a common end are conare oppositely directed.
tained in one and only one pencil, namely, the one having the common end as center and lying in the plane of the two segments. DEFINITION. segment a- is said to be between two noncollinear
segments
a-
cr^
cr
2
if
is
similarly directed with a
the three segments are in the same pencil and segment which is in the pencil and
contained entirely in the triangular region determined by <rt and <T Z direction d is said to be between two noncollinear (Theorem 12).
directions
tions d,
d if dz
if
there exist three
d^ dz respectively such that

extension
of
segments <r, cr^, <TZ in the <r is between <r and cr 2 a

.
direc-
This
justified
the notion
of
betweenness
to
directions is
by the following theorem.

42.
/">
THEOREM
*
If a and
J
I
(3
are two noncollinear segments with

"/w,
a
/Q
J Of
'Tfimln,
^nnnn4a/J
nnn'+ll
/*
rlmfj
is
meant the operation of replacing one of them, say crg by a segment cr4 which is in the pencil and such that neither <T I nor crz is between <ra and cr 4 or similarly directed with <r^ A class consisting
,
of all ordered triads into
which
o\cr 2
og
is
transformable by finite
sequences of elementary transformations is called a sense-class and dz d a are three directions of a pencil, is denoted by ^\> ^(crâ-^. and <r a cr 2 <r s three segments in the directions d^ d z d 3 respec-
>
tively, the sense-class
S(d^dz d^
is
the class of aU triads of directions

of
which
are the directions of triads
segments in the sense-class
S(Ws)THEOREM
43.
If
<r^
cr
2,
cr s
are three segments of a pencil, no two

!
of them being similarly directed, and
cr
s
is
similarly directed with
'v^fjr^S&frt^ THEOREM 44. There
is a one-to-one reciprocal correspondence be-^ tween the directions of a pencil and the points of a line such that two triads of directions are in the same sense if and only if the
corresponding triads of points
ham
of
the
same
sense.
We now
take from
2123
Chap. II the definitions of sepaof
ration, order, etc.,
and on account
Theorem 44 we have
at
once
COROLLARY
applied
to
1.
The Theorems of
21-23 remain valid when

to
the directions
of a pencil instead of
the
points of
line.
COROLLARY
2.
Two
Let
pairs of opposite directions separate each other.

1
DEFINITION.
pencil;
a-
and
o2
by an elementary transformation
,
be two noncollinear segments of a is meant the operation of
replacing one of them, say cr2 by any segment tr g of the pencil such that no segment' collinear with cr l is between <72 and o- 8 The set of
.
all
ordered pairs of segments into which er^ is transformable by sequences of elementary transformations is called a sense-class and
is
segments a~ 1 cr 2 is transformable by elementary transformations into a pair o-^, then cr^ is transformable by elementary transformations into crcr
denoted by ^(cr^cr^. THEOREM 45. If a pair of
not collincar
with
cr,,,
nor
cr
s
with
<r
4,
then either
$(0^) = S(<r s a^
to <r
,
or
to
and
cr
Let d l and dz be two directions of a pencil and let be two segments in the directions d and ^ respectively. 2 l By the sense-class /V(^//,) is meant the class of all ordered pairs of
DEFINITION.
<T
I
and
cr
directions
which are the directions

S(crjr^.
of
ordered pairs of segments
in the sense-class
It is
evident that the last two theorems
may be
restated, without
material change, in
terms of directions instead of segments.
*172. Bundles of rays, segments, and directions. DEFINITION. The

set of
all
rays in a three-dimensional convex region which have a

is
common
Let a,
''ury
origin
called a bundle of rays.
The point
is called
;he center of
b, c
the bundle.
be three noncoplanar rays of a bundle.

is
By
an elemen-
transformation
say
e is a,
meant the operation

such that no ray
a'.
of replacing one of the
ays,
by a ray
a'
of tlie
plane containing
md
nto
between a and
The
set of all ordered triads of rays

trans-
ormations
which abc can be carried by sequences of elementary is called a sense-class and is denoted by S(abc}.
48.
THEOREM
If
abc,
oplanar rays having a

vints of the rays a, f and
and a'b'c' are two ordered triads of noncommon origin O and A, J3, C, A', B C' are
1
',
b, c, a', b', c'
respectively, then S(abc)= S(a
b c
only
if S(OABC] =
49.
S(0A'B'C'}.
are three
b', c'
THEOREM
l(abc)
If
=f=
a,
I,
noncoplanar rays of a
are
bundle,
= S(bca)
50.
S(acb}.
If a,
any oilier
three noncoplanar
ays of the bundle, either
S (a'b'c'}
to
S(abc} or S(a'l'c')
=
S (acb).
THEOREM
nd
If
a, b, c
are three noncoplanar rays of a btmdle
a' is the
ray opposite
the
a f S(abc)
S(a'bc}.
THEOREM
it
51.
If abc are
[x]
of rays of
noncoplanar rays of a bundle, the bundle which satisfy the relation S(xab} =
three
1(zbc]
= S (xca)
with the points of a triangular region
are in such a one-to-one reciprocal correspondence xa that if rays x l} 2

, ,
x t> x s> X 6 corTesP ona
to
points JL I} JC 2
S(^xz x s
A, 3,
= 8(x?F}
the interior
if
and only if
b, c
S(X XV X^ = 8(X XSX^.

J
.A
4,
JL
5
If
are points of the rays a,
respectively,
and
the
region
is
of
the triangle
ABC, F may
to
le
triangular taken as the

in which
it
correspondence in which each x corresponds meets the triangular region.
the point
THEOREM 52. If a, I, c, d are four rays of a bundle such that any plane containing two of them contains a ray between the other two, any other ray of the bundle is between two rays of the set a, b, c, d or
in one of four sets [], [y],
S(xbc]
isfies
[z],
[w] such that [x\ satisfies the condition
= S(xdb), [y~\ satisfies S(yac) S(ycd) = S(yda), [3] sat= S(zbd) = S (zda], [w] satisfies S(wab} = S(wbc} = S(wca}. S(zab)
=S(xcd)
COROLLARY.
Under
the conditions of the theorem if A,

b, c,
_Z?,
C,
are points of the rays a,

is
respectively, the center
of the bundle
interior to the tetrahedron
ABCD.
common end
is
DEFINITION.
set of all segments having a
called
a bundle of segments.
The
set of all directions of
the segments of a
bundle
is
called a bundle of directions.

,
Let <T I} <rz <rg be three segments of the same bundle, but not in the same pencil; the operation of replacing any one of them, say cr z> by a segment <rt of the bundle such that no segment
DEFINITION.
of the pencil containing
ar^
and
<r
z
is
between
cr
&
and
o~
4
or coincident
with
of all
finite
cr
i
is called
an elementary transformation.
class consisting
ordered triads of segments into which o" 1 o" 2 o" 8 can be carried by sequences of elementary transformations is called a sense-class <r ). and is denoted by $(cr,cr * 1 O'
*
<8
Theorems 48-52 to the corresponding theorems for a bundle of segments presents no difficulty. *173. One- and two-sided regions. discussion of the order relaof
The generalization
tions in protective spaces
which
is
closely analogous
both to
168
and to
Let
169
may
details are left
be made according to the following outline. as an exercise for the reader.

Let A, B,
to A,
The
be any point of a planrr region R.
be the
vertices of a triangular region
containing
and contained in R,
B>
and
let a,
/3,
7 be the segments in R joining

a', ft'
C
R
respec-
tively.
Then S(afy = S(j3y) = S(ya).
If 0' is
0' to
any other point of T, and and B respectively, S(<x@)
the segments of
joining
is
said to be like S(a'f3'};
and
S(ap)
i
is
like
#()
A
aud
<% P
like
#(
V3
),
tlien ^(a/S) is said
be
fo'/M
ti(a"ft").
region for
which a given
be two-sided.
sense-class at one
)int is
like the other sense-class at that point is said to be one-
ded.
Any
other region
region
is
is
said to
A convex
gion.
two-sided.
protective plane
is
a one-sided
be any point of a three-dimensional region R. Let A, B, be the vertices of a tetrahedral region T containing and ntained in R, and let a, /3, y, B be the segments in R joining
Let
A, B, C,
If 0' is
respectively.
any other point

0' to A, B,
if
Then S(a/3y) = S(j3a$) = /S(S 7/3) = %8er). of T, and a', /?', 7' are the segments of
respectively,
joining
$ (a/fry)
is
said to
be
like
V/3V);
en
S(afa)
is
like S(a'j3'v'},
and
(a'/3V)
is like S(a"P"<y"),
S(a]3j) is said to be like S(a"V)One- and two-sided regions are defined as in the two-dimensional
se.
Any
174. Sense-classes
region in a three-dimensional projective space on a sphere. The theorems in
is
two-sided.
172 can be
yarded as
carried over to the sphere
denning the order relations among the points of a sphere by letting each point of the sphere
respond to the ray joining it to the center of the sphere. Another y of treating the order relations on a sphere and one which meets directly with 97 is as follows
:
DEFINITION.
Let A, B, C, D be four points of a sphere not all the same circle. By an elementary transformation is meant the
one of them, say A, by a point A' on the BCD. The set of all ordered tetrads into
oration of replacing
ne side of the circle
dch
ABCD
is
.tions is
53.
transformable by sequences of elementary transforand is denoted by S(ABCD).
THEOREM
There are two and only two sense-classes on a sphere.
THEOREM
ere
a',
54.
S(ABCD) = S(A'B' C'D'}
if
t
and only
a'
if
W>
0,
T&(AB, CD)
b, I'
= a + 1 V^T/,
-&(A'B'
C'D')=
+ &'V^
and
are real.
Order relations on complex lines. In view of the isomorism between the geometry of the real sphere and the complex ijective line (cf. 91, 95, and 100) the theorems of the section
175.
)ve
and
of
97 determine the order relations on any complex
line.
One very important

in the real
difference
between the situation as to order

is
is
and the complex spaces or space one sense-class on a line
the following
carried
In a real plane
by
projectivities of a
continuous group into both sense-classes on any other line. So that fixing a particular sense-class on one line as positive does not deter-
mine a positive sense-class on

however, an ordered set
or the
all
other lines.
On
is
a complex line,
of four points
is
ABCD
in one sense-class
other according as I
positive or negative,
"*/^l
= lfc(AB,
CD) and a and
&
are real (Theorem 54).
where a -f- b In conse-
quence of the invariance of cross ratios under projection, a given sense-class on one line goes by projectivities into one and only one
sense-class on
any other
line.
Hence
if
one sense-class
is
called
positive on one line, the positive sense-class can. be
determined on
projective with
every other line as being that sense-class which the positive sense-class on the initial line.
analytic geometry that
is
This connects very closely with the convention for purposes of by Vc is meant that one of the square roots of c which takes the form a, & 1, where a and 5 are real and
+ V
> 0,
or
if
&
that one of
positive.
The symbol V^~I is taken 00 the square roots of 1 for which

a
= 0,
> 0.
to represent
l)
is
176. Direct and opposite collineations in space.
From
the algebraic
definition of direct collineation in terms of the sign of a determinant,
we
obtain at once
55.
jective space
a real three-dimensional proEuclidean space invariant is direct if and only if the collineation which it effects in the Euclidean space is direct.
THEOKEM
Any
collineation of
a,
which leaves
ABCD
In a Euclidean space a point with a point if and only
D
if
is
on the same side of a plane
S(ABCD)
= S(ABCJE).
Hence
a homology whose center is at infinity is direct or opposite according as a point not on its plane of fixed points is transformed to a point
on the same or the opposite side of this plane. Extending this result to the projective space by the aid of the theorem above we have
THEOREM
distinct
56.
homology which carries a point
to
a point
1
A',
from A,
is opposite or direct
according as
A and A
are
the
separated or not separated ly the center of the homology
and
COROLLARY COROLLARY
1.
2.
harmonic ko'niology is opposite. The inverse of a direct homology

is
is direct.
Since any collineation

(
29, Vol. I), it
expressible as a product of homologies follows that
COROLLARY
3.
The
is
inverse of a direct collineation is direct.
Since an elation
the
a product of
it
two harmonic homologies having
same plane of fixed points

4.
follows from Cor. 1 that
COROLLARY
Since
An
elation is direct.
(
line
reflection
101)
is
a product of two harmonic
homologies,
COROLLARY
5.
THEOREM
is
57.
A A
line reflection is direct.
three
skew
lines
invariant
direct.
Proof.
Denote the
lines
by
(
^,
2,
The projectivity on ^ which

three hyperbolic involutions
is effected
by
and the collineation by F. F is a product of two or

is
74).
Each involution on ^
.
effected
by a line reflection whose directrices are the lines which pass through the double points of the involution and meet l and l The product a a II of these line reflections leaves l invariant and effects the same l s
l^
z
,
transformation on
as F.
Hence
1
II~ 1 F leaves
2,
1 3
and
all points
1
1}
on ^ invariant.
l
&,
It also leaves invariant

all
any
l s
line meeting
2,
and
1
and hence leaves
points on
is
the identity, and hence
F = H.
and
invariant.
Hence
H~ T
all
Since the Line reflections are
direct,
is direct. 1.
COROLLARY
lines
Any
collineation
leaving
all
points of two skew
invariant
is direct.
Proof.
Such a
collineation
of
leaves
invariant
three
skew
lines
meeting the given pair
invariant lines.
COROLLARY
2.
Any
collineation transforming
a regulus
into itself
Proof.
Such a
collineation
is
prodi^ct- of
a colliueation leaving
all lines of
the given regulus invariant by one leaving all lines of the conjugate regulus invariant. Hence it is direct, by the theorem.
Corollary 2
is
also
a direct consequence of Cor.
5,
above, and
Proof.
The two
reguli are interchanged by a harmonic

pole
is
homology
56,
whose center and axis are

This harmonic homology
since
its
and polar with regard to the regulus.
opposite
by
Cor. 1,
Theorem
and
product by any collineation F interchanging the two reguli leaves them both invariant and hence is direct by Cor. 2, it follows
that
is opposite.
DEFINITION.
By
a doubly oriented line is
meant a
line
associated
with one sense-class among the points on I and one sense-class among the planes on I. The doubly oriented line is said to be on any point, line, or plane on I.
A
if
A, B,
doubly oriented line may be denoted by the symbol (ABC, ccfiy) denote collinear points and a, /3, 7 planes on the line AB.
For
t-.his
symbol determines the

uniquely.
line
AB
and the sense-classes
r/37)
Since there are two sense-classes
S(ABC) S(ABC]
CB) among the points on a line AB and two sense-classes -- J "'ty ft) among the planes on AB, there are four
Unes '
(ABC, afr), (ACS. 0:7 /3),

7/8),
'hick
transform a doubly oriented
doubly oriented
not meeting
line,
a collineation
T(l).
itself
AB, and V
AB
is
transformed into
The by F
julus containing
be the line harmonically separated AB, I, and V. If V

I and V from met by AB. In either case AB the line reflection whose directrices
lonically separated by
ll'
is
if
is
= Hence AF cum i\\u) Since A and F preserve sense both

^JLJ
'IK,,
I.
leaves both
AB
and
invariant.
in the pencil of points
in the pencil of planes
a/3,
on m. Hence by
74,
AF
and preserves sense both on effects a projectivity on which is a
AF
AB AB
and
AB
product of
A
projectivity
two hyperbolic involutions, {P^P3 } {P,f^, and it effects ou m which is a product of two hyperbolic involutions,
Let
l
i;
(QM'WMrespectively.
2,
t,
be the lines
P&, P&, P^,
P&
The product
{%}
leaves all points
A- r
is
on AP> and on
57.
5,
invariant and
therefore direct
by Cor.
1,
Theorem
by Cor.
1.
All the collineations in this product except
are direct
Theorem
56.
Hence F
is direct.
COROLLAKY
Any
collineation
is direct.
which reverses both
sense-classes
of a doubly oriented line
Proof. Let F be a collineation reversing both sense-classes of a doubly oriented line (ABC, afty). Let a and b be two lines meeting AB but not intersecting each other. The line reflection {ab} reverses
both sense-classes of (ABC, <x/3<y) and is direct. Hence {ab} them both invariant and is direct by the theorem. Hence F
leaves
is direct.
COROLLARY
COROLLARY
2.
Any collineation
and
effects
lines into itself

3.
Any
which transforms each of two skew a direct projectivity on each is direct. collineation which transforms each of two
effects
skew lines into itself and

direct.
an
opposite projectivity on each
is
177. Rightsection can be
and left-handed
figures.
The theorems
of the last
pentads of
used in showing that other figures than the ordered points may be classified as right-handed and left-handed.
is
For this purpose the following theorem
fundamental.
THEOREM
59.
If
the collineations
carrying a figure
into itself
all
are all direct, the figures equivalent to

collineations fall into two classes
under
the
group of
such that any collineation carrying

is direct
a figure of one class into a figure of the same class

collineation carrying it into
and any
be
a figure of
the other class is opposite.
Let [F] be the set of all figures into which F can by direct collineations. There is no opposite collineation collincarrying FQ into an F\ for suppose F were such an opposite
Proof.
carried
eation, let
of
[F] carry
be one of the direct collineations which by definition into F] then P^F would be an opposite collineation Q
itself.
carrying
into
In like manner
it
follows that any collmea-
442
Let
THEOBEMS ON SENSE AND SEPAKATION

[F'~\
[CHAP. IX
site, collineations.
by oppoargument like that above shows (1) that any collineation carrying Jf\ into an F' is opposite and (2) that any It collineation carrying an F' into itself or another F' is direct. follows at once that any collineation carrying an F into an F' or
is
be the set of
all figures into
which
carried
An
an F' into an
F is
opposite.
Since the direct collineations form a continuous family of transformations,
we have
The figures conjugate
to
COROLLARY.
collineations
under
the
group of direct
form a continuous family.

sense-classes of ordered tetrads of
this
The propositions about the

collinear points are
non-
corollaries of
theorem because the only

of noncollinear points into
collineation carrying
itself is
an ordered pentad
all triads of
the identity.
noncollinear lines fall into
By Theorems 57 and 59
classes
two
such that any collineation carrying a triad of one class into a triad of the same class is direct and any collineation carrying a
triad of one class into a triad of the other class is opposite.
It is to
be noted particularly that the triads of lines here considered need not be ordered triads, since by Cor. 2, Theorem 57, the collineation
effecting
any permutation
of a set of three noncollinear lines is direct.
Similar propositions hold with regard to doubly oriented lines,

reguli, congruences,
and complexes
(cf.
178).
Let us
now suppose
that a particular sense-class
S(ABCD)
(cf,
in
a Euclidean space has been designated as right-handed
162).
Any
ordered tetrad of points in this sense-class
is
also called rightis
handed and any ordered tetrad in the other sense-class

left-handed.
called
Let be a point interior to the triangular region BCD, Q the @ the plane APB, 7 the plane APC> point at infinity of the line and S the plane APD. All doubly oriented lines into which (APQ,
AP
is
carried
by direct
collineations
shall be called right-handed
177,178]
EIGHT- AND LEFT-HA]ST DED FIGURES
443
These 'jonvontious give the same determination of right-handed doubly oriented lines and ordered pentads of points no matter what
point of the triangular region
BCD
is
taken as P, because any
col-
Imeation leaving A, B, C, another is direct. In like

of
invariant and carrying one such into manner these conventions are independent
the choice of
A BCD,
lines
so long as
/
x
,
S(ABCD}
is direct,
triad of
skew
l
tl
shall be said to be right-handed or
k/t-hnndeil according as the doubly oriented line (ABC, afty) is is a line meeting l right-handed or left-handed, provided that v / and A, B, C are the points ml v mlz ml s respectively, and a, l. 2 A
Q,
7 are the planes inl^ mln mly respectively. This convention is independent of the choice of
,
in,
by Theorems
carrying a
57 and 58.
By
the
same theorems any
collineatiou
right-handed triad of noncollinear lines into a right-handed triad of lines is direct, and any colliueation carrying a right-handed
triad
of lines
into a left-handed triad
is
opposite.
lines
The reader should verify that a pair

ean space
is
of
skew
Im in a Euclid-
163 right-handed or left-handed in the sense of according as lmlm is right-handed or left-handed, lm being the line of I. at infinity which is the absolute polar of the point at infinity
lmm x is If m^ is the absolute polar of the point at infinity of m, right-handed if and only if lml m is right-handed. Let A be a point of the axis of a twist T in a Euclidean space,
B = T(A), and let C be the point at infinity of the line AB let a be any plane on the line AB, /3 = T(a}, and 7 the plane on AB 163 perpendicular to a. Then T is right-handed in the sense of if and only if the doubly oriented line (ABC, afty] is right-handed.
let
;
This
is
easily verified.
178. Right-
By
Cor. 2,
and left-handed Theorem 57, every

is
reguli, congruences,
and complexes.
is
triad of lines in a regulus
right-
handed or every triad

shall
left-handed.
In the
first
case the regulus

case
to
be said
to be
right-handed and in the second
be
left-handed.
THEOREM
60.
The collineations which leave an
conelliptic linear
gruence invariant are all direct.
congruence through each point of a directrix. Any collineation F which carries each directrix of the congruence into itself effects a is a product of two projectivity on that directrix. This projectivity involutions ( 78, Vol. I). Each involution may he effected by a
line reflection
whose
lines of fixed points are the (real) lines of the
congruence through the (imaginary) double points of the involution; since such a line reflection leaves both directrices invariant, it leaves
the congruence invariant. Hence there exist two line reflections A a A 2 each transforming the congruence into itself, such that
, ,
A^F
A^F
leaves all points on a directrix invariant. Hence transforms each line of the congruence into itself. By Theorem 57,
is direct,
A^F
and by Theorem
56, Cor. 5,
and
are direct.
Hence F
is direct.
If I is any real line not in the congruence, the lines of the congruence meeting / form a regulus, and the directrices are double lines of an involution in the lines of the conjugate regulus. If V
is
the line conjugate to
in this involution, the line reflection {IV}

directrices.
1
must interchange the two

leaves
{U'}
Hence
F'
is
if
F'
is
any colline-
ation interchanging the directrices,
{II }
a collineation which
each
of
F' is direct.
them invariant. Hence by the paragraph above Hence F' is direct. Hence any real collineation
is direct.
leaving an elliptic linear congruence invariant
COROLLARY
1.
The triads of
lines
of an
elliptic linear
congruence
are all right-handed or all left-handed.
For any
collineation.
triad
can be carried into any other triad by a direct
2. If four linearly independent lines are such that of three of them are right-handed or such that all sets of three of them are left-handed, the linear congruence which contains
COROLLART
all sets
them
is elliptic.
An elliptic congruence shall be

of lines in it
is
said to be right-handed
it is
if
every triad
right-handed
otherwise
said to be left-handed.
(
A
is
pair of conjugate imaginary lines of the second kind
109)
deter-
said to be right-handed or left-handed according as
it is
mined by
a right-handed or a left-handed congruence.

(
pair of Clifford parallels
142)
is
said to be right-handed or
they
belong
i.s
agreement with that introduced iu

definitions
parallels
tion
right-lumded or left-handed. This distinction is in 142, because according to both

of right-handed ones is direct, and a collineaof
a collin cation carrying a system of right-handed Clifford
into a
system
a
carrying
system
right-handed
Clifford
parallels
into
system of left-handed ones is opposite.
THEOREM
111.
The
collincu lions
which leave a nondegenerate
linear
complex in.rariitnt are all direct.

Proof.
Let
F
/'.
and let
he any line of
/
he a collineation leaving a complex C invariant, Let I" he any line of C not 6' and l'==T (1).
lines of
meeting
regulus,
stitute
or
The
C which meet
and
I" constitute a
I
and three lines
of this regulus together
with
and
I"
con-
upon which, therefore, all the lines of C are linearly dependent. Hence a collineation which leaves this regulus invariant and inter*
a set of five linearly independent lines
(106,
Vol. I)
changes
larly
and
I" leaves
invariant. Let
I"
I
F' he
I'
a collineation, simileaves
obtained,
The product
direct.
FFT
which interchanges leaves C and
and
and
C invariant.
and T" are
invariant, and
Any
to
C and
invariant leaves invariant the
projectivity II
between the points on I and the planes corresponding them in the null system determined by C. The projectivity II
an arbitrary sense-class among the points on I into an among the planes on L These two sense-classes
transforms
arbitrary sense-class
determine a
points
I,
doubly oriented
carried
line, L
The other
sense-class of the
on
is
by
on II into the other sense-class of planes
and these two sense-classes determine a doubly oriented line I
Since
it
either
any collineation leaving C and I invariant leaves II invariant, transforms this doubly oriented line into itself or into the
is
one obtained
coUineation
particular
that
FFT
direct.
such by reversing both its sense-classes. Hence any In direct by Theorem 58 and its first corollary. and F' are direct, it follows is direct, and since
is
By Theorem 61
are
first
determined by
case
the doubly oriented lines analogous to I which the are all right-handed or all left-handed. In case C shall be called right-handed, and in the second
all
446

algebraic criteria by E. Study referred to in
162.
[CHAP.IX
The
article
in the exercises below are taken from the
See also F. Klein, Auto-
Vol. II, graphierte Voiiesungen iiber nicht-Euclidische Geometrie, Chap. I, Gottingen, 1890.
EXERCISES
1.
Classify parabolic congruences
(107, Vol.
I) as right-handed
and
left-
handed.
2.
For two
lines
f p and p
let
O X) =
is
PuP'o,
+ PuPu + PuPu + fl4/12 + PuP'll + Pa*Plt>

.
where p^. are the Plucker coordinates ( 109, Vol. I) of p, and p' those of Three lines p, p', p" are light-handed or left-handed according as
{
p'.
(Pt P')
(P''
P")
(P"' P)
positive or negative.
3. pair of conjugate imaginary lines of the second kind whose Pliicker coordinates are p^ and p^ respectively are right-handed or left-handed ac-
cording as
is positive
4.
The
(110,
or negative. linear line complex whose equation in Plucker coordinates is Vol. I)

fl
42
^42
is
right-handed or left-handed according as

1234 + 1342 + a !4 a 28 positive or negative. twist given by the parameters of 130 is right-handed 5.
is
if or /?
>0
122,
left-
and the coordinate system is right-handed. 6. The linear complex C determined by a twist according
is
to Ex. 7,
right-handed or left-handed according as the twist handed.
is
right-handed or
figure
*179. Elementary transformations of triads of lines. such that all the collineations which transform
Let
it
be a
into itself
are direct,
and
let
[F] be the
it
set of all figures
equivalent to
F under
direct transformations.
Prom
the fact that the group of all direct
collineations is continuous, familv of figures.
can be proved that [F]
is
a continuous
179,180]
ELEMENTARY TRANSFORMATION
62.
l y
447
THKORKM
plane
If
lf
l
z
are three skew lines,
copla,iar 'with
l.J.^
and such
that
are not separated by
and l^ is a, line the points in which and lz meet the l^ the lines / and then 3 l^ IJ1& can le
/
carried
to
l^J^
l>y
a direct collineation.
l^,
3,
Proof.
Let a be a line meeting
and
l^
(fig.
86) in points
A v A AA
y
,
respectively, which are all distinct.
Let a be the plane conl

a
taining
l
z
and the point
of intersection of
and
l^.
If
is
:
in a,
an elation with
as
as center,
points,
plane of
into
fixed
and
carrying A^ to A^ will carry l^

l z,
l
s
1
1}
ln ,
Z
4
respectively.
By Theorem
elation
i
66,
Cor.
4,
this
is
direct.
by and the point A in which a meets a for by hypothesis A v and the point in which lz meets the line BA are not separated by the lines l and Hence the homology with A as center and a as plane a 4
t
If A is A A and A
not in
a,
the points
are not separated
of fixed points
which
l
l}
carries
l
s
a
lt
to
l
z,
l^
is
direct
(Theorem
56).
This
homology
carries
2,
into
l^
respectively.
An elementary transformation of a triad of skew lines ljz ls may be defined as the operation of replacing one of them, say l & by a line which is coplanar with l and such that 13 and l^ do not sepa,
rate the points in
which
By Theorem 62
by a
their plane is met by ^ and lz an elementary transformation may be effected

.
direct collineation.
sequence of elementary transformations
therefore carries a right-handed triad into a right-handed triad a left-handed triad into a left-handed triad.
and
Conversely,
carried into
it
can be proved that any right-handed triad can be
any right-handed triad by a sequence of elementary transformations and that the two classes of lines determined by a
pair of
skew
is
that alx
lines al according to Ex. 2, 25, are the lines [x\ such right-handed and the lines \_y] such that aly is left-handed.
<s
TnPSP
TYrnnncnf.irm
tiro loff.
fn
f.Vift
TPflnPT
not
make use
if
of the preceding discussions of order in three-space.
DEFINITION.
perspective
1 ,
(A'B'C on a', fi',
doubly oriented lines are said to be douUy be given the notation (ABC, a@y) and a'/3V) respectively in such a way that A, B, C, a, /3, 7 are
they can
Two
and
C' respectively. Two doubly oriented liues 1Q , 7', A', are said to be similarly oriented if and only if there exists a
1
sequence of doubly oriented lines perspective with l^ ^ with l z ,
1
1}
l
l
z,
-,ln such that
is
doubly
I.
n _l
with
n)
and
with
Two
doubly oriented lines which are not similarly oriented are said
be oppositely oriented. From the form of this definition
it
to
follows immediately that
THEOREM
63.
If a doubly
,
.
oriented line
1
2
with a doubly oriented line 1 2 and is similarly oriented with l l^ &
is similarly oriented l^ with a doubly oriented line 1 3
THEOREM 64. If three doubly oriented lines m m^ z no two oj which are coplanar, are such that TO O is doubly perspective with 1 and m 1 with n then z is doubly perspective either with Q or with
,
,
the
Proof.
doubly oriented line obtained by changing both sense-classes on Let ABC be an ordered set of points of the sense-class of
points of
and
let
l^
1
2
be the three lines on A, B,
C respectively
which meet
and
The
sense-class of planes of Q contains either the ordered
triad of planes
or
l Q
0>
IJM Q ,
/
the ordered triad

,
m m
z
fl
Ijn
(fig.
I 2 9
ljti Q
In the
first
case
87) let
m = 7.
lî=a,l^ = ft,
In the second case In both cases

be the -t points
let
1'
,
Ijn^
= 7.
l
s
A,, 1' B,, 1 1' C,

ZjWj,
Lm
m^
respectively, a^ yQ
7j
the planes
^m^ ^m^ ^Wj
FIG. 87
l Q
^ m and 2 2 respectively, 72 the planes /O w 2 ^.j, Z2m 2 respectively. In the first case (ABC, txfiy) is doubly perspective with (A^C^ a i@iVj and this with (A 9 B C2> 2 /S 72 ). Since m (ABC, a0y), and Z 2 Q
respectively,
a,
B C
2,
the points
,
2>
az
>
/52 ,
is
is
w.,
doubly perspective with m^, m l =i doubly perspective with m 2 , n

(A.U I>
*i
iî7i)'
2,
an(^ s ^ uce
^/3272 ).
But
by construction
#2/3^7,,)
C2
tive
ie.
is doubly perspecwith (AI>C, afiy),

is
m^
In
doubly perspecw/
()
.
tive
with
the
second
case
(AB C, a/By) is doubly perspective
with
an(^
r
(A^B^
wi^1
Since
iî7i)
^ n i-s
m = (AHC,
TOJ
is
(â^2^2'
az i'iP^'
a/37),
and
FIG.
doubly perspective
with
m m =
o
,
fi\Y^)'i
an d since
wz^
is
doubly perspective with

,
m m
2,
(A zBz Cy
aSY 2 @
with
2 ).
But by construction (A n B z Cn
ccy/3)
;
(J.C7?,
i.e.
is
doubly perspective doubly perspective with the doubly

cc.f/ z /3 z )
is
oriented line obtained by changing both sense-classes of
THEOREM
(ACB, a
Proof.
2
65.
doubly
oriented
line
(ABC,
afii]
to
is
similarly
oriented with (ACB, ay/3)
and
oppositely oriented
(ABC,
aB
and
Let
Q,
is
on
/ 2
be three lines distinct from
AB
Let
1
Q,
and such
wi
1
that
2
and
a,
^ on
and
7, lz
on C and
/3.
and
^ 2
,
w&
.
be
two
lines distinct
from AS, each

l
of
which meets
let
l
l}
and
Let
,
A it B^, Cl be the points the planes I lf ljnf

k
lt
respectively and a2 A,, 72 the planes Ôm a ljn>t ljn> t Then by construction (fig. 88) and definition the oriented line (ABC,a@y) is doubly perspective with (A^B^, afyj, and this with (A B C and this with (ACB, <xyP). Hence 2 2 2 2 7 2 /3 2 ), (ABC, a/37) is similarly oriented with (ACB, ay0). By a change of
z>
?
î'
m V m
l
and ag Ijn^ Ijn^ respectively

;
^2
,
72
l^
respectively
,
A2
2>
C2
be the points
,
respectively.
notation
it
is
evident that (ABC, #7/3)
is
(ACB,
a/3j).
It remains, therefore, to prove that
with similarly oriented a/fy) is not
(ABC,
similarly oriented with
(ACB,
a/3y).
would If these two oriented lines were similarly oriented, there be a sequence of m, such that , doubly oriented lines 2 l}
m m m ,-,
450

a/3y)
[CHAP IX
m = (ASC,
line of the
and
mn = (ACB,
a/3y),
and such that each oriented
sequence would be doubly perspective with the next one be a doubly oriented line not coplanar with in the sequence. Let and doubly perspective with TO O let be -, n any of Q p
m
,
m m
,
the doubly oriented line obtained by changing both sense-classes
on
TO.
By Theorem 64
is
1
doubly perspective with TO or
TO.
By
a second application of this theorem TO or m, and by repeating this process
doubly perspective with n times we find that mn is

is
doubly perspective with

or
TO or TO.
But
this
means that
n is
(ABC,
(ACS,
66.
ayft).
THEOEEM
larly
There are two and only two classes of doubly oriented
lines such that
any two doubly oriented lines of the same class are simioriented and any two of different classes are oppositely oriented.
Proof.
and
it.
let
K be the class
is
Let (ABC)
<x/3<y)
be an arbitrary fixed doubly oriented line
of doubly oriented lines similarly oriented to
or
(ABC, a<y/3). AS, (ABC, a/3y)
This class contains (Theorem 65) (ACS, ayfi) but not (ACS, a@<y) If I is any line and TO any line not meeting I or
doubly perspective with one
of the
doubly oriented
lines determined
Hence
by any
by TO and this with one of those determined by I. contains two of the four doubly oriented lines determined
Let
line of space.
K'
similarly oriented with
(ACS,
afty).
be the class of doubly oriented lines It also contains two of the four
and doubly oriented lines determined by any line of space. cannot have a doubly oriented line in common, because this would imply that (ASC, a/rty) and (ACS, a/3y) were similarly oriented.
Hence every doubly
oriented line
is
either in
or in K'.
There can be no other pair of classes
of similarly oriented
doubly
oriented lines including all doubly oriented lines of space, because ! one class of such a pair would contain elements both of J"and of
would imply, by Theorem 63, that (ABC, aft<y) was similarly oriented with (ACS, ccfiy). From the construction which determines whether two doubly
this
and
oriented lines are similarly oriented or not,

lineation carries
it is
evident that any col-
any two doubly oriented
lines
which are
similarly
if it
carries
oriented line into
one into an oppositely oriented one, an oppositely oriented one.
it
carries every
doubly
Any
collineation
which
is
carries a
doubly oriented line into a simi-
larly oriented
carries a
to
and any collineation which doubly oriented line into an oppositely oriented one is said be opposite. This definition of direct and opposite eollineations is
one
said to be direct,
easily seen to
be equivalent to that in
of sense.
32.
*181.
in the
More general theory
The theory
of sense-classes
preceding pages can be extended to analytic transformations
by means of simple limiting considerations.

a transformation of a part of a
For example, consider
Euclidean plane
y
where both series are convergent for the point (0, 0). If the determinant
0/(0, 0)
all points in
0/(0, 0)
Sx
dy
0$
\&
dy
not zero, it can be shown that there is a region including (0, 0) which is transformed into a region including (& 00 J 00 ) in such a way
,
that all ordered point triads of a sense-class in the first region go into
ordered point triads of one sense-class in the second region; and if is in the same (a/, y'] plane as (x, y] the two sense-classes will be the
t
same
if
and only
if
J>
0.
By
a similar limiting process the notions of rightcurves, ruled

surfaces,
and left-handedfigures
ness can be extended to
and other
discussion of some of the cases which having analytic equations. 162. arise will be found in the article by Study referred to in
of theory of sense relations belongs essentially to geometry, although the domain to which it applies may 168 and 173, be extended by methods of the type used in
This
sort
differential
The theory
of sense
may, however, be extended in a
different
way
so as to apply to the
geometry of
Tl-i
all
continuous transforma-
tions instead of
merely
to the projective
geometry or to the geometry

QT>fl
flfi
Any
into itself transforms each sense-class

or into the other sense-class.
one-to-one reciprocal continuous transformation of a curve on the curve either into ifcselt
transformation of the
first
kind
is
called direct
and one
of the second
(
formation
is
is a deformation not a deformation.
kind opposite. A direct trans157), and an opposite transformation

consisting
of
Any
simple
closed
curve
points
in
or
on the
boundary of a 2-cell R ( 155) is the boundary of a unique 2-ceil which consists entirely of points of R. A 2-cell can he deformed into itself in such a way as to transform an arbitrary simple closed curve of the cell into an arbitrary simple closed curve of the cell. Any one-to-one reciprocal continuous transformation of a 2-cell and its boundary into themselves
is
deformation
;
if
and only
if
if
it
effects a deformation
on
tlie
boundary
i.e.
if
and only
it
transforms a sense-class on the
boundary into
itself.
If the sense-classes
are designated as positive
on one curve j^ of a 2-cell and its boundary and negative respectively, any sense-class
it
on any other curve jv is called positive or negative according as is the transform of the positive or of the negative sense-class on
by a deformation of
j^
into
are all simple closed curves
/ through intermediate positions which on the 2-cell and its boundary. By the
theorems above, this gives a unique determination of the positive and negative sense-classes on any curve of the given convex region.
A
is
curve associated with
its
positive sense-class is called a positively

its
oriented curve,
and a curve associated with
negative sense-class
called a negatively oriented curve.
Any
transformation of a 2-cell which
is
one-to-one, reciprocal,
and continuous either transforms

positively
all positively oriented curves into
oriented curves or transforms all positively oriented curves into negatively oriented curves. In the first case the transformation is said to be direct and in the second case to be opposite.
is
The transformation
a deformation
its
is
if
and only
if
it
is
direct.
its
2-cell associated
with
positively oriented curves or with

called an
negatively oriented curves
oriented
2-cell.
The
oriented 2-cells of a simple surface fall into
two
classes such
that any oriented 2-cell of one class can be carried
by a continuous
same
class,
two oriented
classes.
but not into any oriented 2-cell of the other class. The 2-cells determined by a given 2-cell are in different simple surface associated with one of these classes of
is
oriented 2-cells
said to be oriented.
2-cells of a
similar
theorem does not hold for the oriented

Instead
the theorem that every continuous one-to-one reciprocal transformation of a projective plane is a deformation. Consequently any oriented 2-cell can be carried into
projective plane.
we have
any other oriented 2-cell by a deformation. The oriented simple surfaces in a 3-cell and
into
its
boundary
class
fall
two
classes
such that any member

of
of
either
can be
deformed into any other member

of
the same class through a set
intermediate positions which are all oriented simple surfaces, but cannot be deformed in this way into any member of the other class.
3-cell
continuous one-to-one reciprocal correspondence which carries a and its boundary into themselves either interchanges the two
classes of oriented simple surfaces or leaves
them
invariant.
In the
first
second case the transformation

case
it is
is
a deformation and in the
not.
3-cell associated
with one
of its classes of oriented
surfaces is called
an oriented
3-cell.
The oriented
3-cells of
a projective space
fall
into
two
classes
such that any member of one class can be carried by a continuous

deformation of the projective space into any member of the same continuous class but not into any member of the other class.
one-to-one reciprocal transformation of the projective space either transforms each class of oriented 3-cells into itself or into the other
class.
In the
first
case
it is
a deformation and in the second
it is
not.
projective space
associated with
one of
its
classes of oriented
3-cells is called
an oriented projective space.

are easily proved
The one-dimensional theorems outlined above

on the basis
of the discussion of the sense-classes
on a
line in
159
and 165. The two-dimensional ones, though more difficult, are consequences of known theorems of analysis situs. They involve, however,
such theorems as that of Jordan, that a simple closed curve separates a convex region into two regions and the theorems of this class do not belong ( 34, 39, 110) to projective geometry. The Jordan
;
theorem in the special case of a simple closed polygon does, however,

r\QI
r\-r\re
4-r\
rw/~*-i/-\/->-f-iTTr
rtAr\-rr\
f\4-
ntr
ot-*
-i
a vmrvtm/^
rol rvmr
"I
Q rt\
The three-dimensional
propositions
outlined
here have not
all
been proved as yet, but are (in form) one- and two-dimensional ones.
Let us note that an ordered triad
direct generalizations of the
of points as treated in
160
may
be
regarded as determining an oriented 2-cell. For the triangular region having the points as vertices is a 2-cell, and a sense-class is determined on its boundary
by the order
of the vertices.
This sense-class determines a sense-class on every

2-cell.
curve of the 2-cell and thus determines an oriented
In like manner an ordered tetrad of points as treated in 160 determines an oriented 3-cell. For the tetrahedral region having points A BCD as vertices is a 3-cell.
The triangular
region
BCD is a 2-cell which
does not contain
A, and is oriented in view of the order of the points on its boundary. This oriented 2-cell determines an orientation of the boundary of the 3-cell, and
thus of the
3-cell.
Likewise an ordered tetrad

2-cell, i.e.
A BCD
of a projective plane determines a
this 2-cell is oriented
that one of the triangular regions by the order of the points
BCD
BCD.
which contains .-1; and Similarly, an ordered

i.e.
pentad ABCDE of points in aprojective space determines an oriented 3-cell, that one of the tetrahedral regions BCDE which contains A, oriented according to the order of the points
BCDE.
of
182.
Broken
,
,
A A2
i}
a 'broken line joining A^ to n The called the vertices and the segments ', points joining them the edges of the broken line. If the vertices are all distinct and no edge contains a vertex or a point on another edge, the broken
to
-
a,
An A n _ l to A n A n are A^
-,
'
and polygons. DEFINITION. A set together with a set of n segments joining

lines
,
n
to
points
A A
, .
is called
line is said to be simple.

closed,
If
A^=A n
the broken line
is
said to be
otherwise
it is
is
said to
be open.
The
If
set of all points
on a
closed broken line
the vertices of a polygon are all distinct and no edge contains a vertex or a point on another edge, the polygon is said to be simple.
called a polygon.
broken line whose vertices are

1
A A
I}
g}
>.-,
are the segments joining
the broken line

-,
AÂ 2
-
A to A Z) A^ to A# A n and its edges are

,
A n and whose edges A n-l to An is called denoted by A A A A

,
.,
A n _Â n
respectively.
-
If
denoted by AÂ Z
A n _^\
the corresponding polygon the vertex A 1 is sometimes denoted
A l =A n
is
A n A 2 by A n +
,
by
l,
etc.
The following theorem is an obvious consequence of the definition. THEOREM 67. The polygon AÂ Z A n A^ is the same as A A
-
A A
tf.m.d.
A A
A A
T~f
is
n.tn.ti
v\n^mt
nf
/!,
/>
//7
.4
.j:
l/myi/
Jly
*.'Vlt,OVVVU,VV
VIW OWIIUU OOK
I//
JJUblWO
U-0
(//(/>
C/M-yti
yi yi
^/
M,
simple polyyon
AÂ a A.A
polygon
A A
n
is
l
such that
the
A A SA
J
are collinear
and A.^ J
in
the
this
]S
is
the
same as
the,
polygon AÂ Q
before
A nA
-which, all the
edges lut AÂ^ are

.
same as
and AÂ &
is
segment AÂ Z A,A

If
DEFINITION.
A, B,
are
an elementary transformation
any three points on a simple polygon, is the operation of replacing any one
of these points, say (J, by a point C' such that C and C' are joined by a segment consisting of points of the polygon and not containing
either of the other
two
points.
A class
consisting of all ordered triads

finite
each
of
which
is
transformable by a
number
of
elementary
transformations into a fixed triad
ABCis
called the sense-class
ABC
and
is
denoted by S(ABC).
There exists a one-to-one and reciprocal correspond-
THEOREM. 68.
line such that
ence "between the points
of any simple polygon and the points of any two triads of points on the polygon are in the same
with respect to the polygon if and only if the corresponding triads of points on the line are in the same sense-dass.
sense-class
Proof.
An _
and
Let the vertices of the polygon be denoted by Let BI} 2 let A be also denoted by A n >, 1
.
lt
A^,
Bn _^
be n
]et
arbitrary points of a line
in the order {B^,
z,
Bn _^}
,
and
Bn
which conr Let jB denote that segment ^-5i+1 tarns none of the other points B. Let the edge joining A { to A i+l correspond protectively to the segment /3{ in such a way that A i and
also
denote
J5
A i+l
are
projectivities
homologous with B and by which two sides

i
Bi+1
respectively.
(In general the
of the
polygon correspond to two
segments on the line will be

to B.
different.)
is
(i=
1,
.,
1),
there
and reciprocal correspondence F is such that each side of the polygon with its two ends corresponds with preservation of order relations to a segment of the line and its two ends. Let P P Ps P^ denote points of the polygon and L I} L2 L 3 L4 I} 2
If we also let A i correspond evidently determined a one-to-one between the polygon and the line
which
the points of
to
which they
correspondence
respectively correspond under F. The an P is so denned that if a t hy f^P^ goes into
PP
1
theorem follows at once irom cue coronary
<.;r
ineorem
&&.
the
COROLLARY
" broken line " interval "
1.
The theorem above remains true if

"
be
words
with distinct ends

line.
substituted
for polygon, and
for
The
definitions
of
points on a line
may now
21 for the separation and order given in be applied word for word to the points
in
OD a simple polygon, and in view of the correspondence established Theorem 68, the theorems about order relations on a line may be
applied without change to polygons. By comparison with the proof of Theorem 15
we
obtain immediately
COROLLARY COROLLARY
2. 3.
A simple polygon A simple broken
is
a simple closed curve. line joining two distinct points
A, A n
is
a simple curve joining A^ and

relations
An
The order
Suppose we
on a broken line which
is
not simple
of a
may
be
studied by the
AA
'
'
'
An
l
simple device. associate an integer with each point of a broken line "With A l and every point of the segment as follows
:
method given above with the aid
joining
to
the
to
ment joining A 2 the number n.

the
number 1 with A 2 and every point of the segA 8 the number 2 and so on, and, finally, with A n
; ;
DEFINITION. The object formed by a point of the broken line and number associated with it by the above process shall be called
;
a numbered point
ment,
line, plane, etc,
and the numbered point is said to be on any segwhich the point is on. If A, B, C are any three
numbered points on a polygon, an elementary transformation is the operation of replacing any one of these numbered points, say C by a point C' such that C and C' are joined by a segment of numbered points all having the same number. A class consisting of all ordered
t
triads of
numbered points each
of
which
is
transformable by a
finite
i?
sequence of elementary transformations into a fixed triad called the sense-class ABC and is denoted by S(ABC).
ABC
By the proof given for Theorem 68 we now have THEOREM 69. There exists a one-to-one and reciprocal correspondence between the numbered points of any broken line and the points
of any interval suck that two triads of numbered points are in the same sense-class if and only if the corresponding triads of points on the, interval are in the same sense-class.
We
tions
of
are therefore justified in applying the theorems and definiabout order relations on an interval to the numbered points a broken line.
EXERCISE
points of a region can be joined by a broken, line consisting entirely of points of the region.
*Any two
183.
theorem on simple polygons.

set of
gon was defined as the

points
and linear segments.
In the last section a polypoints contained in a sequence of This is the most usual definition and
With a view to generalizing so as to obtain the theory of polyhedra in spaces of three and more dimendoubtless the most natural.
sions,
however,
we
shall find
it
more convenient
to use the property
of a simple
polygon stated in the following theorem.*

70.
THEOREM
set
of points [P]
is
if the following conditions are satisfied

distinct points, called vertices,
a simple polygon if and only : (1) [P] consists of a set of
and of distinct segments, called edges, such that the ends of each edge are vertices and each vertex is an end of an even number of edges; (2) if any points of [P] are omitted,
the
remaining subset of [P] does not have the property

It is
(1).
Proof.
satisfies
obvious that a simple polygon, as defined in 182, Conditions (1) and (2), because no edge has a point in comother edge or vertex and each vertex
is
mon with any

exactly
an end
of
two
edges.
Let us
now
consider a set of points [P] satisfying (1) and
(2).
If
two
more edges have a point in common, this point divides each Hence the point may be regarded as a edge into two segments. vertex at which an even number of edges meet. In like manner, if an edge contains a vertex the two segments into which the edge
or
is
divided by the vertex
may
be regarded as edges.
Since there are

this
originally given only a finite
process determines a finite
number of vertices and edges, number of vertices and edges such
that
no edge contains a vertex or any point of another edge.

*This form of the definition of a polygon and a corresponding definition of a polyhedron are due to N. J. Lennes, American Journal of Mathematics,
Vol.
xxxin
(i9i:n. p. 37.
Now
let e
be any edge and

of
P
l
one
of its ends.
Since there are
an even number
as an end, there exists ansegments having P l be its other other distinct from e x let this be denoted by e f Let P 2 as an end, and so on. be a second segment having P end, and let z
; <?
By
this process
we
obtain a sequence of points
and segments
finite
number of vertices is finite, this process must lead by a number of steps to a point P n which coincides with one of the in the points previous points, say P The set of points included
Since the
{
.
and segments
*-i>
i+1>
**+!
"
'>
-*-P
satisfies
the definition of a simple polygon and has the property that

i
each _Z?(/=e,
+1
P _J
n
is
(1).
an end
of
two and only two

(2) it
e's.
Hence
all
it
satisfies
Condition
By
Condition
must include
points of the set [P].
set of -points satisfying Condition (1) of Theorem a finite nunibcr of simple polygons no two of which have any point in common which is not a vertex.
COROLLARY.
70
consists of
Proof.
Condition
In the proof of the second part of the theorem above, If Condition (2) (2) is not used before the last sentence.
be not satisfied, the set of points

e
remaining when the segments
i+v
'
'>
en
an d those
of
the points
t ,
P_
n
which
are not
ends of the remaining segments) are removed continues to satisfy Condition (1). For on removing two segments from an even numHence the process by wliich the ber, an even number remains.
,P n _ l , en was obtained may be repeated and simple polygon P., ei+1 , another simple polygon removed. Since the total number of edges is finite, this step can be repeated only a finite number of times.
-
184. Polygons in a plane.
In the next three sections we shall
prove that the polygons in a projective plane are of two kinds, a polygon of the first kind being such that all points not on it constitute
two
regions,
and a polygon
it
of
the second kind being such
constitute a single region. a triangular region is a polygon of the first kind,

line a
that all points not
on
The boundary of and a projective

t,li
polygon
of the
second kind.
In proving that the points not on
a nolvp-on constitute one or
two
mo-inns.
WR
shall
np.firl
fnllnwincr:
184]
POLYGONS
71.
459
THEOREM
in
a,
"plane,
is
point coplanar with but not on a polygon p in a triangular region of a containing no point of p.
Any
Let the polygon be denoted by An A^_ and the AÂ Z point by P. By an obvious construction (the details of which are left to the reader cf. 149) a triangular region TT may be found
Proof.
;
containing
and not containing

is
or
AÂ Z
In like manner a triangular region

contained in
.
or any point of the edge TZ may be constructed
which contains P,
point of the edge
T^,
and does not contain
or
any
all
AÂ &
By
repeating this construction
we
obtain a
sequence of triangular regions

tain
T^
2,
Tn
each contained in
the preceding ones, containing P,
and such that
Tk
does not cou-
AkJrV any point of the broken line AÂ^ and contains no point of the polygon AÂZ
Thus
Tn
contains
AnA v
p
is
COROLLARY. Any point of space not on a polygon hedral region containing no point of p.
Let the set of lines containing the edges
plane be denoted by be on the same line,
assigned that no
l
lf
in a tetra-
2,
n.
of a simple polygon in a Since more than one edge may
is less
According to Theorem 67
we can
than or equal to the number of edges. first suppose that the notation is so
collinear except
two edges having a common end are
in the case of a polygon of two sides (which is a projective line), for two collinear edges and their common end can be regarded as a single
edge. In the second place, according to the same theorem, we can introduce as a vertex any point in which an edge is met by one of l the lines l^, l z , n which does not contain it.
,
Under these conventions the polygon may be denoted by A tA 3 A mA v where each point A (i = 1, 2, m) is a point of intersection of two of the lines Z l and each edge is a segment joinn) l^ 2 ing two vertices and containing points of only one of the lines
{
, , ,
^>C
1
In like manner,
when two
or
more simple polygons are under conarrange that no two edges of the collinear, and then
sideration, let us denote the set of lines containing all their edges
by
1}
z,
.,
n.
We may
first
same, polygon
which have an end in common are
of
intersection of the
lines
l
lf
/
2
and
to
have no
which contains such a point
of intersection.
are thus led to study the points of intersection of a set of n coplanar lines and the segments of these lines which join the points
of intersection.
We
185.
2
Subdivision of a plane
,
î>
B all
in the
by lines. Consider a same plane TT. The number a

n n
(.
set of
lines
of their points
of intersection is subject to the condition

-= a = o
-=
~ 1)
.
'
the two extreme cases being the case where all n lines are concurrent and the case where no three are concurrent. According to 22, Chap. II, and the definition of boundary ( 150), the points of intersection
bound a number al
1
of linear
convex regions upon the
lines.
The number
is
subject to the condition
al
n (n
1),
the two extreme cases being the same as before.
THEOKEM 72. The points of a, plane which are not on any one a number a 2 of convex l of a finite set of lines l l} /, n fall into regions such that any segment joining two points of different regions
,
contains at least one point of

the inequality
1
l^,
2
The number
a:,
satisfies
n(n
1}
Proof.
The proof may be made by induction.
If
=1
the theorem
follows directly from the definition of a convex region. that it is true for n k 1. k, and prove it for n
We
l
lt
suppose
= +
We
are given k
+l
lines ^,
2,
t+r
The
lines
z,
-,
determine a number
of
k)
not less than k and not more than <L
-+
1,
convex regions. The line lk+1 meets the remaining k lines in at least one point and not more than k points. The remaining points
of
l
k+l
therefore form at least
one and at most k linear convex

all
regions, each of
which
is
the set of
points
3).
common
to
k+:
and
one of the planar convex regions (Theorem
By Theorem
8 each
convex region which contains points
of
k+1 is divided into
two con-
vex regions such that any segment joining two points
of different
it
follows that
COKOLLARY
determine,
1.
If n
n convex
lines of a plane pass through a point, they regions in the plane; if no three of them are
^
concurrent, they determine
-+
-
1 convex regions.
2>
Let us denote the aQ points of intersection of the lines l^
z,
by
or any one of
points determine
the a^ linear convex them by a upon the lines by

;
regions which these
a1
or
a-
...
a1
any one of them by a
1
;
and the a 2 planar convex regions by a2

a2
a2
or
any one of
them by a 2
2.
COROLLARY
line
If
the
lines
l
lf
1
2
are not concurrent,
any
z coplanar with and containing a point of an a has a segment The ends of this segment are on the of points in common with it.
2 "boundary of the a ,
and no
line,
other point of the line is on this boundary.
Proof.
1
2,
,
The given
n
which we
shall call
I,
meets the
lines
l
lt
in at least
two
points, and, as seen in the proof of the
these points as ends is
theorem, one and only one of the mutually exclusive segments having 2 Let composed entirely of points of the a
.
a denote
Theorem
that
l
l
this segment. 10.
Its
lj
Let
l{
and
ends are boundary points be lines of the set l : lz

,
of the
,
-,
ln
a 2 by such
2
cr and the other. All points of the lj from the points of the segment complementary to a by the lines 1. and l} Hence any point of the complementary segment is in a triangular region containing no point of the a* and is
contains one end of
are separated
therefore not a
boundary point
of the
aa
This argument carries with
it
the proof of
COROLLARY
not in the a
2
3.
Any
interval joining a point of
an a 2
.
to
a point
contains a point of the boundary of the a 2

,
l THEOREM 73. If the lines l : 1 2 n are not all concurrent, the 2 boundary of each a is a simple polygon whose vertices are a's and
,
,
whose edges are a^s.

Proof.
The theorem
is
a direct consequence of
;
151 in case
i.e.
nS.
it
Let us prove the general theorem by induction

true for
we assume
n=k
and prove
it
for
= k + 1.
1}
Then any one

l
Let the notation be so assigned that l^ 12 , 1 3 are not concurrent. of the convex regions, say R, determined by 1 / -,
2,
,
,
contained in a triangular region determined by l^ 12 and l & because no two points of R are separated by any two of the lines 1 ,
k is
l^
Let
any
of its
be a line containing no point of this triangular region nor vertices. The segments A ( Aj etc. referred to below do not
contain any point of m. If lk+l contains a point of R, it contains, by Cor. 2, above, a segment of points of R such that the ends of this segment are on the
boundary
of
R.
By Theorem
67, the ends of this
segment may be
R.
taken as vertices of the polygon p which by hypothesis bounds Thus we may denote this polygon by AÂ Z A AjA i} where
{
and
are the points in
which
k+l meets the polygon.
There are just two simple polygons which are composed of the and of sides and vertices of p. For any such polygon segment
A^
it
which contains AÂ contains must contain A Z A S and therefore

t
AA
Z
a
or
i}
A^.;
,
if
it
contains
AA^
it
it
AA
A^^,
l
and since
;
con-
tains
AA
{
must be the polygon
AAA
{
AAj
it
since it
must contain AjA5 _^ and therefore contains A+A. *1 it must be A 1AJA, J

,
AA^ AJ _ Aj _
S 2,
if
,
it
contains
t,
A i+l A
and
Neither of the lines
AA A
2
k+l
i
and
A *1 A,.
AA
{
except
A A^
which
points of this polygon except A if A 1 one of the two regions, which we shall call R'
l
meets any edge of the polygon is contained in l k+r Hence all and those on the edge A A are in
t
and
R' f ,
bounded by
k+l
and
TO.
In
,
like
AA^ except A { A l regions R' and R".

The
points of
A. A
in
the polygon AÂjAj^ and those on the edge AA^ are in one of the two
all points of
manner
R on any
nnp
-nnint. -Frirm
line coplanar with R a oarrmonf rr lC\.r\v
and meeting
oK,vtr\
the
nrViinVi
by lk+l and m. ButP and Q are boundary points of R above. Hence the boundary of R has points in both of the regions R' and R" bounded by l k+l and m. By the paragraph above,
separated
by Cor.
2,
the points of the
boundary
of
in the one region, say R',
must
i
"be
>
the points, exclusive of the interval
AÂ^
of the
polygon AÂ Z
A A^
and those in the other, R", must be the points, exclusive of the interval
AjAy
of the
Let R X and
A A^. polygon AÂj R 2 be the two convex regions formed by the points of
(
.
R not on lk+1 Since these two regions are separated by lk+l and m, we may assume that R I is in R' and R 2 in R". Every boundary point of Rj which is not a point of l k+1 is in R'. For if B is a point
not on
R X it is not on m, by construction, and if it is can be enclosed in a triangular region containing no point of lk+1 or m. Such a triangular region must contain points of R and hence can contain no point of R", since any segment joining 1
of
the boundary of
^.
+1
it
a point of R' to a point of
R" contains a point
of lk+1 or of
m.
Hence
B
in
is
In like manner any boundary point of R 2 not on Zfc+1 is R". But by Theorem 10 every point J5 of the boundary of R is
in R'.
.
on the boundary of R I or R 2 Hence the boundary of R t contains all and by Theorem 10 it contains points of the boundary of R in R'
;
no other points not on lk+r Hence it is the polygon AÂ Z In like manner the polygon is the boundary of AÂ5 Hence the boundaries of the two planar convex regions into any one of the planar convex regions determined by l^ ls
A Ar
{
A^
Ra which
.
>,
is
separated
by
k+1 are
simple polygons.
l
lf
The other planar convex

identical
regions
determined
1 2,
by
,
z,
k+l
are
with regions
determined by l^
k.
COROLLABY
COROLLARY
there is
z Each x is on the boundaries of two and only two a 's. l In case all the lines lv 12 -, n are concurrent, only one a, the common point of the lines ; there are n aPs,
1.
2.
each consisting of all points except a

there are
of one of the
its
lines
lt ;
and
n a 2 's, each having a pair of the lines as THEOKEM 74. The numbers <XQ a^, a z satisfy the
,
boundary^
relation
Proof.
The theorem
"We shall make the proof by mathematical induction. = 1, tf, = 2, # a 2. is obvious if n = 2, for in this case
Let us
for
now assume
l^
l
it
to be true for
n = Tc and prove
set
that
it
follows
n = k-{-l. The lines
z,
',
determine a
of
a' points,
a[
linear
convex regions, and
a[ planar
convex regions subject
to the rela-
a' 1. The line meets a number, say r, of the tion a[ z a[ planar convex regions and separates each of these into two planar is increased to a(+r. The number of convex regions. Hence a' z
is increased by r for the number convex regions on lk+1 and also by a number* s equal to the l number of linear convex regions of the lines l^, l z k which are
one-dimensional convex regions
of
met by lk+v The number of points of intersection of l : lz ., lk+l l also exceeds a' by s. Hence for l i} lz k+1 the numbers <XQ a is az are a' a[+r + s, a'^ + r. Hence ao - a v + a,= (a[ + s) 9 + s,
. . , , ,
186.
The modular equations and matrices.
The
relations
among
the points, linear convex regions, and planar convex regions may be described by means of two matrices of which those given in
we
151 for the triangle are special shall denote by H^ is an array

associated with
of the ith
cases.
of
The first matrix, which #o rows and a^ columns, each
row being
element
row and
an a and each column with an a\ The j'th column is 1 or according as a? is

matrix,
'
or is not an
end
of a}. J
The second
has
2
a:
1
rows and a2
columns associated respectively with the a 1 a and of the i'th row and /th column is 1 or according
on the boundary
of
of
's.
The element
is
as a/
or
is
not
of a/.
1
Since every segment a has two and only two ends, each column contains just two 1's; and since each a 1 is on the 1 boundary
two and only two tains just two 1's.

For each
field
's
(Theorem
73, Cor. 1),
each row
of
con-
of the
a1
'
a let
us introduce a variable which can take

1,
on only the values
and
these being regarded as marks of the

2.
obtained by reducing modulo

aJ , 8
.,
We
denote these variables
akt respectively. There are 2*1 sets of values which can be given to the symbol t (^, #2 , aj aj ).
by
ojj,
*The number
s is less
than r
if
k+ i contains points of intersection of lv
2,
the -variables are zero, these symbols constitute the points of a finite projective space of 1 dimensions in which there are three o^
all
; ktExcluding the one in which
of these symbols (x x -, x a v z ^ corresponds to a way each segment a of the original n lines with a or a 1, the segment / being labeled with the value of xv We shall
Every one
of labeling
regard the
symbol
as
the
notation
for the
set
,
of
,
edges labeled
(y lt y^ the addition
with
>
1's.
ya )
By the sum of two symbols we shall mean (.^ +y 1 \+yz

1
,
(x l
-
xz
xa^) and
-,
a^+y^),
being performed modulo 2. According to our convention the sum 1 represents the set of a 's which are in either of the sets represented
by-(;
x
1(
x
lti
and
(y l}
2/ 2 >
y ai ) but. not in both.
tition of these considerations it follows that the

of
sum
of
By a repen symbols
xai ) for sets of edges is the symbol for a set edges each of which is in an odd number of the n sets of edges. In the sequel we shall say that a polygon p is the sum, modulo 2, of a set of polygons p lt p z p n if it is represented by a
, ,
the form (x^ x z
of
symbol
'
(x^,
.
xz
xa^) which
is
the
sum
is
-,
p
2
>
tt
Let us
>
now
inquire what
of the symbols for p pz v the condition on a symbol

,
(%!>
#a t ) that it shall represent a polygon
every vertex of a polygon there meet two and only two edges. lJ Hence, if we add all the x's that correspond to the a s meeting in 2. This modulo this sum must be zero, equations, gives any point,
At
one for each a, of the form

(4)
(>P>
a,
'
,H
fc
(mod. 2)
a k being the edges which meet at a given vertex), which must be satisfied by the symbol for any polygon. Obviously the For example, in matrix of the coefficients of these equations is r
a %>
'>
the case of the triangle these equations are
(cf.
151)
(5)
38 +B4 +a;6 -M6 =0, x l + xz + x 6 + x6 = 0,

* 1 +02 +
3
(mod. 2)
+tf4 =(X
(4)
We
column
shall denote the set of equations

of
by (H a).
Since each
,
columns
H gives of H are
2 2
2 the notation for a polygon bounding an a the For example, solutions of the equations (H^.
the columns of the matrix
in
151 are solutions
of
(H^.
solution whatever of these equations corresponds to a labell ing of the a 's with O's and 1's in such a way that there are an even number of 1's on the a l 's meeting at each a. Hence, by the
Any
one or more simple polygons.
Hence every solution of
the equations
(H
a simple polygon or a set of simple polygons. Since each column of the matrix H contains exactly two 1's, any
)
represents
one
equations is obtained by adding all the rest. Since the and 1, any linear combination of the only marks of our held are equations (HJ would be merely the sum of a subset of these equaof the
tions.
to the equations in the subset.
Consider such a subset and the points a which correspond 1 Every a joining two points of the
is represented in two equations, and the corresponding variable disappears in the sum. There remain in this sum the variables l corresponding to the a 's joining the points of the subset to the remaining points of the figure. These cannot all pass through the
subset
same point unless the subset consists of all points but one (since any two of the original n lines have, a point in common). Hence
while any one of the equations is linearly dependent * on all the 1 rest, it is not linearly dependent on any smaller subset. Hence #
of the equations
Since the
set of linearly
(H^ number
are linearly independent.

of variables is the a^,
number
all
of solutions in a
independent solutions on which
other solutions are

this
linearly dependent is o^
#
of
-h 1-
By Theorem 74
number
is
2
.t
1. polygons and sets of polygons is 2"" The simple polygons which bound the regions a2 are a set of soluSince each row of the tions, namely, the columns of the matrix g
Thus the
total
number
matrix
contains just two 1's, columns we obtain a solution of

2
it
follows that
in
if
we add
all
the
0.
(H^
which
all
the variables are
On the other hand, if we add any subset of sum will be a solution in which not all the
the columns of
the
variables are zero.
For
2 consider a segment joining an interior point of the region a corresponding to one of the columns in the subset to an interior point
B of a region a corresponding to one of the columns not in the subset; this segment may be chosen so as not to pass through a
2
point of intersection
of
two
of the
lines
l^
2,
ln .
Hence
it
contains a finite
on the polygons corresponding to the columns in the subset. The first one of these in the sense from B
of points * Since the and 1, the statement that only coefficients which can enter are one solution is linearly dependent on a set of others is equivalent to saying that it is a sum of a number of them. tin the modular space of a t 1 dimensions this means that the a 1 inde-
number
186,187]
MODULAR EQUATIONS
a.*
467
to
is
cm an
which,
is
on the boundary of a region in the subset
and a region not in the
subset.
The variable corresponding

a' a
2
to this
interval therefore appears in only one of the so does not drop out in the sum. Hence any
solutions of (4).
all
in the subset
1
and
of the boundaries
of the a 2 convex regions correspond to a ~ a: In other words, 2"

,
set
1,
of linearly independent
or one less
than half
of
the solutions of (HJ, are linearly dependent on the solutions are thus corresponding to the columns of H 2 The solutions of X
.
divided into two classes, those linearly dependent on the columns of H and those not so dependent.
Since each of the lines
l^
l
z
n is
a polygon,
it
a solution of the equations

tion
(H
a ),
but
it
does not correspond
corresponds to to a solu-
is linearly dependent on the columns of the matrix a corollary of the argument used in showing that the sum of any subset of the columns of H, is not a solution in which all the 2 variables are zero. For in that argument we showed that a certain
.
which
is
This
segment AB contains a point on the polygon represented by the sum of such a subset. The same argument applies to the complementary
segment.
Hence the
line
AB
with the polygon or polygons represented by the sum. of columns. Hence this sum cannot represent a line.
to
has two points, at least, in common of the subset
Thus, if we take the solution any one of the lines l^ lz

,
of
,
n,
the equations (Hj) corresponding 1 of the together with any a2

linearly independent set az independent solutions,
of
all
columns of the matrix

solutions.
2,
we have a
But since
this set contains
solutions are linearly dependent on this set.
187. Regions determined

can,
by
184, be regarded as one

ls
by a polygon. If p whose vertices

1
9
,
is
are a's and
any polygon it whose
edges are a
of a set of lines l^
l
,
Two
is
cases arise according as
is
or is not a
first
sum
of a subset of
represented by a symbol which the columns of the matrix 2
corresponds also to the sum of all the remain, 0). ing columns, because the sum of all the columns is (0, 0, It cannot correspond to a third set of columns, for the sum of
In the
case
468

of
[CHAP.IX
Let us denote the two sets
columns
-
of
>,
H
c
3
2,
whose sums are the
and supthat they represent the boundaries of pose the notation so assigned 2 2 a h and a*+i' "> a \ respectively. Let the points of 2 , i such points of the boundthe plane in a?, at,, a%, together with aries as are not points of p, be denoted by [P]. Let the set of the
symbol
for p,
by
lf
ck
and ck+v
respectively,
a* be denoted by [Q]. Clearly, s points analogously related to aj[ +1 the sets of points [P], [Q], and p are mutually exclusive and include
,
,
all points of
the plane.
Consider any point Q of the convex region af corresponding to c r to every It is connected by a segment consisting entirely of points in or on the boundary of a*. If k>l, c 1 has an edge in point
common with
at least
one
of c2 ,
ck ,
and the notation

.
may
be
c Then can Assigned so that c x has an edge in common with a of the common edge by a segment of be joined to any point l
P-points, and
2 the region a2 may be called
P
:
by another segment of P-points and its boundary. If Jc>2 there

c
g,
to every P-point of
is
a solution which
;
with an edge in common with c 1 or c 2 for if not, the solution ^-J-Cg would be one in which all the 1's correspond to the edges of p, and as no subset of the edges of p forms a polygon, c 1 + c g
would correspond to p itself. As before, every point of the region to P by a broken line of at a,l and its boundary can be joined Q
most three edges.
Since there
is
no subset
of c
x,
ck
whose sum
corresponds to TT, this process can be continued till we have any 2 of the convex regions a*, a 2 , ag and their boundaries point is on rr the process of conIf joined by a broken line & to
structing b
if
is
is such that all points of & except in [P] all points of & are in [P].
are in [P],
whereas
Hence any two points

sisting only of
of
boundary
of ,&%, any P can be joined to R of p by a broken line every point of which, except R, A precisely similar statement is true of [Q].
such points one of af, a%,
[P] can be joined by a broken line conand, since every point of p is on the
any point
is
in [P].
Consider
now any broken
line
joining a point
to a point Q.
Since
and
is
1
region,
either in
are within or on the boundary of the same convex [P] or on p. If A^ is in [P] the same con-
sideration
011
shows that
is
in [P] or
on
p.
If
none
of the A's
is in
were
p, this process
would lead
to the result that
Ak
[P], and
of
hence
Q would
is
also be in [P], contrary to hypothesis.
Hence one
on p, and hence any broken line joining a point P to a point Q contains a point on p. It now follows that [P] and [Q] are both regions. For we have seen that any point P can be joined to any other P by a broken line
the A's
is By Theorem 71 any point contained in a triangular region containing no points of p. This triangular region contains no Q, because if it did a segment joining it
consisting entirely of points of P.
to
[P]
satisfies
would, by the argument just made, contain a point of p. Hence the definition of a two-dimensional region given in 155.
A similar
bol
argument applies
75.
-,
to [Q].
Hence we have
THEOREM
(x^
Any
simple polygon
is
x2
xa^) which
the
sum of a
which corresponds to a symset of columns of H is 2
the 'boundary of two mutually exclusive regions which include all points of the plane not on p and are such that any two points of the same region can be joined by a broken line which is in the region.
Any
"broken line
joining a point of the one region
to
<x,
point of the
other region contains
a point of
point
the polygon.
COKOLLAEY
on
1.
Any
of
can
be joined to
any point not

which
to
by a broken line containing no other point of p. meets p in a single point COROLLARY 2. If a segment
ST
is
not
vertex of p,
S and T
are,
in different regions with respect
p.
Proof.
Let 8' and T' be two points in the order {SS'OT'T} and
,
l such that the segment S'OT' contains no point of l l} 12 n ex2 185, S' and T' are in two convex regions a which cept 0. By
,
have an edge in common. Since this edge is an edge of p, the columns of H 2 corresponding to these two a?'s must be one in the if S' set c c c k) and the other in the set c k+v c^. Hence, 1? z
,
and S are in [P], T' and
T
is
are in [Q],
and vice
versa.
THEOREM
bol
76.
.,
Any
)
simple polygon
not the
(x^ xz
aja
which
sum
which corresponds to a symof a set of columns of 8
is
the
not on p.
boundary of a region which includes all points of the plane Any two points not on p can be joined by a broken line
not meeting p.
be enclosed in Proof. By Theorem 71 any point not on p can a triangular region containing no point of p. Hence the theorem will be proved if we can show that any two points not on p are
of such points. If this were joined by a broken line consisting only be any point not on p and let [P] be the not so, we could let set of all points not on p which can be joined to by broken
lines not
meeting p.
have
to consist of a
As in the proof of Theorem 75, [P] would number of regions a 2 together with those points
,
and the boundary of [P] could consist only of points of p. But the boundary of [P] must consist of the polygon or polygons whose symbol is obtained by
of their
boundaries which were not on
adding the columns of H 2 corresponding to the a 's in [P]. By 183 no subset of the points of p can be a simple polygon. Hence p would be the the boundary of [P] and be expressible linearly in
2
a terms of the boundaries of a 's, contrary to hypothesis. l whose n corresponds to a Every polygon edges are on l{) lz xa ^ which either is or is not expressible linearly symbol (i x #2 ,
,
,
in terms of the
columns
of
Hence the
arbitrary simple polygon

lines
76.
p with which
^,
l
-
this section starts
and which determines the
z,
-,
n is described either in
Theorem 75
or in
Theorem
Hence we have
THEOREM 77. DEFINITION. The polygons of a plane a fall into two classes the individuals of which are called odd and even respectively. polygon of the first class is the boundary of a single region comprising all points of a not on the polygon. polygon of the second class is the boundary of each of two regions which contain all
points of a not on the polygon, have no point in common, and are such that any broken line joining a point of one region to a point of the other contains a point of the polygon.
The odd polygons are also called unicursal, and the even polygons are also called bounding. line is an example of an odd polygon, and the boundary of a triangular region is an example of an even
one.
The segments
a, fi,
7 as denned in
26 are the edges
of an
UJJJL>
AJN.D JiVJ^JN JfUJLlbrUJNS
471
polygons of which one is even and which are such that neither polygon has a vertex on the other have an even
78.
(or zero)
THEOREM
Two
number of points
Let p l be
aii
in common.
let p be any other polygon, z the points of intersection of the polygons be R En I} in the order with the theorem n} respect to p^. If n {/^ Z
Proof.
let
even polygon,
and
verified. If n were 1 the edge of p containing would have its 2 ends in different regions with respect to TT^ and hence the broken line composed of all p except the side containing R would have 2 1
is
to contain a point of
p^
li
r
contrary to hypothesis.
If
>
the inter-
val of
p n which has
and
J?
2
as
ends and contains no other points

(except for its ends) entirely to
is
a broken line
which belongs
one of the two regions [P] and [Q] determined by p^ and by Cor. 2, Theorem 75, the interval of p 2 similarly determined by z and K&
belongs entirely to the other of the two regions [P] and [Q]. Thus, if S Sn are a set of points of p 2 in the order {J2 1 *S'1 .fi 2 2 jR 8 v #2 _!#}, and $! is in [-P]> aU- the *S"s with odd subscripts are
,
in [P] Cor. 2,
and
all
the
75,
's
with even subscripts are in [Q].

in [Q] since
But by
is even.
Theorem
Sn
is
Sl
is
in [P].
Hence n
and containing no vertex of an even polygon meets it in an even (or zero] number of points. THEOREM 79. Two odd polygons such that neither has a vertex on
COEOLLARY.
line coplanar with
the other
meet in an odd number of points.
Proof.
the sides of
Let the polygons be p l and p z let the lines containing l and let ln be a line containing no -, l n _ v j? 1 be v
,
vertex of either polygon. According to the results stated at the end of the last section, p^ is expressible by addition, modulo 2, as the sum of lu and a number of boundaries of 2 's. The latter combine
into
edges of
on
even polygons, the edges of which are either Hence these even polygons have no vertices and contain 110 vertices of p z Hence by Theorem 78 they
number
of
pl
or of
n.
have an even
(or zero)
number
if
of points in
common with p,2 Thus

.
our theorem will follow

of points in
we can show
that
has an odd number
common
with
By
the argument just used
p z can be
expressed as the sum,
of points,
and
m
1.
meets
ln
in one point.
Hence p z meets
in
an
odd number
of points.
COROLLARY
in common.
Two odd
If
polygons always have at least one point

there exists an odd number of points and such
COROLLARY
polygon
2.
p
in
is
a simple polygon and

(or zero)
a,
meeting
an even
that neither polygon has
vertex on the other, then
is even.
Since the plane of a convex region always contains at least one line not having a point in common with the region, the last result
has the following special case, which, on account of we shall list as a theorem.
its
importance,
THEOREM
80.
Any
simple polygon lying entirely in a
convex
region is even.
To complete the theory of the subdivision of the plane by a polygon, there are needed a number of other theorems which can be handled by methods analogous to those already developed. They are stated below as exercises.
EXERCISES
convex region R, the points of R p fall into two regions such that any broken line joining a point of one region to a point of the other has a point on p. One of these regions, called the interior of the polygon, has the property that any ray (with respect to R) whose origin is a point of this region meets p in an odd number of points,
1. If
a simple polygon
lies entirely in a
not on
provided
it contains no vertex of p. The other region, called the exterior of the polygon with respect to R, ha.s the property that any ray whose origin is the one of points of this region meets p in an even (or zero) number of points,
provided
2.
it
contains
If
by p, according to Theorem
interior
one of the two regions determined no odd polygon and is called the of p. The other contains an infinity of odd polygons and is called the
is
no vertex of p. any even polygon in a plane
a,
77, contains
exterior of p.
3. If one line coplanar with and not containing a vertex of a simple polygon meets it in an odd (even or zero) number of points, every line not connumber taining a vertex and coplanar with it meets it in an odd (even or
zero)
boundary of a convex region consists of a finite number of linear segments, together with their ends, it is a simple polygon. 5. A simple polygon which is met by every line not containing & vertex in two or no points is the boundary of a convex region.
of points. 4. If the
187, rag]
POLYGONAL REGIONS
473
For any simple polygon A t A a A n A v there exists a set of n - 2 triangular regions such that (1) every point of the interior of the polygon is in
6.
or
on the boundary of one of the triangular regions, (2) every vertex of one of the. triangular regions is a vertex of the polygon, and (3) no two of the
triangular regions have a point in common.
*7. By use of convex regions and matrices analogous to E^ and H 2 prove Theorem 77 for any curve made up of analytic pieces (i.e. 1-cells which
,
satisfy analytic equations).
polygonal region
188. Polygonal regions and polyhedra. DEFINITION. planar is a two-dimensional region R for which there
exists a finite number of points and linear regions such that any interval joining a point of R to a point not in R, but coplanar with these meets one of or linear it, points regions. (three-dimensional) polyhedral region is a three-dimensional region R for which there
exists a finite
number
of points, linear regions,
and planar polygonal
regions such that any interval joining a point of R to a point not in R meets one of these points, linear regions, or planar polygonal regions.
Let R be a planar polygonal region and let Z x l z -, ln be a set with R and containing all the points and linear regions such that any interval joining a point in R to a point not
, ,
of lines coplauar
in
R meets one
of these points or one of these linear regions.
Let
us adopt the notation of 185. 2 If a point of one of the two-dimensional convex regions a is 2 in R, all points of the a are in R, for all such points are joined to
P by
intervals not meeting ^,

Z x,
2,
s
,
n.
l
Since any point not on

of R.
2
n
,
is
l
interior to a triangular
region containing no points of l^
1
2
n,
no such point can be a
boundary point
Let af,
seen, all
,
be the af k points of these
<x 's 2
's
are in R.
its
-,
which have points in R. As we have All points of their boundboundary

Jc)
;
aries are either in
R
(r
or on
I,
boundary
that
is
is,
of
an
a,
may be
for every point of the joined to a point of a^

}
to a point of
R,
by a segment
of
of points of R,
and hence
of one of
either a point of
or of its boundary.
Any
a
2
, ,
point
af.
of the
boundary
is
on the boundary
For any triangular region
containing
contains
474
not have
THEOKEMS ON SENSE AND SEPARATION
[CHAF.IX
B on
its
boundary, any
*
cfl
having a point in
common with
is
T' is one of
?,,?.
1
the boundary of one of a?
Hence every boundary point of R Hence the set of points a]k

,
.
on
of
.,
R
a
and
its
boundary
81.
is
identical
with the
set of all points of a?
and their boundaries.
In other words,
For any planar polygonal region R there is a finite R n such that the set of all set of convex polygonal regions R I} -, Rn and their boundaries is identical with R and points of R
THEOEEM
its
boundary.
As a consequence, any set of points which consists of planar polygonal regions and their boundaries can be described as a set of
points in a set of convex polygonal regions and their boundaries. Therefore no generality is lost in the following definition of a polyhedron by stating it in terms of convex polygonal regions.
DEFINITION.
isfies
set of points [f] is called a polyhedron if it satthe following conditions and contains no subset which satisfies
a, a, ., & segand convex planar polygonal regions a\, a\, 1 als such that each a is bounded by two a's and each a? by a simple polygon whose vertices are a 's and whose edges are a 1 's 1 2 no a or a contains an a and no two of the a l5 s or & 2 's have a
o
,
them: [P]
. .
consists of a set of distinct points

'
ments
,
a\, a\,
; a^,
point in
of
a a 's.
edges,
common; each a is on the boundary of an even number The points a are called the vertices, the segments a 1 the and the planar regions a 2 the faces of the polyhedron.
Just as any point of a polygon can be regarded as a vertex, so
any point of an edge of a polyhedron can be regarded as a vertex, and any segment contained in a face and joining two of its vertices can be regarded as an edge.
The
relations
among the
In the
vertices, edges,
and faces
of a
2
can be described by means of matrices

those of
186.
first
and
polyhedron analogous to
matrix,
the element
77^
is
or 1 according as a?
is
not or
is
an end of
aj.
In the second matrix,
w "
__/ " a\
discussion of these matrices just as that of the projective plane (a special polyhedron) has been derived in the sections above.
Thus
edges
,
of
-
the polygons which can be formed from the vertices and the polyhedron are denoted by symbols of the form
,
(x lt xz modulo
2, of
polygons. and is the same no matter

faces, edges,
186. They are all expressible as sums, the boundaries of the faces together with 1 other The number is called the connectivity of the polyhedron
xai ) as in
and
vertices.
It is determined
how the polyhedron is subdivided into by the following relation
EXERCISES
polygonal region can be regarded as composed of a finite set of triangular regions together with portions of their boundaries, no two of the triangular regions having a point in. common.
1.
Any
2. If R is a polygonal region, every broken line joining a point of R to a point not in R has a point on the boundary. 3. For any three-dimensional polyhedral region R there is a finite set of
R n such that the set of all points of Rj, R 2 , -, polyhedral regions R I( R 2 R n and their boundaries is identical with R and its boundary. R I} R 2
, ,
Rn be tetrahedral regions. 4. If a polyhedron is the boundary of a convex region, each edge of the polyhedron is on the boundaries of two and only two of its faces.
,
may be
so chosen as all to
189. Subdivision of space by planes.

generalize at once into the following.
tion) are left to the reader.
The theorems of 185 The proofs (with one excep-
THEOREM
82.
finite set of planes ir^ 7T8 ,
The points of space which are not upon any one of a 7rn fall into a finite number a of convex s
,
regions such that
contains at least
,7,
any segment joining two points of different regions 7rn The number a satisfies one point of TT^ TTO
3
the inequality
7-,
^ ^n(n-l)(n~2) = n^cc 7p
we
-,
-f-
n.
As
in
185,
indicate the aQ points of intersection of
planes
or any one of them by a by these points upon the
the a x linear convex regions determined
lines of intersection,
by
1 the a z planar convex regions determined or any one of them by a by the lines of intersection upon the planes, by
;
or
and the any one of them by a mined by the planes, by u 8 * l> ("y,,
;
spatial
&
convex regions
deter-
',
U^,
or
any one
of
them by a 3
ir n are not all coaxial, the THEOREM 83. If the planes ir^, 7T 2 z boundary of each a* is composed of a finite number of a 's and of z z l those a 's and a's which bound the a 's in question. Each a is upon
,
boundary of two and only two a*'s. = 0, a^ COROLLARY 1. If the planes TT,, TT Z> -, ir n are coaxial, a Q 2 3 and the boundary of each a is composed of two a 's together with
the
-
0,
the
common
a point
line
of the planes.
2.
COROLLARY
P, Q.
If the planes are not
all concurrent,
any
line through
I of one
& of the regions a meets
the
boundary in two points

a*,
The segment PIQ consists entirely of points of the 8 complementary segment entirely of points not in the a 8 THEOREM 84. If an a 1 is on the boundary of an a
.
and
the
it is
on the
boundaries of two and only two a 's of the boundary of the a*. Any 8 plane section of an a is a two-dimensional convex region bounded by
<t
simple polygon which
is a plane section of the boundary of the a*. a COROLLARY. The boundary of each a is a polyhedron. THEOREM 85. The numbers aQ a^, a z a z are subject to the relation
, ,
Proof.
= 0,
for
7rn+r
The proof is made by induction. In the case of two planes, = 0, az = 2, a s = 2. Assuming that the theorem is true n planes, let us see what is the effect of introducing a plane
tfj
This plane
is
divided by the other planes into a number of
8 convex two-dimensional regions equal to the number of 's in which it has points but it divides each of these 8 's into two & 8> s.
;
Hence the adjunction of these new a z 'a and 8's increases az and txs by equal amounts. The plane vrn+1 according to Theorem 8, Cor. 1, a divides hi two each a which it meets but it has a new a 1 in com, ;
each such region. Here, therefore, a and ai are increased z 1 by equal amounts. The plane ?rn+1 divides in two each a which it
mon with
on
tines of intersection of
(}
number ly which a l
,
'
a for
pairs of the planes TT^ the planes 7r TT^, 2

,
.
7r
,
2,
7rn _ 1 is
the
irn
exceeds
a^
ao
-, rrn _ 1 for the planes TT^ 7T2 1 Proof. New a a are produced by the introduction of
vrn
;
in two
so, this
ways
a
1
(1)
TT,,
may meet an
ir^,
is
separated into two

2
a 1 of TT^ a 1 a and a
'
7r
g
7rn _ t
in a point
;
if
meet an a
of
7r
ir n
^
ir n
in
new a is introduced (2) irn may 1 The only new a's proa new a
.
duced by the introduction of (2) accounts for the increase

190.
are accounted for
of
a,
a1
under a Q as stated above.

,
(1).
Hence
the
The matrices
lt
and
8.
The
relations
among
convex regions determined by n planes which are not coaxial

be described by means of three matrices, which and EL In the first matrix,
may
,
we
shall call
H^ H
H^foJ),
*
= 1,
2,
y = 1,
of aj.
2,
or is not
an end
= aj and rj^ 1 or In the second matrix,

,
according as a\
is
H = W),
2
= 1,
is
or
y = 1, 2, not on the boundary

2,
.
-,
oTjj
-,
and
of #?.
I or according as a] ij} In the third matrix,

),
is
H.
*
=
;
= 1,
is
2,
.,
a2
y = l,
2,
or
not on the boundary of
aj.
and ^- = 1 or according as a\ is Examples of these three matrices are

8
152 to describe the tetrahedron. It will be noted that H has two 1's in each column, and H g two 1's in each row. 1 Corresponding to the matrix H a there is a set of aQ linear
those given in
,
equations (modulo 2)
.7
^
=1
or 1, be taken xa ^, where the a^'s are Let the symbol (# x^ ls 1 and not containing to represent a set of a s containing al if xk 1 it if xk = 0. Just as in 186, this set of a a will be the edges of
, ,
=
'
a polygon or set of polygons
if
and only
sets of
if
(#
lS
xa ^
is
solution of (Hj). Just as in 186, the
sum
of
two
polygons (modulo
2) will be
478

of
f r
[CHAP. IX
taken to be the set

(*!
2>
,)
polygons represented by the sum of the symbols tne fcwo sets ^ polygons. The sum, modulo 2, of two sets of polygons p 1 and p.z is therefore the set of polygons whose
both p and p z edges appear either in p l or in p z but not in 1 of the equations (H ) are linearly By the reasoning in 186, a
.
on these. The independent, and the other one is linearly dependent a5 for the boundaries are the symbols (ajQ o^, columns of t) 2
of
the a2 's and hence are solutions of (H t ).
Corresponding to the matrix

2,
there
is
a set of
al
linear
(H 2)
Let the symbol (x lt % 2
to represent a set of
it if
,
fW
,
(i
= 1,2,
or
1,
-,,).
xa ^, where the xh's are
be taken
xk
set of
a 2 's
= 1 and not con taming containing a\ if xk If this symbol is a solution of (H 2 ), it represents a 0. 1 such that each a is on the boundaries of an even number
2
's
(or zero) of
them
i.e.
it
represents the faces of a polyhedron or a
set of polyhedra.
The columns of H g represent the boundaries of the a 8 's. By Theorem 84 any a of the boundary of an a* is on the boundaries 2 of two and only two 's of this boundary. Hence the columns of
1
are solutions of
(H 2 ).
Corresponding to the matrix

8,
there
is
set
of
linear
Let the symbol
1,
xa^ where the xk 's are

}
or
1,
be taken
to represent a set of a
it
if
xk a8 's
set of
s 1 and not containing containing a k if xk 0. If this symbol is a solution of (H ), it represents a 8 2 of which there is an even mimber on each a It is easily
's
seen that the only such set of a 's is the set of all a s 's in space. Hence the only solutions of (H s ) are (0, 0, 0) and (1, 1). Hence there are a 3 1 linearly independent equations in (H ) on
,
!,,
8
which
the rest are linearly dependent. Let the ranks* of the matrices ,
all
,
H H H
be r
r respectively.
By what
has been seen above
The discussion
in the next section will establish that
V=V a
191. The rank of
o-
Let us
now suppose
that
ir^,
?r
2
.,
7r n
are
not all on the same point and that the notation 7r ?T 7r are the faces of a tetrahedron. 2 TJ-J, 8 4
,
,
is so
assigned that
inspection of
By
the matrices given in

tf
152,
tf
3
it is
= 4,
a 1==12, a g
= 16,
clear that for the case w,==4,
=8, and
,
r2
=8
,
(a set
pendent columns of H 2 upon which the rest depend linearly is the set of columns corresponding to T U TM T U TM r22 r28 Tjfl and T M ).
,
, , ,
of linearly inde-
The number of solutions of (H a ) in a linearly independent set upon which all the other solutions depend is o^ # 4- 1 = 9. Hence one solution which does not depend linearly upon the columns of H
together with a set of eight linearly independent columns of constitute a set of linearly independent solutions of (H ) upon which a all the others depend linearly. Any solution representing a protective
,
line, e.g. (1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0), will
In case
w>4,
the columns of
fall into
;
serve this purpose. three classes: (1) those
(2) those representing representing the boundaries of a 's in TTH 2 1 the boundaries of # 's which are not in 7rn but have an a in 7rn ;
and
(3)
TTB .
those representing the boundaries of

of Class (1) is expressible as a
2
.
's
which have no
a 1 in
Any column
Classes (2) and (3). the boundary of an
sum of columns of For the a 2 whose boundary it represents is on 8 a whose boundary has no other a2 in common
1 Since each a on the boundary of an
with
?r n
(cf.
150).
as
is
2 on the boundary of two and only two a 's of the boundary of the a s (Theorem 84), it follows that the given column is the sum of 2 the columns which represent the boundaries of the other a 's on
the boundary of the a
s
.
These columns are
all of
Classes (2) or
(3).
Each a 1 which
TTj,
is
7r
.,
7r n
_1
not on a line of intersection of two of the planes 2 is the linear segment in which one of the 's
.
determined by TT^
7r
2,
7r
_1
is
1 corresponding to this a contains
met by 7rn Hence the row of H 2 just two 1's in columns of Class
.
the boundary of one of the Moreover, the columns of

this sort, since
's
determined by
TT^
7r
2,
.,
irn _ v
of
every a
ir^
of irn is
2, .,
Class (2) form a set of pairs of either on a line of intersection of

or is
two
of the planes
7r
7r B _j
an edge
of
two and only
2 two a 's not in
?r n .
No
one of such a pair of columns
of
can enter into a linear
relation
among
other does.
a set of columns of Classes (2) and (3) unless the For this column would be the only column of the set
1 containing a 1 in the row corresponding to the a common to the 2 boundaries of the a 's represented by the two columns, and hence the sum of columns could not reduce to (0, 0, , 0).
Let
H
of
and the sums
be the matrix consisting of the columns of Class (3) of of the pairs of columns of Class (2) discussed in
the last two paragraphs. According to the last paragraph the rank is less than the rank of the number of these pairs of 2 2 by
columns; and by the corollary of Theorem 85 this number is the # for TT^ 7r 2 difference between the values of 7r n and for o^
,
T!
"V
'
'
7rB _ r
The columns
mined by
TTj,
,
of
2,
are the symbols in terms of the a

7rn
lJ
deter-
7T
.,
for the boundaries of the
a 2 'a determined
by
TT^
7T
2,
ir n
,r
Hence any two rows
of this matrix
which
cor,
l 1 respond to a pair of a 'a into which an a determined by TT^ 7r2 7r _ 1 is separated by 7rn must be identical; and if one of each
,
rt
7rn _ r
such pair of rows is omitted, H 2 reduces to the H for 2 ir^ Hence H 2 has the same rank as the for w2
TT
.,
Z,
TT^
2,
7r
n_x
Since the difference in the ranks of
for
TT^
7r
ir n
and aQ
of
is the
7r
2,
same
7r n
as the difference
between the values

irn _ v it
of
al
for
Wj,
and
a
for
7r
x,
7r
2>
follows that the introduc-
tion of
7rn
increases the rank of

.
increases o^
Since a1
=r
for
by the same amount that it n = 4, the same relation holds

which are not
all
for all values of n.
Hence we have
set
THEOREM
concurrent,
86.
For a
of planes
TT
7r
2,
irn
*i-*o='V
By Theorem 85
this relation is equivalent to
*-*= rf
192. Polygons in space.
THEOREM
87.
The symbol
fa,
x2
.
. ,
.,
for a
line, is
not linearly dependent on the columns of
xa)
.Let TT
oe
The boundary
and
is
of
TT
if
met by
any plane not containing any or the points a. 187 any a? is an even polygon in the sense of in two points or none, the two points being on
existent.
is
different edges,
gons
p p
lt
each of which
The sum, modulo 2, of two sets met by TT in an even number

is
of poly-
(regardTT
ing zero as even) of points
a set of polygons
TT
p met by
in an
even number
2
7c
of points
if
for if
meets
in 2 kl points and
in
2
,
points,
pz
IT
k s of these points are on edges common to p and must meet p in 2 & x + 2 & 2 2 & 8 points. Hence any polygon
and
which is a sum of the boundaries of the o, 2 's is met by TT in an even number of points i.e. any polygon represented by a symbol sc a (x xz ) linearly dependent on the columns of H 2 is met by Since no line is met by TT in an TT in an even number of points.
;
>
l}
even number
of
of points, the
symbol representing
of
it
cannot be a
sum
any number
of
columns
THEOREM
of r
(!,
(i.e.
88.
All solutions of (H^ are linearly dependent on a
set
a
'
aQ } linearly independent columns of

*<0
and
the symbol
'
'>
for
one line
'
of H is <x 1. The X (H 1 ) is cc^ The number of a set on which all the rest are linearly independent solutions in a +l. Since the rank of H 2 is linearly dependent is therefore o^ aQ a and the columns of H 2 are solutions of (H^), there are a a
Proof.
It has
been shown thai the rank
number
of variables in the equations
of which are solutions of (Hj ; and linearly independent columns 2 since the solution of (H^ which represents a line is not linearly
dependent on these, the statement in the theorem, follows. In the proof of Theorem 87 it appeared that any polygon which
is 's is met by a a sum, modulo 2, of a set of polygons bounding plane which contains none of its vertices in an even number of is met by a plane not containing it in one point, points. Since a line
2
an argument
of the same type shows that any polygon which is a a sum, modulo 2, of a line and a number of polygons bounding a 's is met by a plane containing none of its vertices in an odd number
of points.
Thus we have, taking Theorem 88 into account:
89. DEFINITION. polygon which is the sum, modulo 2, of a number of polygons which "bound convex planar regions is met a vertex in an even number of points "by any plane not containing
THEOREM
points
ana
is called
an
oclcl
polygon. JLny polygon

tlie
.,
is eit/ier
oaa or even.
Suppose a polygon p is a set of convex regions a~,

a^ or on their boundaries
sum, modulo
2,
of the boundaries
of
t
a. The
set of points [P] in a*,
an argument analogous to that given in the proof of Theorem 75) to be a connected set. By an extension of the definition in 150 p may be said to be tlie
is
easily seen (by
boundary of [P]. From this point of view an even polygon is a bounding polygon and an odd polygon is not. 193. Odd and even polyaedra. It has been seen in 190 that
the solutions of
(H 2 )
converse
is
also true, as
represent polyhedra or sets of polyhedra. is obvious on reference to the definition

, ,
The
of
xan ) which reprea polyhedron. The sum of two symbols (xlf xz sent sets of polyhedra is a symbol representing a set of polyhedra. This
is
obvious either geometrically or from the algebraic considera-
tion that the
sum
of
two
for
solutions of
(H 2 )
is
a solution of
(H 2 ).
is
The
set of polyhedra
represented by the symbol which

sets of
.
sum
of the
symbols
2, of j? x
two
polyhedra p t and
pz
is
called
the the
sum, modulo
and
As
in the analogous case of
polygons,
p is a set p and pz
1
of polyhedra
.
whose
faces are in
or in
but not in both.
<z
The number of variables in (H 2 ) is 2 and the rank of H is a a by Theorems 86 and 85. Hence the solutions of (H 2 ) are
on a
set
linearly dependent
of
<z
g
linearly independent solutions.

g
Since any
a set of
as
1 of the
columns
of
are linearly independent,
such
columns, together with one other solution linearly independent of them, will furnish a set of linearly independent solutions of (H ).
The symbol for any plane is a solution of (H 2) linearly independent of the columns of H 3 For let I be any line meeting no or a\ Any column of H, represents the polyhedron bounding an a 8 , and such a polyhedron is met by I in two points or none. By
. 5
reasoning analogous to that used in the proof of Theorem 87, it follows that I meets the sum, modulo 2, of the boundaries of any number
of
's in an even number of points or none. Since I meets each plane in one point, the symbol for TT. is not linearly dependent on the TTj columns of By the last paragraph we now have g
8
THEOREM
90.
Any
solution of
columns of H 3 and
(H g )
is
the
a& -,
1
.
the symbol for
any one of
is the
planes
all
TT
TT
TT
COROLLARY
of a
set
1.
Any
polyhedron
sum, modulo
2,
of
a,
subset
of polyliedra
regions.
TT^
7r
2,
consisting of one plane
and
polyhedra which
bound convex
Proof.
edges,
Let
vrn
be a
set of
and
faces of a given polyhedron
planes containing all vertices,, and such that ?r ?r , TT

,
,
"fl-
are not concurrent.
expressible
as a
By the theorem the sum of the boundaries

,
given polyhedron is eithei of some of the a 8 's deter-,
7r n or as a sum mined by ir lt ir2 s some of the a 's. In the course of the argument ahove
,
of
one of these planes and
it was shown that any poly8 hedron expressible in terms of the boundaries of the & 's was met in 1 an even number of points by any line not meeting an a or an a 7r n is met One of the planes ir^ 7r 2 by such a line in one point.
. ,
Hence any polyhedron which is

of the boundaries of
8 o. 's
the
sum of such
a plane and a
number
of
is
met by
this line in
an odd number
points.
Hence
COROLLARY 2. DEFINITION. A polyhedron which is the sum, modulo @, of a number of boundaries of convex three-dimensional met in an even number of points by any line not meeting regions is Such a polyhedron is said to be even. A polya, vertex or an edge.
hedron which
is the sum, modulo 2, of a, plane and a number of boundaries of convex three-dimensional regions is met in an odd number of points by any line not meeting a vertex or an edge. Such a polygon is said to be odd.
EXERCISE
p be a polygon and IT a polyhedron such that TT contains no vertex of no vertex or edge of TT. If p and TT are both odd they have p and p contains an odd number of points in common. If one of them is even they have an even number (or zero) of points in common.
Let
194. Regions
is
bounded by a polyhedron.
boundaries
of a set of
the
sum
of the
An even polyhedron p convex three-dimensional
and polyhedral regions, regions are denoted by

The polyhedron
we may
a *,
the
a*,
assign the notation so that these
-,*.
of the
is also
sum
boundaries of
because the
sum
of all the
columns of
is
(0, 0,
-,
0).
There
is
no other
of
linear expression for

is
in terms of the boundaries of the
a 3 's, because there
only one linear relation
among the columns
8.
a direct generalization of what is said at the beginning 3 3 3 a* 187, it is easily seen that the points of ai a 2 together with those points of their boundaries which are not on p, constitute a region bounded by p; and that the points of a k 3 a k+ i 3
This
is all
of
187.
As
in
together with those points of their boundaries which are not on p, constitute a second region bounded by p. With a few additional details (which are generalizations of those given in the proof
'
aa3 3
of
Theorem
75) this constitutes the proof of the following theorem.

91.
THEOREM
polyhedron.
Any
even polyhedron is the boundary of each of two

all
and only two regions which contain
points of space not on the
These regions are such that any broken line joining a point of one region to a point of the other contains a point of the polyhedron. Any two points of the same region can be joined by a broken line consisting entirely of points of the region.
By
a similar generalization of Theorem 76, we obtain

92.
region containing
odd polyhedron is the boundary of a single points not on the polyhedron. Any two points of this region can be joined by a broken line not containing any point
THEOREM
Any
all
of the polyhedron.
COROLLARY.
point not on
except P.
195.
it
Any point P on a polyhedron can be joined to any by a broken line containing no point of the polyhedron
for the projective plane. DEFINITION.
The matrices Ei and E 2
segment, interval, broken line, polygon, two-dimensional convex region, or three-dimensional convex region associated with a senseclass
among
its
points
line,
interval,
broken
is called an oriented or directed segment, polygon, two-dimensional convex region, or
three-dimensional convex region. DEFINITION. Let a 1 be any segment which, with its ends A and B, 1 denote the oriented segment is contained in a segment s, and let s obtained by associating a 1 with one of its sense-classes. The sense1 class of s is
contained in a sense-class of s which

is
is
either
is
S(AO)
or
S(OA]
if
any point
of a 1
In the
first
case
said to be
oi/w/t/ew/
L<J
1
.
t>
a .LIU u
i/uc oouuiiu. t/itso a. ID
Salu. to
uv
tively related to s
To
aid the intuition,

of
we may think of
is
an.
oriented segment as
is
marked with
an arrow, the head

oriented segment.
which
at the
end which
positively related to the
to
Obviously, if one end of an oriented segment is positively related it, the other end is negatively related to it, and vice versa. DEFINITION. The sense-class S(A 1 Z 8) of a polygon.^^ n l
AA
AA
and the sense-class

case
of
on the edge A 1A 2 are said to agree in the order {AÂBA^ and to disagree in -case of the order
S (AB)
185, the segments a*, Returning to the notation of each he associated with two senses. They thus give
directed segments.
1 each a
,
a*
may
rise to
2a
<
Assigning an arbitrary one of the two senses to we have a l oriented segments to which we may assign the
3
,
notation s a
-,8^.
We
l
shall denote the oriented
segment obtained
the negative of s\ may be indicated
by changing the The relations
sense-class of si
of the s
's
by
s/
and
call it
,
to the points
a,
call
a
ft
by means
of a
matrix which we shall
E^
In the matrix
E^^
the element of the ^th row and /th column shall be 1, 1, or 0, according as the point a? is positively related to, negatively related
to,
or not an
It
is
end
of,
the oriented segment
sj.
clear that the signs 1

of this
and
1 are interchanged in the yth

Sj
column
matrix
if
the sense-class of
is
changed.
Since the
sense-class of each
segment is arbitrary, a matrix equivalent to E t can be obtained from the matrix H I} 186, by arbitrarily changing
1. one and only one 1 in each column to In the case of the triangle, by letting the se fo ^
give rise to s lt
respectively,
we
derive the
from HjOf
151:
The elements
of the
matrix
E may
}
be regarded a
of a set of linear equations analogous to the equations
Let
(ic 1 ,
xz
-,
x ai ) be a symbol in which the
x's can take
on any
integral values, positive, negative, or zero, and let this symbol represent a set of oriented segments comprising s/ counted x. times if
xt
si
is positive,
s/
counted
zero, i
x times
i
if
is
negative,
1, 2,
and neither
,
nor
s/ if
is
taking on the values
a^
The
ments.
sense-class of
class of each of its
an oriented polygon agrees with a definite sensesides and thus determines a set of oriented seg(x^,
The symbol
xz
is
0i )
for this set of oriented segments
may
also be regarded as a
symbol
for the oriented polygon.
Each
vertex of the polygon

another.
are
positively related to one of the oriented

,
x ai ) and negatively related to segments represented by (x^ xz Thus if sf and s> meet at a certain vertex to which they
,
that
(a?j,
both positively related according to the matrix E 1 we have .= l and #,. = 1 or that ^.= 1 and ^=1 in the symbol t
,
ai )
for
any directed polygon containing the
sides
a{ and
a,..
The
x'a corresponding to the
segments not in the polygon must of

,
course be zero.
linear equation
Hence the symbol (a? 1 must satisfy the ) whose coefficients are given by the row of E x cor,
responding to the vertex in question.
If s {
a
and
,
are oppositely
C{
related to a vertex according to the matrix
E we
must have
for
=1
and 2y=l or xi
and #y =
1 in the
symbol
of
any directed
be
satis-
polygon containing the sides a{ and ajt Hence in this case also the
linear equation given
fied.
by the corresponding row
E must
x
Finally, the equation given by a row of E 1 corresponding to a point which is not a vertex of the polygon is satisfied because all the aj/s corresponding to edges meeting at that point are zero.
Hence
matrix
the
symbol for a directed polygon must
be
a solution of In the case
the
linear equations whose coefficients are the elements of the rows of the
E r These equations shall be denoted by (E^. the triangle they are = 0>
of
^+^ +^+^
(6)
By
reasoning entirely analogous to that of

of
186,
it
follows that
(E^ in integers represents one or more directed simple polygons. The situation here differs from that described in
any solution
the modulo 2 case, in that the same
side may enter into more than one polygon and the same polygon may be counted any number of
times in a set of polygons. Since each column of the matrix
contains just one 1 and
one
1,
the
sum
of
vanishes identically. relation among the equations (E^, because the matrix Ej and the equations (E x) when reduced modulo 2 are the same as H a and
linear relation among the equations (E ) would a (H^, and so any imply one among (H^.* Hence the number of linearly independent a 1. The number of variables being a^, the equations of (E^ is Q
the left-hand members of the equations (E ) There can be no other linear homogeneous
number of linearly independent solutions Theorem 74, this number is equal to

.
is
o^
tf
+l. In view
of
It will be recalled that in the
modulo 2 case one
class of solu-
tions of Equations
given by the columns of the matrix 2 These columns are the notation for the polygons bounding the coneach of these be vex regions at,' , &\polygons replaced by one of the two corresponding directed polygons, a set of solutions is
(H^
is
determined for the equations (EJ. These solutions are obtained minus signs so that the matrix 2 by introducing directly from the
columns become solutions of (EJ. This is possible in just two ways 2 for each column, because each polygon bounding an a has two and shall be denoted by so obtained matrix A only two sense-classes. E In the case of the triangle such a matrix is
.
It
is
to the
number
evident on inspection that the rank of this matrix is equal the of columns. That is to say, unlike those of 2 ,
* The coefficients of any linear homogeneous relation among the rows of E x maymodulo 2 it be taken as integers having no common factor. Hence on reducing would yield a linear relation among the rows of H r But as the only linear relation linear relais no there is rows all the of zero, among the rows of Hj is that the sum not involving all the rows. There could not be two such tion among the rows of derive a we could them E because by combining relations among all the rows of x
488
THEOEEMS ON SENSE AND SEPARATION

E
2
columns of
good
for the
:
matrix
are linearly independent. The same proposition hold; E in the general case. This can be proved as
2
follows
H^ x By the reasoning used above for the rows lows that any linear relation among the columns of E 2 implies on< th< among the columns of H 2 Since the only such relation among involves all the columns, we need only investigati columns of
of
it fol
.
and
linear
homogeneous
relations
among
the columns of
If
Ea
in
which
al
the coefficients are different from zero.

of E 2 common would have
such a relation existed
two columns
corresponding to regions having an edge ii numerically equal multipliers in the relation
else
the elements
cancel.
But since any two
corresponding to the common edge would no 2 of the convex regions a can be joinei
of points of these regions and o the edges of their bounding polygons, it follows that all the coeffi cients in the relation would be numerically equal, i.e. they could al
by a broken line consisting only
be taken as
-f-
1 or
^,
1.
l n containing all the points and segment our figure are not all concurrent three of them, say l^ lz l &) forr a triangle. Let us add together all the terms of the supposed rela
Now
the
lines
of
tion corresponding to regions
a 2 in one
of the four triangular region
determined by l^
to this triangular
The elements corresponding to edges interio region must all cancel, because they cannot canct
2,
against terms corresponding to regions a

region.
exterior to the triangula

of
The sum must represent an oriented polygon

all
which th
on the boundary of the triangular region. This oriente 183, must be identical with the boundary of the tr: polygon, by angular regions associated with one of its two sense-classes. If w
edges are
by lv
operate similarly with the other three triangular regions determine 1 , 1 we obtain three other oriented polygons. But since th 8 2
,
linear combination of the
columns of E 2 is supposed to vanish, eac edge of the four triangular regions should appear once with on sense and once with the opposite sense, and this would imply the
triangle there
of
in the case of a
would
2,
exist a linear
homogeneou
relation
among the columns

case the
2
Hence in every
contrary to the observation abov columns of E 2 are linearly independent.
dependent on the columns of \.
This is in sharp contrast with the 186. property of the equations (HJ) stated at the end of 196. Odd and even polygons in the projective plane. Let us apply the results of the section above to the theory of odd and even polygons. Since any polygon is expressible in terms of the columns of
E2
an odd polygon must be so expressible. Let us write this expres=
sion in the form

(?)
pp
sf,
,
where
represent the columns of
E p
2,
.,
is
the symbol for
the given oriented polygon, and p and y^, may be taken so as not to have a common
y^
are integers
which
factor.
Since the coefficients do not have 2 as a common factor, (7) does not vanish entirely when reduced modulo 2. But since an odd polygon is not expressible in terms of the columns of H must contain the 2 p
,
factor 2,
and
of
(7)
.
columns
must reduce, modulo 2, The only such identity

.
to
is
an identity among the the one involving all
the columns of
2
Hence the
T//S
are all odd.
But in order that

y's
the edges not on the odd polygon

ing to
s 's
Since any two having an edge in points not on p can be joined by a broken line not meeting p
p shall vanish, the common must be equal.
correspond-
(Theorem
:
76), it follows that all the y's are equal.

k, it is
If
they are
all
taken equal to theorem
obvious that p
= 2h
Hence we have the

in
THEOREM
the
93.
The symbol p for any odd polygon

aa
is expressible
form
(8)
*P=$<*1
e i is
where each
+1
may
or
1.
This theorem
the matrix
be verified in a special case by adding the columns of
E2
given above for a triangle.
The sum
is (0, 0, 2,
2, 0, 0),
which
represents a line counted twice. The number 2 is called the coefficient of torsion of the two-sided polygon (cf Poincare, Proceedings of the London Mathematical
.
Society, Vol.
etc. is
XXXII (1900), p. 277. The systematic use of
the matrices
E p E2
due to Poincar6).
:
Another form of statement for Theorem 93 is the following If the region bounded by an odd polygon p be decomposed into convex regions each bounded by an even polygon, each edge ofp is on the boundary of
Aii even polygon
is
also
all

is
(7).
Aside
from a common factor of pression for p of the form
only one ex(7), for if not, by eliminating p we could obtain a linear homogeneous relation among the columns of E 2
the coefficients, there
.
Let R be one of the two regions bounded (Theorem 75) by p, which contains one of the convex regions a] for which the corre-
sponding y in
equal
a
(7) is
having an edge in
y's
in (7) in
not zero. Any two s 's corresponding to a?'s common must be multiplied by numerically order that the symbol for the common edge
Since any two points of R can be joined by
of points of
2
shall not appear in p.
broken line consisting entirely
R, this implies that
the coefficients yi corresponding to the a
's
in
are all numerically
equal to an integer k. terms in the right-hand
From this member of
it
follows that the
sum
of the
(7)
equal to p, because each edge of R. Since the equality just found is of the form y a are all numeri(7), and (7) is unique, we have that p and y^, cally equal to k. Obviously the factor Jc can be divided out of (7).
is
2 one of the a 's in
, ^
which correspond to 2 's in p is an edge of one and only
Hence we have
the
THEOREM form
(9)
94.
The symbol
for an even polygon
is expressible in
?=%'& i=l
e { is
where
or -f
or
1.
The
af's such that the e{ 's

t/ie
with the same
subscripts are not zero are the af's in one of
regions
R referred
to
in Theorem 75.
DEFINITION.
By
the interior (or inside] of an even polygon
is
meant
that one of the two regions determined according to Theorem 94 which contains the f s having the same subscripts as the non-zero e/s in (9). The other region is called the exterior of the polygon.
EXERCISE
Identify the interior of a two-sided polygon as defined above with the interior as defined in 187.
197.
be a polygon which
One- and two-sided polygonal regions. Let A A n An is the boundary of a convex region R for which
, ,
0' are any two points of R, then S(OAÂ 2 ) ^(O'AÂJ (cf. 161) with and 0' are on the same side of the nne A A in respect to R' because
D'
'
'
1-2
gam
'
8(0 A^) =
because A^ and
are
on opposite sides
of the line
OA A
,
and
are on opposite sides of the line
OA
S,
etc.
sense-class in
R,
which we shall
call positive,
determines a
i.e.
positive sense-class in
any convex region
R'
containing R,
the
sense-class containing the given sense-class of R.
This, in view of
the paragraph above, determines a unique sense-class on the polyis positive, where gon bounding R, by the rule that if
^(OA^j
is
in R, then
S(A 1 A Z A 8 )
is
positive
on the boundary
of
and
if
S(OA From
is positive, Z A^)
161
it
then S(A z A lA n ) is positive on the boundary of follows without difficulty that this determination
of
is
R.
is

Conversely,
it
the convex region R'. obvious that by this rule a sense-class on the
a definite sense-class in R.
boundary
of
R determines
2 Let a be any planar convex region which, with its in a convex planar region R, and let a.1 be contained is boundary, 2 Let s 1 denote the oriented any segment on the boundary of a 1 segment obtained by associating a with one of its sense-classes,
DEFINITION.
and
one
of
denote the oriented region obtained by associating a 2 with The sense-class of s 2 is contained in a its sense-classes.
2
certain sense-class of
is
in a
s
1
and
A
a
and
s
1
,
1
,
then
and
are
the sense-class of
R which may be denoted by S(OAB], where B are on a\ If S(AB) is the sense-class of said to be positively related and if S(AB) is not they are said to be negatively related.
;
As
of R.
pointed out above, this definition is independent of the choice Let R t and R 2 be two convex regions having no point in com-
mon and bounded by two

respectively
polygons
which have
in
A A 2A S common only the

1
.
A m and A A B
vertices
l
Bm
A and A and
the points of the edge AÂ Z Suppose, also, that R R and their I} boundaries are contained in a convex region R. These conditions are
satisfied if
of lines four of
R and R 2 are 2 's, and AÂ Z is an a\ determined by which are such that no three are concurrent.
:
a set
The rule given above for determining positive boundaries of Rj and R 2 requires that if 8 (O.-l A n ]
sense
is
on
the
positive for
of
it
A is a point ot the edge AjA f 11 O' is any point ot K 2 in R. Hence from on the opposite side of the line z S(0'A Z AJ is positive, and hence S(A 2 AA^ must be the positive senseR I} where
,
is
A^
class
Let R
I
,
on the boundary of and R be two

L
,
R.
of the a?'s
determined by a set of lines
let
sense-class
(x v
as
the boundary of R 1 associated with the positive determined in the last paragraph be denoted by
xa ) according to the notation of 195; and let the boundary of R 2 associated with the positive sense-class determined at the same time be (y^, yz ,*, ya^~ The notation may be assigned
%2
>
',
and y l refer to the edge AÂ Z common to the boundaries ajj R and R 2 In this case, if x v 1, y^= 1, and if x^ = 1, +1 for the positive sense for the boundary of R I is S (AÂA^ and for
so that
of
l
.
y^
the boundary of
bols
(aj l}
R2
is
S (AÂA^. Hence
(y^ yz ,
-,
the
sum
of the
two sym-
x2
xa ) and
y a)
I}
ary of the region R' composed of

this
the symbol for the boundR 2 and the common edge Z

is
AÂ
boundary being associated with a
sense-class S'
which agrees
of
with the positive sense-class on any edge of the boundary R 2 which is an edge of the boundary of R'.
RI
or
By
repeated use of these considerations
it
follows that
if
a set of
a 2 's with their boundaries constitute a convex region R and its xai) for the boundary of R assoboundary, the symbol (x t xz
. . , ,
ciated with a sense-class

of the
which
is
designated as positive,
is
the
sum
8 symbols for the boundaries of the 's, each associated with its positive sense-class. In other words, the symbol for the boundary of R associated with its positive sense-class is the sum of a set of
columns
each multiplied by -f 1 or 1 so that it shall be the 2 boundary of the corresponding a associated with the sense-class which is positive relatively to the positive sense-class
of
2,
symbol
of
R.
for the
By comparison with Theorem

also)
is
94,
it
follows (as
is
obvious
is
from other considerations boundary

of a
that
even.
any polygon which
the
convex region
The argument
modification to
in the paragraph above applies without essential
any region bounded by a polygon and having a 168. Hence any polyunique determination of sense according to gon bounding a two-sided region is even. Moreover the steps of the argument may be reversed as follows
:
If the
symbol
for
any oriented polygon
be expressible in terms
197, 198]
MATRICES OF ORIENTATION
493
of the columns of
e,, ,
where the non-zero coefficients e ik p is the are e^, 2 boundary of the region R consisting of 2 2 a 2 and those points of their boundaries which are not 3 on p. If R' is a convex region contained in R, and its positive sense-class be determined as agreeing with the positive sense-class of one of the regions a^ 2 a<2 2 a ik2 it must agree with that of every a 2 with which it has a point in common for otherwise the symbols
2,
H
,
in the
form
(7),
'
'
-!
>
*,
-,
two of the a 2 's would not cancel in (7). If R" is any other convex region contained in R, and its positive sense-class is also determined by this rule, the positive sense-classes of R' and R" must, by definition, agree in any region common to
for the
common
edges of
R' and R". Hence R is two-sided according to have by comparison with 196
168.
Thus we
THEOREM
95.
The
interior of
an even polygon
is
a two-sided
region.
198. One- and two-sided polyhedra. Let the vertices of a polyaao the edges by ai 1 a 2 l hedron be denoted by Oi, a 2 aai l a 2 2 Assigning an arbitrary one of its and the faces by i 2 a2 2
-
sense-classes to each edge, there
segments
si
1
,
Sz
1
,
,
sai :
is determined a and a matrix
set of oriented
in which i
0,
= 1,
of
s,-
2,
a,1
.
-,
= 1,
2,
-,
a\
and
e*,
is
1,
1,
or
according as
is
positively related to, negatively related to, or
not an end
is
Assigning an arbitrary one of its sense-classes to each face, there saa a determined a set of oriented planar convex regions Si2 s 22 ,
,
and a matrix
in which
TT.
&2
,
,N
(Gif),
a2 and e,- 2 is 1, 1, or 0, * 0.1 j 1, 2, -, 1, 2, 1 197), negatively related according as s* is positively related to (cf. 2 By the last section each column to, or not on the boundary of s
; ; ,
za]), in the sense explained in the symbol (x i} x 2> -, 195, for an oriented polygon obtained by associating the polygon bound2 ing one of the a 's with one of its sense-classes. Changing the sense2 class assigned to any a to determine the corresponding s 2 amounts to
of
2
is
494
[CHAP. IX of
,
In this case there are just two non-zero elements in each row Hence the sum of the columns of E 2 will reduce to (0, 0,
E2
0) if
and only
if
the sense-classes have been assigned to the faces of the
polyhedron in such a way that one of these elements is + 1 and l This means that each s is positively 1 in each row. the other
related to one of the s
related to the other.
2
's
on whose boundary
it
is
and negatively
are
Thus the
faces are related as
the
a?'s
which constitute a two-sided region bounded by an even polygon

in the plane
(
197).
DEFINITION.
polyhedron for which the sense-classes can be
assigned to the edges and faces in such a way that eacli edge is positively related to one of the faces on whose boundary it is and
negatively related to the other,
is
said to be two-sided, or bilateral',

is
and one
for
which
this
assignment of sense-classes
not possible
is
said to be one-sided, or unilateral.
Changing the assignment of sense-classes on an edge amounts merely to multiplying the corresponding column of E X and row of E 2 by ~ 1> au(i changing the assignment of sense-classes on a face
amounts
to the same operation on a column of E Consequently the polyhedron is two-sided if there is a linear relation whose coeffi1's among all the columns of E cients are 1's and and it is one.
sided
if
there
is
no such
if
relation.
It is
also obvious
from these
considerations that
a polyhedron satisfies the definition of two-
its edges, it
sidedness (or of one-sidedness) for one assignment of sense-classes to does so for all assignments. therefore infer at once
We
THEOREM
rank of
96.
is
polyhedron
2
is
one- or two-sided according as the
a z or
1.
By
reference to
195 we find
is
COROLLARY. The protective plane
a one-sided polyhedron.
In the case of any polyhedron in which each edge is on the boundary of only two faces, it is seen that the only possible linear relation among the columns of E2 reduces to one in which each coefficient is + 1 or 1, for any other relation would imply that a subset of the faces determines a polyhedron.
THEOREM
97.
VtnliiTi.p.fJ.rn'r):
"hnii.'n.rln'n.n
n fnvimanf wonnnw
o/i7ii'/>7j
I'Q
There
of of
is
no
difficulty in proving that
if
C and
an edge the two
si of
the polyhedron bounding R, and

sides of the plane
D are two E and E'
points
points
faces having this edge on their boundaries, then
and
E' are on opposite
OCD. Hence
Hence the
that
sense-classes are assigned according to the rule above to

s/
the two faces having the edge

si is positively related to
on their boundaries in such a way
one and negatively related to the other.
DEFINITION. By an oriented polyhedron is meant the set of oriented 2 two-dimensional convex regions [s ] obtained by associating each face of a two-sided polyhedron with a sense-class in such a way that if
sense-classes are assigned arbitrarily to the edges to determine directed
segments, each of these directed segments is positively related to one of the oriented two-dimensional convex regions on whose boundary
it
is and negatively related to the other. oriented faces of the oriented polyhedron,
The
s 's
are called the

s
l
and the
's
its
oriented
edges.
CoKOLLARY. given two-sided polyhedron determines two and only two oriented polyhedra according to the definition above.
DEFINITION.
is
contained with
3 Let a be a three-dimensional convex region which 2 its boundary in a convex region R, and a- a two8 3
.
3 Let s denote a dimensional convex region on the boundary of a 2 2 associated with one of its sense-classes, and let s denote a associated
with one
one
of
of its sense-classes.
The
sense-class of s
is
contained in
8
,
be a point of a and the sense-classes, say S, of R. Let 2 A, B, C three points of a such that S(OABC) is S. Then if S(ABC) 3 2 2 2 is the sense-class associated with a to form s , s and s are said to
,
be positively related.
Otherwise they are said to be negatively related.
is independent of any particular choice 8 convex region R containing a and its boundary. From what 2 has been proved above it follows that if each a on the boundary 8 of an a is associated with a sense-class in such a way as to be posi-
By
161 this definition
of the
tively related to the oriented region determined
3 by a and one
of its
in
The definitions made in this section are extended to polyhedra which each edge is on an even number of faces (instead of only
two, as
we have been
DEFINITION.
polyhedron
supposing) as follows: is said to be two-sided
if
sense-classes
can be assigned to the edges and faces in such a way that each resulting oriented edge is positively related and negatively related to
equal numbers of the resulting oriented
faces.
EXERCISES
1.
An
odd polyhedron
2.
Make
one-sided and an even polyhedron is two-sided. a discussion of one- and 'two-sided polyhedral regions in space
is
analogous to the discussion for the two-dimensional case in
197.
199. Orientation of space. ized to the three-dimensional
The matrices
case.
of
195 can be
,
general-
Let
si,
s,
si.
be the oriented
.
segments obtained by associating each of the segments al, a%, with an arbitrary one of its sense-classes. In the first matrix,
.,
a\
^=(4),
i
= l,
Sj.
2,
#o /=!,
;
2,
a^,
and
e],
is
+1,
1,
or
according
as .a* is positively related to, .negatively related to, or not an end
can be formed from Hj by changing one 1 to a 1 in The choice of the 1 in the /th column amounts to the choice of the sense-class on aj- which determines si. As an for a tetraexercise, the reader should form Ej from the X given
ut
t
each column.
hedron in
bols
of
152.
1
Sets of oriented segments s
are represented as in
se^),
195 by symare
positive
or
the form (x v
a? , 2
where the
x's
negative integers. By the same argument as in 195, if this symbol represents a set of oriented segments each of which is an edge of a
polygon associated with that one of its sense-classes which agrees with a fixed sense-class of the polygon, it is a solution of the equations,
and, conversely, any solution of these equations is the symbol for one or more such sets of oriented segments. Thus any solution of may be regarded as representing one or more oriented polygons.
Let s s be the oriented two-dimensional convex ., s; 2 regions obtained by associating each a 2 with an arbitrary one of its senseclasses. The oriented two-dimensional regions obtained by associating 2 the s' 's with the opposite sense-classes be denoted
, ,
may
by
sf,
s|,
.,
s^ respectively.
In the second matrix,
= 1,
2,
.,
al j
= 1, 2,
to,
.,
az
and
e| is 1,
1,
is
positively related
negatively related
to, or
of sf.
E2
can be formed from

of
(x^,
or according as s\ not on the boundary

of the 1's in
each column
be a symbol
by changing some
2
,
to -1's in such a
-
way
l!
that each column shall
x2
xa ^)
for a set of s s
whose
sense-classes all
agree with that of the oriented polygon determined by associating 2 the boundary of s with one of its sense-classes. This is
possible
at the beginning of 197, since each column 2 the symbol for the boundary of one and As an 2 only one a, exercise, the reader should form E2 from the for a tetra2 given 152. hedron in
of
by the argument
is
symbol of the form (x^ x2 , ., x a ) in which each x is a positive or negative integer or zero may be taken to represent a set of
X times
L
oriented two-dimensional convex regions which includes sf counted if x { is positive, s? counted x{ times if xi is negative,
sf if x. is zero.
and does not .include

oriented polyhedron
(
197),
it is
If this symbol represents an a solution of the equations
CE } \ g/
V
.7
y j jj e?..
UJ
(i
\
=l
-*-j
2,
~*)
')
=1
a} l/*
For consider the iih
of these equations
If
an oriented face
s],
of the oriented
polyhedron
is
positively related
to
it
;
contributes
a term
+1
to the
left-hand
member
;
of this
equation for if s| is this oriented face, xk 1 and e?t= 1 and if s| is this oriented face, x l and e|.= .1. An oriented face which k is related to contributes a term 1 to the left-hand sj negatively
member
?
of this equation; for

;
if
s% is this oriented face,
fc
=l
and
and
if
is
this oriented face,
#A =
and
eik
1.
Hence
there are as
many terms
equal to
+1
as there are oriented .faces
oriented faces negatively related to
s}.
If neither s
s\
2
fc
nor
its
is
in
the oriented polyhedron, or
if
sg
does not have

the
on
boundary,
the
Mi
term
of this
t
equation
is zero, for in
first
case xk
=
,
and
in the second case eik
0.
Hence by the
definition of
an oriented
,
xa ^ polyhedron, each of the equations (E 2 ) is satisfied if (x^, xz represents an oriented polyhedron. In particular (Theorem 97) the symbol for either oriented polyhedron determined by the boundary
of
an
is
a solution of (E 2 ).
One-sided polyhedra do not give rise to solutions of (E 2 ). s 8 s be the pairs of s2 s* s 2 and Let s and sJ3 and 3 oriented three-dimensional convex regions determined by a, *, a^
}
respectively according to the definition in
197.
In the third matrix,
= 1,
2,
o;
3
and
e?.
is
+1,
1,
or
according
as s f
is
positively related to, negatively related to, or not on the

of sf.
boundary
1's
The matrix
can be formed from
by changing
resulting
solu-
to
1's in
the columns of
in such a
way that the

As an
columns are the symbols

tions of (E 2 ).
for oriented polyhedra
and therefore
This
is
possible
by Theorem
96.
exercise, the
reader should form
E8
from the
of
H
E
x
The sum
of
of the
columns
is
152. given for a tetrahedron in because each row , 0) (0, 0,
contains one
+1
and one
1.
There can be no other linear
relation
among
the columns of
reducing modulo 2, of r Hence the rank of
because this would imply, on more than one linear relation among the columns
1,
is
x
1,
and the number
of solu-
tions of
in a linearly independent set on

is
2
which
all
the solutions
are linearly dependent
a1
a +l.
Since the rank of

relation
of
1
is
ai
is
among
the columns of
H
c
z>
the rank of
at
aQ and since every homogeneous linear E 2 implies one among the columns aQ It is, in fact, at least least a 1
,
.
+ 1 because, by Theorem
'
93, the symbols for a set of columns

all
c
i>
'
>
ck
which represent oriented polygons bounding

form
*
the
s 's
of a protective plane satisfy a relation of the
(10)
e c
1
+ A +"-+*fc a=2
where
I is
the symbol for a line in this plane and
e
lf
2,
en
are
Reducing modulo 2, this gives rise to a homogeneous linear relation among the columns of H. which is not one of those
or
I.
+1
Thus there
)f
f
t
These are
+
a1
ihan
+ l linearly independent columns (E 1 ), and as there are not more than 1 linearly independent solutions -of there are not more (Ej), a Q + 1 linearly independent columns of E2> Hence the
are at least o^
all .solutions of
rank of
is
2
a
l
+ 1,
which by Theorem 85 is the same as
*-, +
gon
!.
, ,
# ffi ) for any oriented polyIn consequence, the symbol (x^, x z is linearly expressible in terms of the symbols for oriented polygons which bound convex planar regions. It can easily be proved
that in case of an odd polygon this expression takes the form (10) c k are not neceswhere, however, the polygons denoted by c lt c z
,
sarily all in the
same
a3
plane.
Since the
number
#2
of variables in the equations
(E 2 )
is
a2 and the
rank
of
is
2
+ 1,
the
number
of solutions in
a linearly inde-
1. pendent set on which all solutions are linearly dependent is a 8 Hence the rank of o all solutions of (E,). The columns of E o B are n &' It cannot be less than a 1. cannot be greater than # b 1, because, g
^
less
on reducing modulo 2, this would imply that the rank of H 8 was Hence the rank of E g is # g 1. Since the symbol 1. than a 8 2 2 's is for any oriented polyhedron whose oriented faces are s 's or a solution of (E 2 ), it follows that it is expressible linearly in terms of the symbols for oriented polyhedra which bound convex, threedimensional regions. Since the rank of
Eg
is
ag
1,
the set
of
equations
(E 8 )
j=i
|>>, =
,
<t'
= l, 2,...,g
When
reduced
190,
must have one solution distinct from (0, 0, modulo (2) this solution must satisfy (H a ) and
reduce to
(1,
0).
therefore,
by
Since each equation in the set (E 8) has only two coefficients different from zero, and these coefficients are 1, it follows that all the aj's are numerically equal in a solution
1).
!,,
(#!>
#v
x's
'
'
the
may
*") ^ (^s)' Sin ce the equations are homogeneous, all 1. be taken to be +1 or
The
ith of these equations is of the
form
500
e?.
t

+1
or
[CHAP.IX
being
1 according
e?.
s
as or
sf
is
positively or negatively
related to s?,
and
being
s?.
+1
or negatively related to
Hence,
x's are
if
1 according as sf is positively the set of regions represented

1 includes that one of
sj*
by a
s?
solution in
which the
and
s? to
which
Sj 3
s?
is positively related, it also
includes that one of
and
to
which
to
s? is negatively related
and
if it
includes that
one of sft and

that one of s? and
which s? is negatively related, it also includes s? to which sf is positively related. a

(0, 0,
,
Hence the
existence of a solution of (E g) other than

8
0)
s s such that each s is implies the existence of a set of s 's and positively related to one of them and negatively related to another.
Since the notation sf and

the yth column of
(1,
sf
1,
by
may be interchanged by multiplying the notation may be so arranged that
!,,
s
2
1)
is
a solution
of
With the
3
each
is
positively related to one
and negatively
notation so arranged, related to another.
We thus
T
7r
I}
,
have
98.
7r n
THEOREM
-,
If each of
a,
the a?s determined ly a set of planes

2
,
in
protective space is arbitrarily associated with one
of
its sense-classes to
&
determine an oriented planar convex region
each of the a 's can l>e associated with one of its sense-classes to determine a three-dimensional convex region s 8 in such a way that each s 2 8 is positively related to one s and negatively related to another.
The
lateral
set of s
's
described in this theorem

198.
is
a generalization of an
oriented polyhedron as defined in
If the definition of uni-
dimensions,
and bilateral polyhedra be generalized to any number of it is a consequence of this theorem that the three-
dimensional space is a bilateral polyhedron. In general, it can easily be verified, by generalizing the matrices E^ E 2 E g etc., that projective spaces of even dimensionality are unilateral polyhedra and
,
protective spaces of odd dimensionality are bilateral polyhedra.
EXERCISE
INDEX
About, 159 Absolute conic, 350, 371 Absolute involutions, 119 Absolute polar systems, 293, 373 Absolute quadric, 369, 373 Addition of vectors, 84
Affine classification of conies, 186 Affine collineation, 72, 287 Affine geometry, 72, 147, 287 Affine groups, 71, 72, 287, 305 subgroups of the, 116
;
Asymptotes
of
of
a conic, 73
;
Axis, of a circle, 354
of a conic, 191
;
;
of a parabola, 193 of a quadric, 316 radical, 159 ; of a rotation, 299 of a translation, of a 320 317; twist,
;
;
a line
reflection, 258
Backward, 303
Barycentric calculus, 40, 104, 292, 293 Barycentric coordinates, 106, 108, 292 Base circle, 254 Base points of a pencil, 242 Beltrami, E., 361
Bennett, A. A.,
iii
Agree (sense-classes), 485 Alexander, J. W., iii, 405 Algebra of matrices, 333 Algebraic cut, 15 Alignment, assumptions of, 2 Analysis, plane of, 268 Angle, 139, 231, 429, 432 numbered, 154 Angles, equal, 165 sum of two, of rotation, 325, 327 154 Angular measure, 151, 153, 163, 165, 231, 311, 313, 362, 365 Anomaly, eccentric, 198 Antiprojectivities, 250, 251, 253 Apollonius, 235 Arc, differential of, 366 Area, 96, 149, 150, 157, 311, 312; of ellipse, 150 Assumption, Archimedean, 146 Assumption A, 2 Assumption C, 16 Assumption E, 2
;
Between,
15, 47, 48, 60, 350, 387, 430, 433 Bilateral polyhedron, 494 Bilinear curve, 269
Biquaternions, 847, 379, 382 Bisector, exterior, 179 interior, 179 perpendicular, 123 BOcher, M., 256, 271 Boger, R., 168 Bolyai, J., 361
;
Bonola, E., 59, 362, 363, 371, 376 Borel, E., 60 Boundary, 392, 474, 482 Bounding polygons, 470, 482 Broken lines, 454 directed, 484 oriented, 484 Bundle, of circles, 256 center of, 436 of directions, 436 of projectivities, 342 of rays, 435 of segments, 486 Burnside, W., 41
;
Assumption Assumption Assumption Assumption Assumption Assumption Assumption
H
I,
H, 11
,
H, 33
80
J, 7
Calculus, barycentric, 40, 104, 292, 293 Carnot, L. N. M., 90
K, 3
;
commutative law of P, 2 multiplication equivalent to, 3
Assumption Q, 16 Assumption R, 23 Assumption R, 29 Assumption S, 32

cateAssumptions, of alignment, 2 gorioalness of, 23 consistency of, 23 ; of continuity, 16 for Euclidean geometry, 59, 144, 302; of extension, 2; of order, 32 independence of, 23
;
Carslaw, H. S., 362 Cartan, E., 341 Casey, J., 168 Categoricalness of assumptions, 28 Cayley, A., 163, 335, 341, 361 oriented, 462, 463 Cells, 404 Center, of a bundle of rays, 435 of a of a conic, 73 of circle, 131, 394 curvature, 201 of gravity, 94 of a of a rotation, 122 pencil, 429, 433 of similitude, 162, 163 of a sphere, 315 Center circle, 231 Centers, line of, 159
; ;
; ;
mental ineorem
;
lor,
zz
Ti-umieii-
equuiLiuii ui, û.6,
three-, 284 sional, 250 Circle, 120, 131, 142, 145, 148, 157, 269, axis of, 354 base, 254 354, 394 bundle of, 256 ; center, 231 center
;
zvo BALBIIUI ui, 11, 174, 176; focus of, 191; interior of, 171,174,176; invariants of, 207; latus
;
a.
rectum
;
circumference of, 148 of curvature, 201 degenerate, 253, 266 director, 200 directrix of, 192 Feuerbach, 169, 233 focus of, 192 fundamental, 254 ; imaginary, 187, intersectional 229 at infinity, 293 length of, 148 properties of, 142
of, 131,
394
metric properties of, of, 198 normal to, 173 81 nine-point, 82 ordinal and metric properties of, 170; outside of, 171; parameter of, 198;
;
; ;
projective, affine,
sification of, 186, 210,
ami Euclidean clas212 a simple

;
closed curve, 402
vertex
of,
191
Conjugacy under a group, 39 Conjugate imaginary elements, 182

Conjugate imaginary lines, 281, 282, 444 Conjugate points with respect to a chain, 243 Connected set, 404 of sets of points, 405 Connectivity of a polyhedron, 475 Constructions, ruler and compass, 180 Continuity, assumptions of, 10 Continuous, 404 Continuous curve, 401 Continuous deformation, 406, 407, 410, 452 Continuous family of points, 404 Continuous family of sets of points, 405 Continuous family of transformations,
;
limiting points of pencils of, 159; nine-point, linearly dependent, 250 pencils of, 169, 233 orthogonal, 101
;
157, 159, 242
power
;
of a point
of,
with
respect to, 162 sides Circular cone, 317
245
direct,
248
U/iu/nv^l Wll u.^ VîXJJ.V.tU. Clockwise sense, 40 Closed curve, 401 Closed cut, 14
Ul
444
406
Coble, A. B., iii Coefficient of torsion, 489 Cole, F. N., 222 Collinear vectors, 84 ; ratio of, 85 Collineations, affine, 72, 287; direct, 61, 64, 65, 107, 438, 451; direct, of a quadric, 260 ; equiaffine, 105 ; focal properties of, 201 ; involutoric, 257 ; opposite, 61, 438, 451 ; in real projective space, 252 Commutative law of multiplication equivalent to Assumption P, 3
Continuous group, 406 Continuum, 404 Convex regions, 385-394
linear, 47
;
sense in overlapping, 424 oriented or directed three-dimensional, 484 ; oriented or directed two-dimensional,
Complementary segments or intervals, 46 Complex elements, 156 Complex function plane, 268 Complex geometry, 6, 29 Complex inversion plane, 264, 265 Complex line, 8 order relations on, 437 and real Euclidean plane, correspond; ;
ence between, 222 implex plane, 154 inversions in, 235 implex point, 8, 156 'one, circular, 317 Confocal conies, 192 Confocal system of quadrics, 348 Congruence of lines, 275, 283 elliptic, 443 right-handed and left-handed elliptic, 444
"
Congruent figures, 79, 80, 94, 124, 134, 139,

144, 297, 303, 352, 369, 373, 375, 394 Conic, 82, 158, 199 absolute, 350, 371
; ;
asymptotes
of,
of.
73
73;
central,
73;
axis of, 191 center confocal, 192;

;
484 Coolidge, J. L., 229, 360, 362 Coordinate system, positive, 407, 408, 416 right-handed, 408, 416 Coordinates, barycentric, 106, 108, 292 polar, 249 rectangular, 811 ; tetracyclic, 253, 254, 255 Correspondence, between the complex line and the real Euclidean plane, 222 between the real Euclidean plane and a complex pencil of lines, 238 between the rotations and the points of space, 328 perspective, 271 projective, 272 Cosines, direction, 314 Cremona, L., 168, 251, 348 of separation, 55 Criteria, of sense, 49 Crossings of pairs of lines, 276 Cross ratio, equianharmonic, 259 of points in space, 55 Curvature, center of, 201 circle of, 201 Curve, 401 bilinear, 269 closed, 401 a conic a simple closed, 402 equidistantial, 356 normal, 286 path, 249, 356, 406; positively or negativelv oriented, 452; rational, 286; 401 simple,
;
Cut-point, 14, 21 Cuts, open and closed, 14 Cyclic projectivity, 258
algebraic, 15
Darboux, G., 251, 324 Dedekind, R.,60

Deformation, continuous, 406. 452 Degenerate circle, 253, 266 Degenerate sphere, 315 Dehn, M., 290
407, 410,
De
Paolis, R., 362
Describe, 401
Diagonals of a quadrangle, 72 Diameter, of a conic, 73; end of a quadrilateral, 81 Dickson, L. E., 35, 339, 341
Differential of arc, 366 Dilation, 95, 348
of,
151;
Direct 451 Direct Direct Direct
collineation, 61, 64, 65, 107, 438, of a quadric, 200 projectivitics, 37, 38, 407
similarity transformations, 135 transformations, 225, 452 Directed, oppositely, 433 similarly, 433 Directed broken line, 484 Directed interval, 484 Directed polygon, 484
;
Edges, of a broken line, 454 of a polyhedron, oriented, 495 Eisenhart, L. P., 368 Elementary transformations, 409 411414, 418, 419, 421, 423, 430, 43 1', 434437, 447, 455, 45 restricted, 410, 414, 420, 430 Elements, complex, 156 imaginary, 7, 156, 182 ideal, 71, 287 improper, 71 Eleven-point conic, 82 area of, 150 foci of, Ellipse, 73, 140 189 imaginary, 187 Elliptic congruence, 443 Elliptic displacements, parameter representation of, 377 Elliptic geometry, double, 375 Elliptic geometry of three dimensions, 373 Elliptic pencils of circles, 242 Elliptic plane, 371; double, 375; single, 371, 375 Elliptic plane geometry, 371 Elliptic points, 373 Elliptic polar systems, 218
;
Elliptic projectivity, 5, 171 Elliptic transformations, direct circular,
248
Ernch, A., 230 End of a diameter, 151 Ends of a segment or interval, Euriques, F., 302
Directed segment, 484 Directed three-dimensional convex region, 484 Directed two-dimensional convex region, 484 Direction-class, 433 Direction cosines, 314 Directions, bundle of, 436 pencil of, 433 Director circle, 200 Directrices of a skew involution or line reflection, 258
;
45,
427
Envelope of lines, 406 Equation of a conic, 202, 208 Equations of the affine and Euclidean groups, 116, 135, 305; linearly independent, 406 and matrices, modular, 464
;
Directrix, of a circle, 192; of a conic, 191 ; of a parabola, 193
Equiaffine collineations, 105 Equiaffine group, 105, 291 Equianharmonic cross ratio or set of points, 259
Equidistantial curves, 356 Equilateral hyperbola, 169
Disagree (sense-classes), 485 Displacement, 123, 129, 138, 143, 297, 317, 325, 352, 369, 373 parameter representation of, 344 parameter representation of elliptic, 377 parameter
; ;
representation of hyperbolic, 380 types of hyperbolic, 355 Distance, 147, 157, 311, 364, 373 algeof translabraic formulas for, 365 unit of, 147 tion, 325, 327
;
Equivalence, of ordered point triads, of ordered tetrads, 290 96, 288, 290 with respect to a group, 89 Euclid, 360
;
Doehlemann, K., 229, 230 Double elliptic plane, 375 Double elliptic plane geometry, 375
Doublepoints of projectivities, 5,114,177
Doubly oriented
449
line, 440, 442, 445, 447,
Doubly perspective, 448 Down, 303
Euclidean classification of conies, 186, 210 Euclidean geometry, 117, 118, 119, 135, assumptions for, 144, 287, 300, 302 59, 144, 302 ; as a limiting case of noil-Euclidean, 375 Euclidean group, 117, 118, 135, 144; equations of, 116, 135, 305 Euclidean line, 58 Euclidean plane, 58, 60-63, 71 and complex line, correspondence between, 222, 238 inversion group in the real, 225 sense in. 61
;
; ; :
504
Even polygons,
470, projective plane, 489 482, 483
INDEX
482, 489; in the
Gravity, center of, 94
Even polyhedra,
Group, affine, 71, 72, 287, 305; conjugacy under, 39; continuous, 406; of
displacements, 129 equiaffine, 105, 291 equivalence with respect to, 39 Euclidean, 116, 117, 1 18, 135, 144, 305 geometry corresponding to, 70, 71, 78, homothetic, 95 inver199, 285, 302
;
Expansion, 348 Extension, assumptions of, 2 of a conic, Exterior, of an angle, 432 171, 174, 17(3 of a polygon, 472 of an even polygon, 490 of a quadric, 344 Exterior bisector, 179
;
; ; ;
Faces of a polyhedron, 474 oriented, 495 Family of points, continuous, 404 Family of sets of points, continuous, 405 Family of transformations, continuous, 406 Fano, G., 11, 285, 286 Feuerbach, 169, 233 Field, Galois, 35 Fine, H. B., 3, 18 Finzel, A., 369 Focal involution, 195 Focal properties of collineations, 201 Foci of an ellipse or hyperbola, 189 Focus, of a circle, 192 of a conic, 191 of a parabola, 193 Follow, 13, 37, 47, 48 Forward, 303 Foundations, of complex geometry, 29 of general projective geometry, 1 Fubini, G., 362 Function plane, 268
; ;
; ;
Euclidean plane, 225, one-parameter continuous, 406 parabolic metric, and geometry, 119, the projective, of 130, 135, 144, 293 a quadric, 259 special linear, 291 subgroups of the affine, 116 Groups, algebraic formulas for certain equations of parabolic metric, 135 the affine and Euclidean, 116, 135,
sion, in the real
226
305
Half turn, 299, 370 Half twist, 324 Halstead, G.B., 361
Harael, G., 28
Hamilton,
W. B.,
339
Harmonic homology, 257 Harmonic separation, 45 Harmonic sequence, 10,

point of, 10 Hatton, J. L. S., 168 Heath, T. L., 360 Heine, E., 60 Hermitiaa forms, 362 Hesse, 0., 284
Hilbert, D., 103, 181, 394
33, 34
limit
Functions, trigonometric, 154
Fundamental circles, 254 Fundamental theorem of

for a chain, 22
projectivity
Homology, harmonic, 257 Homothetic group, 95 Homothetic transformations, 95

Horocycle, 356 Horosphere, 370 Huntington, E. V., 3, 33 foci Hyperbola, 73 equilateral, 169 of, 189 rectangular, 169 Hyperbolic direct circular transformations, 248 Hyperbolic displacements, parameter representation of, 380 types of, 355 Hyperbolic geometry, of three dimenand inversion plane, 357 sions, 369 Hyperbolic lines, 350 Hyperbolic metric geometry in a plane, 350 Hyperbolic pencils of circles, 242 Hyperbolic plane, 350 Hyperbolic points, 350 Hyperbolic projectivity, 5, 171 Hyperbolic space, 369
;
Galois field, 35
Gauss, C.F., 40, 361 Generalization, by inversion, 231 projection, 167, 231 Geometrical order, 46 Geometries, projective, 36
by
Geometry,
assumpafflne, 72, 147, 287 tions for Euclidean, 59, 144, 302 complex, 6, 29 ; corresponding to a group, double ellip70, 71, 78, 199, 285, 302 tic, 375; elliptic, 371 ; Euclidean, 117, 118, 119, 135, 144, 287, 300, 302; Euclidean, as a limiting case of nonfoundations of genEuclidean, 375 eral projective, 1 ; generalized, 285 history of non-Euclidean, 360 hy: ;
;
perbolic plane, 350 inversion, 219 inversion plane and hyperbolic, 357 modular, 253 of nearness, 303 nonEuclidean, 350; parabolic metric
;
Imagining ellipse, 187 Imaginary one-dimensional form, 156 Imaginary lines, conjugate, 281, 282, 444 Ima :uv dints, 8, 156 Imaginary sphere, rotations of, 335 Improper elements, 71 Incomplete symbol, 41 Independence of assumptions, 23; proofs of, 24-29
i
Left-handed Clifford parallels, 374, 444 Left-handed conjugate imaginary lines,

444
Left-handed doubly oriented

445
lines, 442,
Left-handed elliptic congruence, 444 Left-handed ordered pentads of points,

442
Left-handed ordered tetrad of points,

442
Infinity, circle at, 293
line at, 58, 71
plane at, 287; points at, 71, 241, 268, 287, 352 space at, 58 Inside, of a conic, 171 of a qnadric, 344 of a conic, Interior, of an angle, 432 of an interval or seg171, 174, 176 of a polygon, 472 of an ment, 45 even polygon, 490 of a quadric, 344 of a triangle, 389
; ;
; ;
Left-handed regulus, 443 Left-handed sense-class, 407, 416 Left-handed triad of skew lines, 443, 447 Left-handed twist, 417, 443
of a circle, 148 Lennes, N. J., 18, 457 Lewis, G.N., 96, 138, 362 Lie, S., 341 Like sense-classes of segments, 436, 437 Limit point of harmonic sequence, 10 Limiting points of pencils of circles, 159 Lindemann, P., 366, 368, 369 Line, of centers, 159 complex, 8 doubly oriented, 440, 442, 445, 447, 449 Euclidean, 58, 60 hyperbolic, 350 ideal, 287, 350; imaginary, 166;
Length
Interior bisector, 179
Intermediate positions, 407 Interval, 45, 46, 47, 60, 456 directed, 484 ends of, 45, 427 oriented, 484 Intervals, complementary, 46
; ;
;
Intuitional description of the protective plane, 67 Invariant subgroup, 39, 78, 106, 124 Invariants of a conic section, 207
Inverse matrix, 308 Inverse points, 162

Inversion, 162, 241, 266 generalization in a complex plane, 235 by, 231 Inversion geometry, 219 real, 241, 268 Inversion group in the real Euclidean plane, 225, 226 Inversion plane, 268; complex, 264, 265; hyperbolic geometry and, 357; real, 241 Inversor, Peaucellier, 229 Involution, absolute, 119; focal, 195; order relations with respect to, 45 orthogonal, 119 skew, 258 axes and directrices of skew, 258 Involutoric collineations, 257 Involutoric projectivities, products of pairs of, 277 Involutoric rotation, 299 Irrational points, 17, 21 Isogonality, 231 Isomorphic, 3
; ;
at infinity, 58, 71 ordinary, 71, 287, sides ; oriented, 426 ; real, 156 of, 59, 392 ; similarly oriented with, translation parallel respect to, 426
;
350
to,
288
;
Line Line
pairs, measure of, 163 direcreflections, 109, 115, 258 trices, or axes of, 258 orthogonal,
;
Isotropic Isotropic Isotropic Isotropic
lines, 120, 125, 265,
294
plane, 294
rotation, 299 translation, 317
Jordan, C., 453 Juel, C., 250, 251

Klein, F.. 71, 249, 278, 284, 285, 361, 362, 374, 375, 446
120, 122, 126, 299, 317, 362, 370 Linear convex regions, 47 Linear group, special, 291 Linearly dependent circles, 256 Linearly dependent solutions of E x 488 Linearly independent columns of E 2 488 Linearly independent equations (Hj), 466 Lines, broken, 454 congruence of, 275, 383 conjugate imaginary, 281, 282, 444 crossings of pairs of, 276 envelope of, 406 ideal minimal, 265 meetings of pairs of, 276 minimal or isotropic, 120, 125, 265, 294 ; negative pairs of, 417 ordinary minimal, 265 orthogonal, 120, 138, 293, 350, 352; pairs of, 50, 163; parallel, 72, 287, 351 perpendicular, 120, 138, 293, 369, 373; positive pairs of, 417; singular, 235 elementary transformations of triads of skew, 447; right- and lefthanded triads of skew, 443 subdivision of a plane by, 51-53, 460-464 ; vanishing, 86
,
Logarithmic
spirals, 249
Lower side of a cut, Loxodromic direct

mations, 248
Liiroth, J., 9
14 circular
transfor-
MacGregor, H. H., 260 Magnitude of a vector,

Malfatti, G., 236
86, 147
Manning, H.
P.,
362
;
;
modular Matrices, algebra of, 333 equations and, 464 sum of two, 333 Matrices E, and E 2 for the projective
plane, 484 Matrices H 19 2 and H 8 396, 398-400, 477 Matrix, inverse, 308 orthogonal, 308 rank of, 478 scalar, 334 Measure, of angles, 151, 153, 163 angular, 163, 165, 231, 311, 313, 362, 365 of line pairs, 163 of ordered tetrads, 290 of ordered point triads, 99, 312 of a simple n-point, 104 of triangles, unit of, 99, 140, 319 99, 149, 312 Median of a triangle, 80 Meetings of pairs of lines, 276 Menelaus, 89 Metric group and geometry, parabolic, 119, 130, 135, 144, 293 Metric properties of conies, 81 Mid-point, 80, 125 Milne, J. J., 168 Minimal lines, 120, 125, 265, 294 Minimal planes, 294 Minimal rotation, 299 Minimal translation, 317 Minkowski, H., 394 MObius, A. F., 40, 67, 104, 229, 252, 292, 293 'Model for projective plane, 67 Modular equations and matrices, 464 Modular spaces, 33, 35, 36, 253 Moore, E. H., 24, 35 Moore, K. L., 59 Morley, P., 222 Motion, rigid, 144, 297 screw, 320 Moved, 406
Negative twist, 417 Negative of a vector, 84 Negatively oriented curve, 452 Negatively related sense-classes, 485, 491, 495 Net of rationality, 35 cuts in, 14 order in, 13 Neutral throw, 246 Nine-point circle, 169, 233 Nine-point conic, 82 Noncollinear points, 96 Nondegenerate circle, 266 Nondegenerate sphere, 315 Non-Euclidean geometry, 360 Euclidean geometry as a limiting case of, 375 history of, 360 Nonmodular spaces, 34 Normal to a conic, 173 Normal curve, 286 Null vector, 83 Numbered angle, 154 Numbered point, 456 Numbered ray, 154
;
;
;
Numbers, complex, 219
Odd Odd
polygons, in a plane, 470, 482 the projective plane, 489 polyhedra, 482, 483
in
On, 440 One-dimensional form, imaginary, 156 order in, 46 real, 156 One-dimensional projectivities, 156, 170173; and quaternions, 339; represented by points, 342 One-sided polygonal regions, 490 One-sided polyhedra, 493 One-sided region, 437 Open cut, 14 Opposite, 433
;
^"-dimensional chain, 250 ^-dimensional segment, 401 JV-dimensional space, 58 ^"-dimensions, generalization Nearness, geometry of, 303
to,
304
Negative ordered pairs of lines, 417, 418 Negative of an oriented segment or region, 485 Negative points, 17 Negative relations between points and segments, 485 Negative rotations, 417 Negative sense-class, 407, 416
of a transformations, 452 452 Oppositely directed, 433 Oppositely oriented, 448, 450 Oppositely sensed, 245 Order, 40 assumptions of, 32 geometrical, 46 in a linear convex region, 47 in a net of rationality, 13 in any one-dimensional form, 46 on a polygon, 456 of a set of rays, 432 Order relations, on complex lines, 437 in a Euclidean plane, 138 in the real inversion plane, 244 with respect to involutions, 45 Ordered pair, of points, 268, 271 ; of rays, 139 Ordered projnctive spaces, 32 Ordinary lines, 71, 287, 350
:
Opposite Opposite Opposite Opposite Opposite

2-cell,
collineations in space, 438, 451

projectivities, 37, 88 to a ray, 48
sense, 61
Ordinary planes, 287 Ordinary points, 71, 265, 268, 287, 350 Ordinary space, 58
Orientation of space, 496 Oriented, oppositely, 448, 460 similarly, 448 similarly, with respect to a line,
; ;
426 Oriented broken line, 484 Oriented 2-cell, 452 Oriented 3-cell, 453 Oriented curve, 452 Oriented edges of a polyhedron, 495 Oriented faces of a polyhedron, 495 Oriented interval, 484 Oriented line, 426 doubly, 440, 442, 445, 447, 449 Oriented points, 426 segments of, 426 Oriented polygon, 484 Oriented polyhedron, 495 Oriented projective space, 453 Oriented segment, 484 Oriented segment or region, negative of, 485 Oriented simple surface, 463 Oriented three-dimensional convex region, 484 Oriented two-dimensional convex region, 484
;
;
119, 130, 135, 144, 293 Parabolic pencils of circles, 242 Parabolic projectivities, 5, 171 Parabolic direct circular transformations, 248 Parallel to a line, translation, 288 Parallel lines, 72, 287, 351 Parallel planes, 287 Parallelogram, 72
Parallels, Clifford, 374, 375, 377, 444 Parameter of a conic, 198 continuous one-parameter family of sets of points,
;
Origin of a ray, 48
Orthogonal circles, 161 Orthogonal involutions, 119 Orthogonal line reflections, 120, 122, center of, 126, 299, 317, 352, 370
;
122
pairs of, 126
Orthogonal lines, 120, 138, 293, 350, 352 Orthogonal matrix, 308 Orthogonal plane reflections, 296 Orthogonal planes, 293 Orthogonal points, 352 Orthogonal polar system, 293 Orthogonal projection, 313 Orthogonal transformations, 308 Outside of a conic, 171 Outside of a quadric, 344 Owens, P. W., 59, 371
Padoa, A., 44
Pairs of lines, 50, 163 crossing of, 276 measure of, 163 meetings of, 276
;
405 continuous one-parameter family continuous of transformations, 406 one-parameter group, 406 Parameter representation, 344 of ellipof hyperbolic tic displacements, 377 displacements, 380 of parabolic displacements, 344 Paratactic, 374 Pascal, E., 186, 235, 279, 280 Path curve, 249, 356, 406 Peaucellier inversor, 229 Peirce, B., 341 center of, Pencil, base point of, 242 of lines, of directions, 433 429, 433 correspondence between the real a and Euclidean plane complex, 238 of segments, 433 of rays, 429 Pencils of circles, 157, 159, 242 limiting points of, 159 Pencils of projectivities, 343 Pentads of noncollinear points, right and left-handed, 442 Permutations, even and odd, 41 Perpendicular bisector, 123 foot of a perpendicular, 123 Perpendicular lines, 120, 138, 293, 369, 373 Perpendicular planes, 293, 369, 873 Perpendicular points, 352, 369, 373 Perspective, doubly, 448 Perspective correspondence, 271 Pieri, M., 244 Pierpont, J., 3 Planar convex regions, 386 Planar region, 404 Plane, of analysis, 268 complex, 154 complex inversion, 264-268; corre;
negative ordered, 418 positive ordered, 417 separation of plane by, 50 Pairs, of orthogonal line reflections, 126 of points, ordered, 268, of planes, 50 271 of points, unordered, 271 Paolis, E. De, 362 Pappus, 5, 103, 118 directrix axis of, 193 Parabola, 73 focus of, 193 Steiner, 196; of, 193 vertex of, 193
negative, 417
positive, 417
;
spondence between a complex line and the real Euclidean, 222, 238 double elliptic, 375 elliptic, 371 Euclidean, 58-63, 71 function, 268 hyperbolic, 350 hyperbolic geometry at and inversion, 357 ideal, 287
; ; ; ; ;
;
287 intuitional description of the projective, 67 inversion, 268 inversion group in the complex Euinversion group in clidean, 235 the real Euclidean, 225, 236; iso294 minimal, 294 model for
infinity,
;
; ; ;
tropic,
a order relations order relations in the real inversion, 244 ordinary, 287 orthogonal, 293 projective, 208 real, real inversion, 241, 208 140, 156 sense in reflections, orthogonal, 295 sides of, 59, 392 a Euclidean, 61 single elliptic, 371, 375; subdivision of a plane by lines, 51, 53, 400-404 of symmetry, 295 Planes, pairs of, 50 parallel, 287 perpendicular, 293, 369, 373 subdivision of space by, 50, 54, 475-477; vanishing, 348 Pliicker, J., 292, 326 Poiucare', H., 341, 362, 489 Point pairs, congruence of parallel, 80 mid-point of, 80 separation of, 44-47 Point-plane reflection, 257 Point reflection, 92, 122, 300, 352, 414 Point triads, measure of ordered, 99 equivalence of ordered, 96, 288, 290 sum of ordered, 96 circular, 120, Points, complex, 8, 156 155; double, of a projectivity, 5, 114, 177 equianliannonic elliptic, 373 set of, 259 hyperbolic, 350 ideal, imaginary, 8, 71, 265, 268, 287, 350 156; at infinity, 71, 241, 268, 287, 352 inverse, 162 irrational, 17, 21 negative, 17; noncollinear, 96 numone-dimensional projecbered, 456 ordered tivities represented by, 342 pairs of, 268, 271 ordinary, 71, 265, 268, 287, 350 oriented, 426 ; orthogoof a pencil, base, 242 of nal, 352 pencils of circles, limiting, 159 ; perpendicular, 352, 369, 373 positive, 17 projection of a set of, 291; rational, 17; real, 8,156; rotations represented by, 342, 343 ; segments of oriented, 426 singular, 235 in space, correspondence between the rotations and in space, cross ratios of, 55 the, 328 unordered pairs ultra-infinite, 352 of, 271 vanishing, 86 Polar coordinates, 249 Polar system, 215 absolute, 293, 373 elliptic, 218 orthogonal, 293 Polygon, 454-459, 480, 481 ; bounding, 470, 482 directed, 484 even, 470, interior and exterior of, 482, 489 472, 490; odd, 470, 482, 489; order on, 456 oriented, 484 regions determined by, 467 ; sum modulo 2 of 481 unicursal, 470 Polygonal regions, 473 ; one- and twosided, 490 Polyhedra, odd and even, 482, 483 ; oneand two-sided, 493 oriented, 495 oriented edges of, 495 ; oriented faces sum modulo 2 of. 482 of, 495
projective, 67
;
Euclidean, 138
Polyhedral regions, 4M conbilateral, 494 Polyhedron, 474 nectivity of, 475 edges of, 474 faces oriented edges of, 474 one-sided, 494 and faces of, 495 two-sided, 494, 490 unilateral, 494; vertices of, 474 Positions, intermediate, 407 Positive coordinate system, 407, 408, 416 Positives ordered pairs of lines, 417 Positive pairs of lines, 417
;
; ; ;
Positive points, 17 Positive relation between points and oriented segments, 485 Positive rotation, 417 Positive sense-class, 40, 407, 416, 491 Positive translation, 416 Positive twist, 417 Positively oriented curve, 452 Positively related sense-classes, 485, 491,
495
Power,
;
of a point with respect to a circle, 162 of a transformation, 87, 230 Precede, 13, 15, 37, 47, 48, 350, 387 Product, of pairs of involutoric projecof two vectors, 220 tivities, 277
_
Projection, generalization by, 167, 231 orthogonal, 313; of a set of points,

;
291 Projective classification of conies, 186 Projective correspondence, 272 foundations Projective geometry, 36
;
of general, 1 Projective group of a quadric, 259 intuitional deProjective plane, 268 matrices E x and E 2 scription of, 67
; ;
for,
484
;
Projective space, collineations in a real, 252 sense in, 64 senseProjective spaces, ordered, 32
;
classes in,
418
Projectivities, bundle of, 342 ; cyclic, 258 direct, 37, 38, 407 ; double points of, 5, 114, 177; elliptic, 5, 171 ; hyperbolic, 5, 171 ; one-dimensional, 170, 171 ; opposite, 37, 38 ; parabolic, 5,
;
171 pencil of, 343 powers of, 87 products of pairs of involutoric, 277 of a quadric, 273 real, 156, 1701 73 representation by points of onedimensional, 342 representation by quaternions of one-dimensional, 339
;
; ; ; ;
;
Projectivity, assumption of, 2 Prolongation of a segment, 48
Proofs, independence, 24-29
Quadrangle, diagonals
;
of, 72
;
axes of, Quadrics, absolute, 369, 373 316 conf ocal system of, 348 direct
; ;
collineations of, 260 interior and exterior of, 344 ; projective group of.
259: -Droiectivities of. 273: real. 262;
ruled, 259 sides of, 344 sphere and other, 315 unruled, 259 Quadrilateral, diameter of, 81 Quaternions, 337-341, 378 and the onedimensional projective group, 389
; ; ;
;
Ricordi, E., 360 Riemann, B., 361 Right angles, 153
Radical axis, 159 Radii, transformation by reciprocal,

102 H 2 479 of a matrix, 478 Rational curve, 286 Rational modular space, 35, 36 Rational points, 17 Rationality, net of, 35 order in a net
Rank, of
liight-handed Clifford parallels, 374, 444 Right-handed conj ugate imaginary lines, 444 Right-handed coordinate system, 408, 416 Right-handed doubly oriented lines, 442, 445 Right-handed elliptic congruence, 444
Right-handed ordered pentad of points,

442
Right-handed ordered tetrad of points,

442
of, 13 Ratios of collinear vectors, 85 Rays, 48, GO, 143, 350, 372, 387, 429; bundle of, 435; numbered, 154; oppoorder of a set of, 432 site, 48 ordered pair of, 139 origin of, 48 Real and imaginary elements and transformations, 156 Real inversion geometry, 241 Real inversion plane, 241, 268; order relations in, 244 Real line, 156 Real one-dimensional form, 156 Real plane, 140, 156 Real points, 8, 156 Real projective transformations, 156 Real quadrics, 262
; ; ;
Right-handed -regulus, 443 Right-handed sense-class, 40, 407, 416,

442
Right-handed triad of skew

447
lines, 443,
Right-handed twist, 417, 443 Rigid motion, 144, 297 Rodrigues, O., 330
Rotation, angle of, 325, 327 ; axis of, 299 center of a, .122 involutoric, 299
; ;
;
isotropic, 299 minimal, 299 ; nega417; positive, 417; sense of, 142 Rotations, 122, 128, 141, 299, 321, 328tive,
337 ; correspondence between the points of space and, 328 of an imagrepresented by inary sphere, 335
;
points, 342, 343
Reals, geometry of, 140
Ruled quadric, 259

Ruler-and-compass constructions, 180 Russell, B., 41 Russell, J. W., 168, 201
Reciprocal radii, transformation by, 162 Rectangle, 123 Rectangular coordinates, 811
Rectangular hyperbola, 169 Reflections, axes of line, 258 ; center of directrices of orthogonal line, 122 line, 258 line, 109, 115, 258 orthogo; ; ;
Same
Saccheri, G., 361 sense, 61 Scalar matrix, 334

Schilling, M., 67
nal line, 120, 122, 126, 299, 317, 352, 370 orthogonal plane, 295 pairs of orthogonal line, 126 point, 92, 122, in a 300, 352, 414 point-plane, 257 three-chain, 284
; ;
; ; ;
Region, convex, 385-804 ; negative of an oriented segment or, 485 onesided, 437; order in a linear convex, 47; planar, 404; polygonal, 473; polyhedral, 473 sense in overlapping convex, 424 simply connected threedimensional, 404 tetrahedral, 54, 398, 399 ; three-dimensional, 404 triangular, 53, 389, 395 trihedral, 307; two-sided region, 437 vertices of a triangular, 53 Regions, bounded by a polyhedron, 483 determined by a polygon, 467 Regulus, right- and left-handed, 443 Restricted elementary transformations, 410, 414, 420, 430 Reye, T., 168
; ;
Schweitzer, A. R., 32, 415 Screw motion, 320 Segment, 45, 46, 47, 60, 350 or interdirected, val, complementary, 46 interior of 484 ends of, 45, 427 interval or, 45 n-dimensional, 401 oriented, 484 prolongation of, 248 Segments, bundle of, 436 of oriented
; ;
points, 426; pencil of, 433; classes of, 436, 437
sense-
Segre, C., 9, 250, 251 Self -conjugate subgroup, 39, 78, 106, 124 Sense, 32, 41, 61, 387, 413; clockwise, 40 criteria for, 49 in a Euclidean in Euclidean spaces, 63 plane, 61 more general in a linear region, 47 theory of, 451 in a one-dimensional opposite, 61 ; in overform, 40, 43
;
;
lapping convex regions, 424 positive, in a projective space, 40, 407. 416 64; richt-hanrled, 40, 407, 416, 442;
; ;
491 right-handed, 40, 407, 416, 442 Sense-classes, agree or disagree, 485; negatively related, 485, 491, 495; positively related, 485, 491, 495 ; in projective space, 418 Sensed, oppositely, 245 similarly, 245 Separated, 392, 432 Separation, algebraic criteria of, 55 harmonic, 45 ; by pairs of lines, 51 by pairs of planes, 51 of point pairs,
;
tionof, 496; oriented projective, 453; polygons in, 480, 481 rational modusense in a Euclidean, 63 lar, 35, 30 sense in a projective, 64 sense-classes in projective, 418 singular, 58
; ;
;
Spatial convex regions, 386 Special linear group, 291 Sphere, center of, 315 degenerate, 315 and other quadrics, 315 ; rotations of
;
an imaginary, 335
Spirals, logarithmic,
44,47 Sequence, harmonic, 10, 33, 34; limit point of, 10 Sets of points, connected, 404, 405 continuous family of, 405 Shear, simple, 112, 293
;
249
Square, 125 Statements, vacuous, 24 Staudfc, K. G. C. von, 9, 40, 251, 283

Steiner, J., 196, 229 Steinite, E., 35, 69 Stephanos, C., 286, 324, 342, 344 Study, E., 40, 327, 341, 347, 362, 374, 416, 446
11., 168 Subdivision, of a plane
Sheffers, G., 341 Sides, of a circle, 245 of a plane, 59, 392
of a line, 59, 392 of a quadric, 344

;
Similar, 119
Sturm,
Similar and similarly placed, 95 Similar figures, 293 Similar triangles, 134, 139 Similarity transformations, 117, 119, 293 ; direct, 135 Similarly directed, 433 Similarly oriented, 448 with respect to aline, 426
;
by
lines, 51-53,
460-464 475-477 Subgroup,
of
a space by planes,
50, 54,
Similarly sensed, 245 Similitude, center of, 162, 163 Simple broken line, 454 Simple curve, 401
or invariant, 39, 78, 106, 124 Subgroups of the affine group, 116 Sum, modulo 2, of polygons, 481 modulo 2, of polyhedra, 482 of ordered point-triads, 96 of two anof two matrices, 333 of gles, 154 two vectors, 83
; ;
self -con jugate
Surface, simple, 404

;
Simple polygon, 454-457 Simple shear, 112, 293 Simple surface, 404 oriented, 453 Simplex, 401 Simply connected element of surface, 404 Simply connected surface, 404 Simply connected three-dimensional region, 404 Singular lines, 235 Singular points, 236
;
404 404 Symbol, incomplete, 41 Symmetric, 124, 297, 300, 352
simply connected, simply connected element of,

;
Symmetry,
123, 124, 129, 138, 297, 300, 352, 373 plane of, 295 ; with respect to a point, 300
;
Tait, P. G., S41
directrices or ; 258 elementary transformations of triads of, 447; right- and lefthanded triads of, 443 Smith, H. J. S., 201 Sommerville, D. M. Y., 362 Space, assumptions for a Euclidean, 59 ; collineations in a real projective, 252 correspondence between the rotations and the points of. 328 cross ratio of points in, 55; direct collineations
Singular space, 58 Skew involutions, 258
axes
of,
Skew
Taylor, C., 168 Taylor, W. W., 82 Tetracyclic coordinates, 253, 254, 255 Tetrad, measure of an ordered, 290 of points, right- and left-handed
;
lines,
ordered, 442 Tetrads, equivalence of ordered, 290 Tetrahedral region, 54, 398, 399
Tetrahedron, 52, 397 volume of, 290, 311 Three-dimensional affine geometry, 287 Three-dimensional convex region, 386 Three-dimensional directed convex region, 484 Three-dimensional elliptic geometry, 373
;
Three-dimensional Euclidean geometry, 287 Three-dimensional hyperbolic geometry, 309 Three-dimensional region, 404 simply connected, 404 Throws, 40 neutral, 245 similarly or oppositely sensed, 245 Torsion, coefficient of, 489 Touch, 158 Transference, the principle of, 284 Transformations, of a 2-cell, circular, 225; continuous family of, 406 direct, direct similarity, 185 225, 452 elliptic direct circular, 248 elementary,
; ;
left-handed, 417, 443
negative, 417
positive, 417; right-handed, 417, 443
Two-dimensional convex region, 380 directed, 484 Two-sided polygonal regions, 490 Two-sided polyhedra, 493 Two-sided polyhedron, 494, 496 Two-sided region, 437
Ultra-infinite points, 352
Uriicursal polygons, 470 Unilateral polyhedron, 494 Unit of distance, 147
Unit triangle, 99, 149, 312 Unit vector, 220
409, 411-414, 418, 419, 421, 423, 430, 431, 434-437, 447, 455, 450 ; homotheloxodromic direct circular, tic, 95 248 opposite, 452 orthogonal, 308 parabolic direct circular, 248 power of a projective, 87 power of a circular, 230 ; real and imaginary, 156; by reciprocal radii, 1G2 ; restricted ele; ; ; ; ; ;
Unordered pairs of points, 271 Unruled quadric, 25'J

Up, 303
Upper
side of a cut, 14
Vacuous statements, 24
Vailati, G., 44
mentary, 410, 414, 420,430 117, 119, 293

; ;
similarity,
Translation, 74, 117, 122, 288, 321, 374, 414 axis of, 317 distance of, 325,
Vanishing lines, 8(5 Vanishing plant's, 348 Vanishing points, 86 Vector, magnitude of,
tive of, 84 83, 220
;
80,
147
nega-
null, 83; unit, 220; zero,
327;
isotropic,
;
negative, 410
positive, 416
minimal, 317; parallel to a line, 288
317;
Transposition, 41 Transversals, 91 Triads of lines, elementary transformations of, 447; right-handed
and
left-
handed, 443
Triangle, area of, 149, 312 interior of, 389 measure of, 99, 149, 312 median separation of a plane by, 52 ; of, 80
;
Vectors, 82, 83, 85, 147, 219, 288 addition of, 84 collinear, 84 product of two, 220 ratios of coiiinear, 85 sum of, 83 Vertex, of a conic, 191 of a parabola, 193 Vertices, of a broken line, 464 of a polyhedron, 474 of a triangular region., 53 Volume, 290, 311
; ; ;
unit, 99, 149, 312
Triangles, congruent, 134, 139

134, 139
similar,
Whitehead, A. N., 32, 41 Wiener, H., 94, 280, 322, 327

Wilson, E. B., 96, 113, 138, 362
Triangular region,
of,
53, 389,
396
vertices
53
Trigonometric functions, 154 Trihedral regions, 397 Turn, half, 299, 370 Twist, 320, 321 axis of, 320 ; half, 324
;
Young, Young,
J. J.
W.,
iii,
250
W.
A., 146
Zermelo, E., 27 Zero vector, 83, 220
PROJECTIVE GEOMETRY,
Norbert Wiener expressed
1964 by M.
his
Volumes
and
II
Oswald Veblen and John Wesley Young

admiration for Protective Geometry
in his autobiography Ex-Prodigy (paperback edition published in

I.
T. Press) as follows:
"Professor Huntington ha'd
recommended
to
me two mathe-
matical books for

Russell.
summer
reading before starting
my work
with
They were
Young's Projective heart as the most consistent exposition of the postulational stand. . .
Modern Algebra and Veblen and The second book I took to my Geometry.
Bocher's
I
point that
had found anywhere.
worked out almost
all
the
problems of the first volume, which was the only one existing at the time the personality of the book was chiefly that of Professor
.
Oswald Veblen
of Princeton.
He
*nathematical school of Princeton as well

of the Institute for
was the founder of the great as" the scientific founder

also at Princeton.
Advanced Study,
He
is
without doubt one of the fathers of American mathematics."

4
Other Blaisdell Books
PROJECTIVE GEOMETRY,
by H. S. M. Coxeter. A synthetic treatment of general projective geometry in which relations of incidence and projective transformations are stressed, (clothbound)
9
$5.00
BASIC CONCEPTS OF GEOMETRY,

Meyer Jordan. This
by Walter Prenowitz and
is a presentation of geometry as a branch of contemporary mathematics involving interrelated study of specific
geometrical systems, characterized by properly chosen postulate systems, and which, within the framework of such, gives a treat-'
ment
of Euclidean geometry, (clothbound)
$7.50

Veblen, Youn - Projective Geometry

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Veblen, Youn - Projective Geometry

Загружено:

Авторское право:

Доступные форматы

PROJECTIVI

Oswald Veblen and John Wesley Youn

JOHN WESLEY YOUNG

BLAISDELL PUBLISHING COMPANY DIVISION OF GINN AND COMPANY

(the properly projective spaces)

Among the spaces satismay be mentioned the

spaces, the rational

space, the real space,

the complex space.

in the complex projective plane

of all-continuous one-to-one reciprocal transformations

groups connected with Euclidean geometry

example, conic sections) to which cross references could be made.

*12. Chains in general *13. Consistency, categoricalness,

ELEMENTARY THEOREMS ON ORDER

Direct and opposite projectivities on a line The two sense-classes on a line

Sense in any one-dimensional form

Segments and intervals

24. Algebraic criteria of sense 25. Pairs of lines and of planes

and the tetrahedron

27. Algebraic criteria of separation.

28. Euclidean spaces

*32. Sense in a projective space

THE AFFINE GROUP IN THE PLANE

The geometry corresponding

36. Parallel lines

39. Self -conjugate

Congruence of parallel point pairs Metric properties of conies 42. Vectors

Extension of the definition of congruence

The homothetic group

Measure of ordered point The equiaffine group

*51. Algebraic formula for measure.

*52. Line reflections

Geometries of the Euclidean type Orthogonal lines

Displacements and symmetries. Congruence

58. Pairs of orthogonal line reflections

65. Intersectional properties of circles

The Euclidean geometry.

Area The measure of angles The complex plane

One-dimensional projectivities Interior and exterior of a conic

77. 78. 79.

186 189 193 196

82. Eccentricity of a conic

Synoptic remarks on conic sections Focal properties of collineations

Euclidean classification of point conies

88. Classification of line conies

*89. Polar systems

Vectors and complex numbers

real inversion plane

97. 98. 99.

244 246 250 253 267

100. Tetracyclic coordinates 101. Involutoric collineations

imaginary lines of the second kind

AFFINE AND EUCLIDEAN GEOMETRY OF THREE DIMENSIONS

112. Vectors, equivalence of point triads, etc

The parabolic metric group. Orthogonal

114. Orthogonal plane reflections 115. Displacements 116. Euclidean

and symmetries. Congruence

n dimensions the affine and Euclidean groups

119. Distance, area,

The sphere and other quadrics

122. Rotation, translation, twist

Hyperbolic metric geometry in the plane

12. Chains in general 13. Consistency, categoricalness,

169. Oriented points in a plane 170. Pencils of rays