Professional Documents
Culture Documents
GEOMETRY
HUNT
PROJECTIVE GEOMETRY
Projective
BY
Geometry
II
VOLUME
OSWALD VEBLEN
OSWALD VEBLEN
Late Professor of Mathematics, Princeton University
London
A,
1L
A,
Ji,
J1
A, &, f
or A,
Ji,
r,
fying A, E,
modular
nonmodular
of
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.
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
we
find
among
others,
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
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
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
...
23
29 32 33 36
CHAPTER
II
19. 20.
37 40
43
21.
44
45 47
22.
23.
49
50 52
The
triangle
....
55
58 59
61
Assumptions for a Euclidean space 30. Sense in a Euclidean plane *31. Sense in Euclidean spaces
29.
63
64
CHAPTER
SECTION
34.
III
group of transformations
...
70
71
35.
affine
group
72
parabola
73 74
The group
of translations
subgroups.
Congruence
78
80
81 82
85
Theorems
of Menelaus, Ceva,
and Carnot
89
92 94
Point reflections
46.
47.
95
96
triads
48.
49. 50.
99
106
Barycentric coordinates
106
109 115
116
CHAPTER IV
EUCLIDEAN PLANE GEOMETRY
55.
56.
57.
119
120
123 126 129
131
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
66.
67. 68.
69.
70.
set of
assumptions
144
147
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
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
Homogeneous quadratic equations in three variables 86. Nonhomogeneous quadratic equations in two variables
85. 87.
202 208
210
212
215
CHAPTEE VI
INVERSION GEOMETRY AND RELATED TOPICS
90.
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
Order relations in the real inversion plane Types of circular transformations Chains and antiprojectivities
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
CHAPTER
SECTION
111. Afflne
VII
287
288
and planes
293
295
297 301
geometry
*117. Generalization to
118. Equations of
304
305
311
315
121. Resolution of a
....
317 321
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.
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
364
365
366 369
371
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
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
388 392
396
397
155.
156. 157.
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
423
424
425
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.
*181.
182.
183.
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
H H
z,
and
JET 8
477
479
SECTION
192.
193.
PAGK
480
482 483 484 489 490 493 496
601
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.
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
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
a theory
ourselves
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
(cf. p.
16, Vol.
I).
ASSUMPTIONS OF ALIGNMENT:
A
loth
line
1. If A and B A and B.
is
on
A 2.
If
A and B
A and
C
(D
C,
is
on both
B. are points
A3. If
not
all
A, B,
the
on
same
=
line,
and
and
E]
are
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
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
line
invariant,
leaves every
ASSUMPTION
H
fl
As 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.
of
completing the
list
:
of
K.
geometric
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
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
ever, incomplete.
there given
cit.)
is,
howholds,
if
that
if
but
it is
is satisfied.
THEOEEM
1.
Assumption
is
valid in
and E and such that multiplication Assumptions. in a geometric number system (Chap. VI, Vol. I).
Proof. It
is
commutative
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
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
sequence
of perspeetivities
where
Chap.
[Jf]
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.
By
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
H HH
y
J^
is fixed, S,
V,
it is
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.
HH
of
PT;
Q
let
RT;
and H'
is
TRSP
determines
I)
(JfQ
HH
v
ir
in the scale
lf
Hm H H'}
v
ffx
Since multiplication
is
-Hv = H'.
commutative,
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
4.
Double points of
DEFINITION.
A projective trans-
was proved in
58, Vol.
px
i
= ax + ox
1
'
of the
equation
f-( A = ad
be.
where
its
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
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-
order
tions of order
is
to be
found in Chap.
IY
tiva of Enriques.
EXERCISE
A>
:
is
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)
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
we
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
it is
its
One way of founding this geometry is to replace Assumption by another assumption of au equally summary character, namely,
J.
is
isomorpMc with
the
complex num-
Since this
geometry
Assumption
P,
and
all
for the
com-
A, E,
of
The problem
projectivity
finding the double points of a one-dimensional completely solvable in the complex geometry for
;
may
,
ax
* o
v,
and therefore
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.
an
to
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
+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
By
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
of
view
of a general space
and
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
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
.
set of all
,
complex points
,
of the
form
KX^,
KXZ
KX^j
where XQ
points S r
tions of
x^,
x 2 xs are
,
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.
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
is
projective space
Sc
Obviously S
may
where
also
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.
knowledge or study
fact,
of the
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
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
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,
reals.
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
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+
= ITi+1
is
called a
is
is
The
dently parabolic to sr O im
and
carries
THEOREM
2.
gard
to the
other two.
.
Proof.
By
57,8]
HARMONIC SEQUENCE
to
11
the
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
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
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.
The
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
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,
and E.
If
on a line
projection
I,
let a line
I
/'
meet them
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
is
m then a"* +
= am
is
the identity,
period, there
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
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,
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
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
H-%> H-fr
5)
H-v
-&
-#i>
projective
(fig.
with
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
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"
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
r>
Two
subsets, [A]
B]
nd only
3
if
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]
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]
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~],
* 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
by the coordinates
of
The
sets [A]
and
[/*]
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
D)
If a cut
the
(A,
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
D} and
(C,
D) precedes a
cut
(JE,
7'
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
is
said to be algebraic
if
a xn
a^z"-
+
,
an
,
0,
with integral
all
coefficients,
\_A~\
points of
between
AQ
and
all
B make
the left-hand
member
of this
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,
I,
for refer-
ence as Assumption Q.
ASSUMPTION Q. There
is
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
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
which precedes
all
preceded by
all
other S's.
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
symbols
ail d
z>,)
ASSUMPTION
C.
If every
I
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)
^A,S->
If two open
cuts (A,
B) and
(C,
D) are
distinct,
the points
and
J5
and
(C^,
C2 any
)
open
of
R^^H^H^}, and
if
is
a projec-
tivity
such that
associated with some cut
then
(A v
(D^
2)
between
lows
.f
_//
is
called positive,
is
called neyativi'.
,
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
^
\Gy
where d
the coordinate of D.
which follow
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
is
THEOREM
9.
Hx
>
0,
-7^=
1,
#"
the real
number system of
I,
analysis.
The
operations of addition and In that place we derived all the fundamental laws of operation, except
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
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
+*
b
condition
in
C^/^-frJ, suppose
first
and
an irrational point.
that a
^(H^H^H^)
(4)
x*=x+a
of points [ajj into
it.
a],
which
is
the
same
+ zj.
,
Similarly,
(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
+ a,
and
which defines
,
yz +
a.
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
(^ 4-
2,
y^ 4-
y2)
is
& 4- a.
Thus we have
identified
geometric addition
points in
^(H^H^H^
be stated as follows
cuts (x v
y/j)
If
a and
let
I are positive
and
the
(,/;,
[x[]
Then ab
tive
is
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
= 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
[aajj, a&,
and
[a?/,],
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/= :(!)
.
Hence, by Assumption
C (2),
By
tive
Hence,
C^H^), (l)x = x(
on Assumptions
,
C(H H H^}.
1).
it is
to be
if
rememis
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).
THEOREM
into
10.
Any
projectivity
Q,
chain
(^(H^H^)
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
-'-to
'.
a point of
C^H^).
I,
it
follows that
Moreover, any
of
ormation
'ransforms
>,
C (HflJlJ
Q)
H^
II
and
II',
the prod-
liemselves.
Hence, by
mid be the
me
'ct of
ging
Q,
projecfcivities of
'~"r
o}.
Then
PHP-
leaves ff , o
H^
&
n)
invariant
andhence
shown
theorem.
Hence
2.
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
* 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
(X) = X'
is
independent
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
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
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)
is
chain.
a unique
the
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
Any
into
chain
C(ABC)
one
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^H^ABC.
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
ito points of
ictivity
C(QPA)
into
C(PQR).
In
like
manner C(QPA)
= C(PBA) =
COROLLARY.
cut in
any
THEOREM
any
have
A CHAIN. If A, B,
(A, B, C,
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 ,
we
D'=D' r
H^ be
the two pro j activities mentioned in the theorem. C(ABC) fixed; for it leaves every
n^
point of
R(ABC)
fixed,
(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
Any
Our
of reals
ASSUMPTION
R.
On
follows, by
we have
THEOREM
deducible
14.
from Assumptions A,
all
theorems
Let us
now
duction (Vol.
I) to
apply the logical canons explained in the Introthe foregoing set of assumptions.
C,
(%
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^
if
one
other
members
is
are given
8)
,
all
8 ),
where
any
real
number not
zero, shall
be called a point.
tetrads satisfy
Any
two
whose component
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
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.
I.
Since
is
a logical consequence of A, E,
H,
C,
is
number
homo-
Now
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
if
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
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,
* 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
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.
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
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
Hence 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
,
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.
pseudo-points can be found identical with an ordinary harmonic sequence and net of rationality on any line not passing through 0.
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
A3
is
true
1,
E 2, E 3, E
H,
C,
E
false
1.
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
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.
lines of
and
16,
72, Vol.
I,
and
below).
and
are true,
is false,
and C and
shall
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
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
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
_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
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
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
we
H H
coordinates
'R(H^H1H^)
(js,
consists of
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
of the
two open
x
1
numnumber
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
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 ,
number
The ordinary
that
is,
aft
/3's
+ ^a. 4 /^ +
+ pm akm
:
where the
are rational.
numbers
by the
The laws of combination for the number may now be stated as follows Express the two
;
add or multiply
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
ao 4
a^+ a a'^----z
{
(-
an a(n
new law
of multiplication,
which we
shall denote
by x
is
now
a''s
according to
a!
4 a^ H---- 4 #<
whenever
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
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
We may
number systems
.
selecting the
= V2
then choosing the a' 'a so that a[ = a z * 14. Foundations of the complex geometry. Let us add to Assumptions A, E,
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
P 1 X) C(PP
However,
x' = = or
it is
clear,
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,
two
cases, to
Theorem
4,
J be
any point
of
).
[X] be the
set of
all points in
PP C(P Q 1
a>
Then, by Theorem 11
30
FOUNDATIONS
is
[CHAP.
[X+J]
with
a chain.
l
a3
P n
in
commo]
C(PPP ),
because,
m X+J=X'J= P
it
X'
a point of
C(PP P
Let us denot
the chain
by
C'.
In order
we need
tion of closure
ASSUMPTION
I.
Through a point
I
of any chain
C
C.
of the line
J on
P in common with
r>
Now
1
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
Let
he
this poin
In case
(12)
X^ P
the projectivity
-X-
-X)
transforms
forms
itself,
and I
into itself.
Hence
it
tram
into
is
C^JT^), and
in particular P,
If
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.
is expressible
in the form
and
2.
LEMMA
in
1
A, B, A!,
at
PP X C(P Q
For
if
')
are identical if
and only if
A' and
1
B = B'.
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 )
transof
into
C(P (P -J)I^.
C(P P I^
1
be any point
where A"
is also in I) it
>
).
By
5,
follows that
A-AJ=A"-JA".
By
(14), this reduces to
2, it
follows that
Hence
AJ=JA. From
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.
of multiplication.
Hence,
is
satisfied in
any space
satisfying
number system
is
expressible
A + JB, we B
are in
have
J^A^JB^
and
where
C (J?J?^,). Thus J is
of the involution
(16)
XX'- %B (X+X')-A =
X C(PP P
1
0,
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>
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
J-=-P.
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
C
is
there is
a,
sense
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. Camhrid^e "IQOB
f
With
II.
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,
S 7 reduce
to
II.
This
real
satisfied
by a rational or a
is
shown
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
of points, every
We have seen ( 7) that, in any and E, any two harmonic sequences one harmonic sequence contains an infinity
spaces.
of points,
and
by
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:
exists, at least
one con-
ing
H nonmodular.
Theorem
5,
It follows, just as in
is
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
than
n.
m = k,
we have
and hence
rP = 1.
the harmonic sequence are contained in the set
Hence
In case k
integers,
is
is, if
k 1}
different
=k
&a
let
us con-
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,
when we
arrive at a
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 )> '"' ^o' LL (
where II
is
projectivity is
With
reference to a scale in
which Jf
is oo,
= x-\-l.
harmonic sequence are
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
I,
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-
...,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, E,
A, E,
A, E,
H H
Q Q
in
t
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, E, H, C, E,
or A, E,
I\
J
f is
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
to
CHAPTER
II
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
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
to be
coefficient of
is
positive
if
and only
if
precedes B.
(2)
the form
/3,
where a
of the
is positive,
This group
is
a subgroup
group
P x
group contains
The group
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
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
and afterwards
23) show
how
tions of
The
definitions for the general case, like those for the special one, will
may
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
by multiplying them
2
of
the othf
this merel
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 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
<
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}
,
Q Q Qw
,
yjy^ =
fy(Q<a Q
2/0
=
0.
(4)
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
Hence
it
follows that if XQ
and
P^, P*,,
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
R
Q>
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
x[),
where
(x' , x[)
= S(x
,
Q>
xj.
The
are (y
coordi,
QQ Q Q x
I}
T(xQ
ajj)
and
Q , y[) (y'
= T(x'
y^
x[} respectively.
Hence, by substitution,
of T.
The determinant
of
the transformation
TST~
is
where
as A.
Thus the
is
important in
DEFINITION.
lent
Two
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
any
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
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
to be proved.
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.
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
same sense-class
are said
Tw
sent
same
sense.
Two
sam
One
sense-class
chosen arbitrarily
may
as riyht'handed, clockwise, positive, etc.f The term " sense," standing by itself,
"
The
senses are
any
set of objects
is
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
* "
two senses
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
dx
dx
THEOREM
then
ABCis in
the sense-class
S(AHC)=ti(A
o ).
If S(AIi(,')
= 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(ABC) = S(A"B"C").
0"
C'
B'
ABC
FIG. 7
A"
B"
A'
Proof.
opposite.
of
If
S(ABC)^
S(A'B'C'}, the projectivity ABC-^A'B'C' is follows from the fact that the product
two opposite
projectivities is direct.
sense-classes
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.
that
S(ACB}^X(BCA}.
'
It follows,
= 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
is
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-
and only
> 0.
form
of
(1,
1), is
(1, 1
+ a).
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
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
is
said
to
if
and only
if
S(ABC]
of
= S(A'
is
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
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,
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
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.
DEFINITION.
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-^A'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
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
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
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
they separate
C and
D.
II
By Theorem
II
7 (2)
AD\\BC.
imply
we have
(1)
and
BD \\ACy
THEOREM
An
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
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.
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
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.
contain
its ends.
2.
in
ABC,
then
COROLLARY
3.
If
is
in
ABC,
then
and D
and
C.
THEOREM
of which
Proof.
nionically.
10.
If A and
line, thi
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
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.
AB
are said
COROLLARY. If A, B,
line consists
are
any
and
line,
ABC, BCA, CA
one
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
+1
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.
(i
= l,--,
n), is
set
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
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
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
A region is
said
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
AB.
set of points of
n}
if
{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
S(ABA^)
is
S(ABB
ai
}>
and
the segment
AA^B is
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"),
then
* S(A"B"},
then
If S(AB}
* S(BA). = S(BG],
then
S(AB)
= S(AC).
Theorem 7
:
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
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
2. If three points of
the order
3.
a convex region are in the order {ABC}, they are in in the order {ACS} or {CAB}.
4.
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 l^Uul
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,
(a
,
(>
a a ),
B = (b
Q,
6 t ),
and
(7=
(C Q , Cj)
are
any three
(6)
changes
only
if
and
(1, 1)
into A, B, and
respectively
if
and
that
is, if
Pi
With
is of
By
is
&i)
direct if
and only
c[]
if
is positive.
Now
A'
= (a[,
a[],
B'
= (&
',
C'
= (c'
Q
are
of
=
If S' is of the
possible.
same sign
as S, the projectivities
which
(7) (8)
(1,0) (0,1)
(1,0)
(1,1)^^,
l^A'B'C'
projectivity in
(0, 1)(1,
which
is direct.
(7)
and
direct (8) is
is
,
opposite.
Hence
(a' , a[],
THEOREM
'
Let
'
A = (a
a,j,
J5
= (6
bj,
C = (o
^ vj
A'
'
C Q , c[) be
collinear points.
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
respectively, if
1
(c
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,
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
and b'
respectively if
and only if a
and
C
Q
a'
b liaw
th same
sign.
Proof. Set a
aja^
bjb^
THEOREM
17. A,
is
separate C,
D
B
if
and onfy
if
the
cross ratio
B (AS, CD)
Proof.
negative.
By
separate
C,
JD
if
and only
d
if
and
are opposite in sign.
k
these
same sign as Tb(AB, CD] (cf. p. With the aid of this theorem the proof much more simply than in 21.
the
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.
and
of planes.
THEOREM
a and
/3
OP
Proof.
By
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 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
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
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
S l on
/j
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
52 on
:
(Theorem
6).
The
^ and
fall
S2>
those by which triads in the sense S i are projected into and those by which triads in the sense S^ are projected
Sy
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.
THEOREM
line
If a
line
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
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
meets them.
IC^B by
7,
&
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
a/3<y, afiy,
all
three
segments
in
any
of
a Pj-
The lines of the plane, exclusive of those through the vertices, therefore fall into four classes:
(1) (2) (3) (4)
/3,
7,
7,
7,
/?, 7.
No two
1?
lines
1
l^,
of the
same
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 /
to the vertices.
is
perhaps more
intuitively striking
when put
:
THEOREM
20.
The points of
sides
of a
four
classes
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 L^I^ meets the sides of the triangle.
sides
DEFINITION.
is
Any
one
of
the four classes of points in Theorem 20 The vertices of the triangle are also
of
Theorem 19 can
also ser
+ l)-poinb
I}
in ?t-space.
,
teti
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}
be a plane not passij the six segments which it meets may be denoti
<;
"
>
tlie
%,
-,
Then
as a corollary of
Theorem 19 we have
that any plane
which
nonco-
meets
three
u'
all
>
^
the
-,
meets
the rest of
no plane meets
^2- If
all
we observe that
FIG. 17
it
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.
DEFINITION.
is called
Any
one
of
a tetrahedral region.
HJ.IU
uu
juuve uuau
is
tiie
uuuuuxuy
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
The
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
,
,
(cf.
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,
the points A^
A A A
2, a>
into
Y into
(12)
s)
is
carried
by
this projectivity.
By
substituting in (11)
<>, '
we
see that
\*
x,
if
~~
i, V
,
if
34
From
a,
a 8 &J,
\a^, %,
taken as the homogeneous coordinates with respect reference whose vertices are A A 2 , A^.
I}
to the
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
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 A^A^A^A.^ and are separated By Theorem 17, the points
by the planes
and A^A^A^ if and only if &14 is negative. They will be separated by A^A^A^ and A^A^A Z if and only if is negative, and by 04 & A^AyA.^ and A^A 2 A S if and only if 3i is negative. Hence, by Theorem 21,
3
/fc
A 2 A A^
we have
THEOREM
J-
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
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
70, Vol.
I),
we
(x)
see that
if
ax
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)
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),
=
,
0,
N
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
,y 4-
a 3z
and denominator
state this as a
THEOREM
Theorem 18,
system
23.
Ity
The two
classes
the singular
-\-
and a plane ax
by
}-
+d=
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
space*
of
on a
siuj
(n
1)
of n dimensions.
but one
of
and the
set
on a single
line,
called a
Euclidean plane.
DEFINITION.
(n
1)
(n
1) -space
S"
is
-space or the (n
V)-spi
Any
figure in S
is
said to
?i-sp
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
and
B are
"
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
COROLLARY. If a
of a are either on
In like
I
is
or on the
a point of
I,
all poii
manner Theorem 18
25.
THEOREM
a plane
TT
three-space
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
DEFINITION. The two classes of points determined by a line / in a to Theorem 24, are called the two sides
The two
by a plane
TT
in a Euclidean
sides of
TT.
three-space, according to
Theorem
two
The two
sides of
TT
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
The rays
if the rays coincide, it is called a zero angle. k are called the sides of hk, and their common origin the
;
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
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
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.
by
F.
actions of the
American Mathematical
Society, Vol.
I.
W. Owens, XI (1910),
Transp.
141.
Compare
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
AB consists
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]
{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
C not
in
any of
the orders
points collinear with pairs of points on the intervals called the plane ABC.
AB, BC, CA
is
VII. If A, B,
not in the
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
pair of points A,
C such
that if
is
any
infinite set
of segments of
AC, having
AC
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.
which
effects a
direct projectivity
on 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
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
called a sense-class
and
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.
sense-classes in a Euclidean
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'.
S(AC) = S (A'B'C').
The
triads to
which
ABC
is carried
direct all
form a sense-class,
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,
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)
hence
not direct
and hence
*S (ACS).
D
are on opposite sides of a
=
S(ABC)
8(ABD).
25
can also be derived from the following algebraic considerations. Let us choose a system of nonhomogeneous coordinates in su
is
the sam
The group
of all pr<
itself
reduces
67, Vol. I) to
(15)
If
=0.
to the
we change
homogeneous
coordinates for
which x
and y
= xjx^
= 0,
and
(16)
_ =
/yi
>
4-
4i
On
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
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 ^e same
sign.
b.
1 1
and
c[
b>
c' z
as a corollary of
Theorem
23,
27.
EXERCISES
2.
(^45(7) S^A'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.
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
;
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
*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
projectivities.
is
Nevertheless, the
basis of the
algebraic criterion
definition
A>
made the
which
follows.
With
of
reference to a
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
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
is
the
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
DEFINITION.
The group
which
the
determinant a
>0
is
called the
lt
A A
a,
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
A, B,
C,
be
a z a & ),
,
(6^
2,
(c^
z,
8 ),
(d v d^,
(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-
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.
of all
(20)
for
^=i/y^ =
J
(=<),...,)
from
zero.
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
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
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
which
>
is
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
can be transformed by direct collineations. If the initial points are to be transformed respectively into the points whose coordinates are the
(21)
the collineation
(22)
where the
/o's
+/ i a ol +
)
(23)
W" + P ^ n + P
this
determinant
is
(a^a 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
) (
which
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,
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
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
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
and
is
6T
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
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
10.
Since any
tri-
triangular region
may
be represented
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
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
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.
O would
ORST move
is
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
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
chapter,
deformation
of
complete
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,
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.*
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
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-
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
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
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
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
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
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
all
projectivities
18).
The group
(1) -will
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
.
TT.
line at infinity.
The
points of 1 M shall be
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
affine collineation
transform-
B,C
is
of a triangle
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/^ap +
'
where
A=
DEFINITION.
=0.
Two
ordinary lines not meeting
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
which
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
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
parallel,
parallel to its
3.
is
With
by
is
ix
+
4.
%+c
0.
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
itself.
DEFINITION.
A conic meeting L
ellipse.
n two
me
in
no point an
The
Ellipse
Parabola
Hyperbola
FIG. 22
pole of
is
The tangents
intersection with
Any line through the hyperbola at its points asymptotes. A conic having an
to a
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
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
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
into itself.*
15.
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
said to be parallel to
I.
COROLLARY.
line
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
AB respectively.
AM*
the
to
1
The
translation carrying
A to A'
line
Hence which
tion of
is is
A'
the
intersection of
A'M^withBL*. Hence
the points A' and B' to
which
tively
and
respec-
<
are
carried
by
FIG. 23
ABB A'
1
is
a parallelogram.
translation carrying
to A', the
and only one same reasoning shows that whenSince there is one
ever
ABB' A'
is
and
to A'
5.
THEOREM
An
AB,
if
and
only if
Q(L X AA',
L^B'B),
of AB.
Proof.
Lx
being
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.
f.lip
f.r ft
n si
ft fvi
nn
r>.flrrviTor
A to B
of
carries
P
t
to Q,
QNa
with AB.
Hence N^,
and 5'
are collinear,
and hence we
THEOREM
6.
If A, B,
C
to
translations carrying
and
to
is the
translation carrying
to C.
Proof,
Let
Ax
B^,
Ca
BC, GA,
A,,,
,
AB
respectively
with
lw .
Suppose
first
Bm Cm are all distinct. The the line ABX into the line BBX
,
translation carrying
and the
.
translation carrying
B to
CBn
Hence
the line
ABm
is
invariant
two
translations.
Consider
line
now any
other
let it
f ' t
through
Bx
and
OmC
be the point of
of
intersection
BC (fig.
A
25).
We
(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
.
two
;
and
m invariant
that
is,
it
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
Hence we have
COROLLAEY. The
THEOREM
Proof.
7.
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,
,fn^i,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'
T and T
x
are
commutative.
In case B'
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
implies that
also,
T
1
and
= ^',
and
&
THEOREM
is
8.
If
OX and OY are
parallel
is
translation, there is
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
P parallel
OX and Y respectively.
T
x
:,
and T 2 the
6,
translation carrying
Y^
T[
it
follows,
by Theorems 4 and
that
T T 2 =T.
a
On
if
any translation
parallel to
to
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
is the
Proof.
The point
into
which
is (0, 0)
may
be denoted by
By Theorem
(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
THEOREM
in which
the
lw
10.
is the
With respect
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
by a transformation
of
the form
x'
x,
EXERCISE
Investigate the subgroups of the group of translations.
Congruence. DEFINITION.
to be
Any
2
sub-
group under
G G
of
if
a group
and
self-conjugate or invariant*
is
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
Fio. 26
3]
CONGRUENCE
J
79
nd 2T2~
.gures
_/''/
(/<'/)
l^.
Therefore,
if
G'
is
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-
ided that
f
is
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
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,
I)
2T2-
is
translation.
COROLLARY
1.
The group of
the affine
translations
is
self-conjugate
it.
under
ny subgroup of
COROLLARY
here exists
2.
For any
affine collineation 2,
T,
2T=T'2 and
translation
T"
uch that
Proof.
v/
T2 = 2T".
Let
2T2~ = T'
a
and 2~ a T2
= T". By
J
are translations.
But
= T'
nply
DEFINITION.
and 2-
2T = T'2
Two
and
T2 = T" T2 = 2T"
respectively.
if
The
figure consisting c
two
to
distinct points A, J3
may
be looked at in two
ways with
respec
congruence.
We
and
AB
to
and
BA
AB
the
second case
AB
BA mean
BA
On
ordered pair
lation carrying
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(J^AC, P,DB]
point pair
Proof.
is satisfied is
is
an
AB
congruent
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
AB with respet
A and
B.
Then
the
AO
case
is
congruent
to the
pair OB.
DEFINITION.
point
of
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
be the points at infinity of the lines AB an B^ and Cl be the mid-points of the pairs A.
of "mid-point,"
Hence the
and
lines
B C are
BQ,
that BJ.
parallel.
DEFINITION.
The
of a triangle
ABC
the mid-point of
BC is
The
called a
median
of the triangle.
THEOREM
13.
three
40, 4ij
CONGRUENCE
Let the triangle be
81
,
Proof.
ABC]
let
A a Bx
;
AB
and
(7(7,,
AA X
and
BBX respectively
sets
(
(fig.
27).
harmonic
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
parallel to
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
Let us
line
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
as mid-point.
contains a
number
of theorems on conies
82
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
HA
Z
is con-
The point
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
X^X^
The
The
A BCD
are
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
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
ordered pair of points determines a set of pairs it under the group of translations. In
order to study the relations between such sets of pairs we introduce The term " vector " appears in the literature
A conic is said to
if
" the term " ordered pair of points is to be understood to include the case of a single point counted twice.
DEFINITION.
A planar field
which are
(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,
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.
vector corresponding to the point pair is a definite object, and we shall denote it as
AB
or,
in symbols,
Vect (AB).
Since
any point
of
a Eu-
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,
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
is
that
0.
Vect (AB)
+ Vect (BA) =
Hence we
define
DEFINITION".
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,
is associative;
Proo/.
definition,
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,
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
OC, where
is
such that
a,
vector
00 such
AB.
that
Q^AO,
J^BC), where
is the
point at infinity
of
the line
M be
is
two points
such that
OBML
a parallelogram.
Then,
by
definition,
= Vect
is,
such
is
ACML
be
the
parallelogram. Let
LM
ideal
OL and BM,
(7.
and
let
be the ideal
AL and If
to
any
in which P^
is
and
A+B = C.
Theorem
1,
Cf. Cor. 1,
I.
17.
b
vectors is
commutative ; that
are vectors, a
+b=b
-f-
Theorem
15,
and
if
case of addition
we should he
vectors
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
--
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
Theorem
now
gives
THEOREM
If A
s>
A
A
THEOREM
sides
20.
"^2"'^8
4)
AB, BO, CA
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
line
5^
is
parallel to
<?
BC. Thus,
if
Ba
be the point at
AB, and
AC,
_ = _ = _,
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
an involution transforming
the involution.
THEOREM
21.
DEFINITION. If
and
43]
VANISHING LINES
I' ,
87
I,
parallel* line
to tin
and
X is
a variable point of
which
1
X is
OX
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
;
'
X^
18,
of
2
= __I.
of vectors,
magnitude
O'X!,
OX
COROLLARY
COROLLARY
point
the
1.
= OA\
O'X/.
center
27t0
and
a conjugate pair
2.
AA^
OA
be
is
OA^.
Let II
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
Proof.
Let
Q and
which the
line
FP meets the
vanishing lines
n~
By
the theorem,
FP = QF
P'Q''
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'
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
4. Six points
(/1 2
# 2 C2 J 1 B 1 C'
,
1)
if
and only
if
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
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
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
FIG. 30
Proof.
of
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
is
now
obvious.
THEOREM 23
and
sufficient condition
,
Suppose
first
that
C"
is
an ordinary point.
of
Menelaus,
(5)
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
is
parallel to
AB
and,
by Theorem 20,
(7)
I.
The condition that C" be harmonically conjugate to C' with regard to and B now takes the form C'A
-
C'B
On
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
B C
Z,
respectively.
The formula
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
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
(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.
AB
5,
2,
and
1,
23.
EXERCISES
1.
The
A, B,
A^A^B^Ry C^CZ on
and only
2.
if
AB
'
of a triangle are
tangent to a conic in
Av
Bv
C, respectively,
3. If a line .BC*
meets a conic in
and
C, respectively,
meet
it
^B
respectively,
==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
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
A A V A^A V
,
A nA
respectively, of a simpL
polygon A^A-^A^
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
respectively, in th<
^^o AtC A
9. If a conic is
A
3
AB
Ci
A Bn A Q Cn
,
A A V A^A Z
A nA
respectively,
ii
the points
Bv
Bn respectively,
00
AiBo
A 2 B,
A Bn
nn =
'
DEFINITION.
reflection.
n is called
a.
point
From
The
THEOREM
25.
point
reflection is
fully determined
center is the
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.
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^
all
reflections is
translation.
THEOREM
one of which
Proof,
tion,
27.
is
Any
translation
is the product
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.
reflections
a point reflection.
Proof.
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
.
P
l
such that
p r2 r p
l
'
'
'
p ^Zn +
pp rjr 2n + p2n +
l jr
p *
a,
COROLLARY
is
2.
point reflection
a point
reflection.
3.
COROLLARY
a,
of
all
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.
of point
reflections
and
translations
is
THEOREM
29.
With
respect
to
coordi-
nates in which
I* is the
singular
.
equations of a point
reflection
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.
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.
group
of
reflections.
4.
reflections
P p P2
7l
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
of translations
and point
is
reflections.
This definition
It will be completed in
IV.
The main
it
significance of the
of
is
that
Pn
The center of T is the " center of gravity " of the cenCf H.Wiener, Berichte der Gesellschaft der Wissenschaften zu
.
THEOREM
30.
Any
AB
is
congruent
to
the
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
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
is called
imothetic transformations.
letic
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
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.
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
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
book
of Euclid's
we
shall
work
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
word
" area."
The
triads of points
which
are
Our
triangle
definitions
ABC is
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-
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
such
t
lt
and t^t', or
(4) if
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
that
t'.
is
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.
ty^ABC^ABC.
the form of the definition
34.
35.
From the last two theorems and from we now have at once
THEOREM THEOREM
,
the mid-point
of
pair
AB
then
AOC^OBC.
Proof.
Let
is
changed
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
AOC sc OBC' *= BC'O = BC'M + BMO OB C = OB M + OMC. reflection with M as center carries OMC
into
C'MB.
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
the
.
C2 C3
By Theorems 34 and
definition,
35,
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
^O'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
--B'
and T(<7)
<7'.
If
'
is
be the intersection
(fig.
parallel to
line
PS
is
on the
PQ,
let.
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
on PQ,
it
follows,
by the
36,
By
r
Theorem 36
the
line
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
The theorems
In order to
c,
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
OPQ
as a unit is a
number m(ABC)
deter-
mined
as follows
If
the line
BC is
and C1
to
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
OP
magnitude
of the vector
B^C^
*
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.
IfABC^A'B'C
1 ,
then
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
C^A 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 A^A 2 BB
lt
,
are parallel,
it
follows from
43
that
BC
Hence
or
*
44 ^
m(ABC]
=AA ^
l
V=
1
Cl
The
factor \
is
we
In case
BC is
parallel to
OP, the
last clause of
[n case
CA
is
parallel to
OP,
AB
and
B C are
But,
by
definition,
ELence
(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
m
et
the line
BC (fig. 39).
;he line
neets
;he
BC, and
let
B G D
I}
I}
be
points in
which the
lines
through B, C,
parallel to
respectively,
m(ABiy)
is parallel
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
IN
THE PLANE
and
[CHAP,
ill
THEOREM
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).
OQ (fig.
m(ABD)
But,
by
43,
In case
BC is
parallel to
OP,
let
(ABC)
m (ABD)
COROLLAEY
1.
BC SA ~ = BC ~~ = m(BCA] = ~
-
'
m (BDA)
BD SA
BD
IfB,
C,
collinear
with a point A,
AC ,AC,
COEOLLABY
2.
If B,
C,
D are points no
is the
P^
B AC AC>
'
THEOREM
Proof.
40. If
m(ABC} =
m(A'B'C'}=i-
0,
then ABC^A'B'C'.
By Theorem 37 there exists a point C" on the line A'C' such that AB C ^A'B'C", Hence A'B' C' ^A'B' C", and by the last
theorem, C'
= C".
of
In consequence
the last
triad
may
to
to
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
C',
parallel to
By Theorem
line
38,
m(ABC} = m(ABC'},
AB.
C[.
Hence the
CC'
is
parallel to
42.
THEOREM
the line
AC, then
BC'
is
parallel
to the line
if
Proof.
By
C"
is
a point of
AC' such
that
BC"
is parallel
ABC^AB'C".
41 the only points C such that ABC^AB'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
Vol.
When
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
CA'
parallel
to
C'A.
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
upon
elemen-
The notion
points,
i.e.
of
of
11
(cf.
14, Vol. I) to
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.
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.
C,
2.
C,
Av Az
An
the number
is
all
We
it
define it to
be the measure
of
A^A 2
An
and denote
by
m(A l A z
An
~).
6.
m(A A Z
1
'
A n)
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'
respec-
tively parallel,
'
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
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
a consequence of the
triads
THEOREM
If an
affine collineation
it
triad into an equivalent point triad, triad into an equivalent point triad.
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'
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
is
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
P^
of
Since
m(ABC] = m(A''C'),
it
follows that
ADC^A'D'C*.
DEFINITION.
Any
affine collineation
subgroup of the
Proof.
affine
group.
By
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
B and A
B';
a
B'.
be
line
is
on
! ,
but not on B,
and
let a' be
on
There
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,
ABC^A'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
ABO^A'B'C'.
Hence (Theorem
44) there
EXERCISE
Any
affine collineation leaves invariant the ratio of
the measures
of
any
two point
triads.
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
=
In case the line
{
0.
3s
2,
the point
l
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.
case
BC
is
parallel to
OP
is
eadily verified.
Now
(11)
f
if
C
ft
respectively
by a
ransformation
/lf
A=
(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}
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(\,
27
x y
I 1
0=
x
c
i
y
C
2
in a system for
is also equiaffi
{L^}
is equiaffine.
THEOREM
reflections.
49.
An
equiaffine collineation is
a product of two
Proof.
Let
If
there be any pc
which
,
is
be such a point.
A A A
z
,
By
the hypothesis on
the points
equiaffine
A A A
Q>
I}
are uoncolliuear,
and
is
AA4>-A^M-^A
,
Hence, by Theorem 41, the line A A 3 is parallel to A^A Z or else A of the pair A. be the mid-point of the pair A^A Z and Let z i
Let L^ be the point at infinity of the line A A Z> L z of the line A^A^ 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,
at infinity of
A^M^ and
mid-point of
M^M By
.
and
Hence
But
since
{Z,y
{
,
it
follows,
by Theorem
1,
that
if
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
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 A^A Z and B^B^ and F is
A^A 2
at infinity
and
let &
be the line
B^
B
l}
and
the point at
infinity of
into
tively,
'
Mj
FI<J _
4g
If
an ideal point,
into 0,
1.
A B
I}
A B
2
,
let be the line A^B^ and let m be the line A^A 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*
for
of A
is either
is
a point
reflection or
at infinity
and whose
an ordinary
Proof.
line.
it
if
the point
is
ordinary,
If
F is
I
and
F = {Om}
;
{01}.
an elation with
as center
to the point
Im
as axis.
112
[CHAP,
DEFINITION.
is
elation
is
whose center
is
at infinity
and whose
axis
an ordinary line
COROLLARY
with
I
If F
is
{/}
{L^,
then
for
every
^ne
concurrent
and
which
ofT
and
M and
r = {Mm}
There also exist a point
{LI}.
M' and a
F
-
line
= {Ll} {M m }.
If
be
taken as variable,
Proof.
choice of
follows
The
first
from the
fact that
The
projectivities follow
theorem for
A A
Q,
z,
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
T=
THEOREM
of
is either
{M'm }.
which are products
this
50.
The
set
of
line reflections
form a
group.
Every transformation of
group
an
an
equiaffine
transformation by a line
Proof.
By Theorems 48 and 49
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
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
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 //
If
M
3.
T=
(
with respect to
if
and
if
{/VH^
Z2
on
and
and only
0,
1
is
on
is
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
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 -^i
= 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
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.
,
If the line
C2 C
5,
parallel.
9.
of equivalence.
10.
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
ax
where
a, 6, c,
d are rational.
The
cz
8
2 ax
-b=
0,
and both
will be rational if
is
rational in
o2
+ be =
k2
Now
of
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
,
ab')_
',
(5c'
'
aa')
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
0,
(1) of the
affine
group become
(13)
2
A=
==
^2^0
'
az
=0.
^2^1
On
the line
ln
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)
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
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
= ctjX 4- 5 y
t
-f-
Cj,
a,
&
= 1.
EXERCISES
1.
The
a self-conjugate subgroup
= a^x +
o^y
bz y
y'=az x +
where k
2.
is
+ +
cv
c2 ,
The
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
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
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
number
systems,
all distinct in
where
A= ^
Z
0.
The
(2)
A>0;
(
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
A = l;
2
A = l,
(
51).
~"
2
1
~~~
'
2'
(47).
7i
ft
C\
rt
'
7i
rt
1
1
'
2J
45).
the group of translations ( 38). The principal groups connected with the Euclidean geometry are
(55
&1
and
62).
Its
similarity transformations.
(11)
al
+&
2 2
0,
a 151
+a &
2
=0,
A>0,
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)
a aa
= 1,
aj.
=&
>
a2
=~&
i
A/fine
Euclidea
A dotted
Homothctic
Displacements
conjugate subgroup.
The fundamental importance of the group of translations is indicated by the fact that it is a self-conjugate
subgroup
of
Translations
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,
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
1;
CHAPTER IV
EUCLIDEAN PLANE GEOMETRY
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
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.
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
we
as geometries of
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,
.
in this
connection
is
explained in a
where the
elliptic
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.
DEFINITION.
Two
be orthogI* in
onal or perpendicular to each other if and only conjugate points of the absolute involution.
they meet
THEOREM
are the
through any point, 0, pairs of an involution. Through any point there is one and
1.
Tlie
pairs of perpendicular
a given
line.
A line perpendicular to
Two
lines
one
is
perpendicular
to the other.
perpen-
same
DEFINITION.
points,
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.
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
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
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
PX and P Y,
:
ti
and
J^X and
1
Y,
POPP.IPQPX, 2 ^
1
and
are
hence
the
points PI
and
X and
P
Q.
Hence
must
X
Y.
Hence
must be
satis-
on any conic
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,
C 2 passes
2.
through them.
THEOREM
An
is
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.
transform en
inf.n ihsfilf
hv
nrth ocron al
reflection
THEOREM
to the axes.
3.
The product of two orthogonal line reflections whose is a translation parallel to any line perpendicular
Proof.
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
Hence
any
on
fc
all lines
through
L^.
Hence
it is
a translation parallel to
4.
through
L^.
THEOREM
line reflections^
is
L is
the center
If
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'
line reflections.
Proof.
If
\=OL'
and
1
2
=
inva-
to
T(0).
{Ll^
3,
Chap. Ill,
If
is
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
two orthogonal
the point
is
THEOREM
5.
A rotation
which
is the
line
Proof.
{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
of
invariant.
of period
two on the
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
AB, BC, CA
of a tri-
meet in a
4.
by a fixed orthogonal
57. Displacements
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
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
(3)
the product of an
of
even number
by an even number
by
definition, a displacement.
The group
T
If {Ll} is
an
2 a similarity
2"1
transformation,
and L'
1
= 2 (L),
= 2 (1),
I'
then
{.LZ}
is
a harmonic
homology
as axis.
But
-
since
at infinity of V.
reflection.
Hence 2
{Ll}
^~ l
{L'l'} is
an orthogonal
1
line
If
similar
n orthogonal
reflections
of
I(
and
is is
displacements
group.
COROLLARY
a
1.
The
set
form
self-conjugate
subgroup of
DEFINITION.
Two
figures
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
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-
of the theory
of
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
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
on this
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
0,
and
reflection
whose
of the line
QP
through
0,
($)
= P.
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.
would
that
in general
have exceptional
we have adopted
III.
the definition
mid-point given in
40,
Chap.
similar
remark
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-
itself.
Two
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.
pairs
BD and A C respectively.
and only
is
58. Pairs of
THEOKEM
is
X
8.
If A x
2,
Aa
a point
line
reflection
AAA
8
2
an orthogonal
passes
through
0.
Proof.
In
case
the
the product
translation,
AA
8
is
and
so
by
in the form
is
an orthogonal line
reflection
whose
to
axis
is
parallel
axes.
the
other
Hence
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,
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
(A^A
Q l is an
arbitrary point of
2
(7
(A^A^l.
The transformation
AAA
3 2
is
1
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?
an orthogonal
line reflection.
1.
COROLLARY
reflections
If
A A
2,
and
are
any
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
jr
In like manner,
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
n^
3,
whose axes
is
_ 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
COROLLARY
COROLLARY
4.
5.
An
The
is
not a displacement.
common
center
is
a,
comimttative group.
Proof.
reflections
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
and hence
(2)
But
since
A A A = \\\,
8
AAMi^VAA
which combined with
(2) gives (1).
THEOREM
9.
Any
invariant
is
rotation about 0.
Proof.
is
AM
Ar
whose
axis
t
is
Then
the
product
T = AA'
Thus
where each T^
if is
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
VA^A:-..^
,
where
Aj,
0.
A^
through
59.
By
Cor. 3,
Theorem
8,
A^
A/
is
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
T' xV
reflection
where
i '(-i
= l,
tt
2 n)
is
an orthogonal line
,
whose
axis
= 1,
2 n)
is
8,
Since
it
invariant, it is a rotation.
1.
COROLLARY
Any
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
it is
clear that
and ^ cannot be
is
parallel,
and hence
In case
PT is T is a
a rotation.
translation
it
whose
center
absolute involution,
Chap. Ill) as
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
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
Cz
is
uniquely determined by
these conditions
(cf.
the
lemma
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
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
of
distinct
from
0,
and
let
through
P having
<7
as center
invo-
displacement leaving
axes meet in 0,
must leave
it
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,
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
OP as A = A' A.
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
is, if
the line
is
whose axis
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
which
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
60. Circles.
DEFINITION.
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.
OP
THEOREM
points.
A circle consists
sec-
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
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 which
OP
as axis
Theorem
0.
8), it
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
no double points,
In
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
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
is
P with
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.
P P v
PP^
Q
and
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^
circle
through
P
G 2
and
must be
I.
In like
Jfw
with
regard to any circle containing P^ and bisector of JJJ* Since the pobts ,
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
by that with
0-ZjJ
as axis
OP
l}
P with
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.
common
center
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
common
10.
of an
point
The
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.
(0, c,
a).
In
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).
W
For the rest of
I~T~
2~~
j
this section
we assume
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
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
(3).
(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
is
satisfy
must
^+^,=
(a^+ 5^) (a^+ \x + (0,^+ & A (a&+ l^ =
2)
)
0,
which
is
2
the same as
2
Z>
(a,
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 - a^i =
2
Thus we
infer
a2 =
5
1
and a t = T
Hence
group are
THEOREM
(6)
19.
The equations of
where
= 1.
If-
Any
(66,
Vol. I)
line
XQ
2.0
_a
Comparing with
(4),
we have
~~
11
22
>
a iz ~
i.e.
If this circle is to
have
(1, 0, 0) as center,
as pole of #
= 0,
the
equation
(7)
must
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
,
is
the form
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
symmetries are
(9)
^'
where a?
Z
-j-
{3
1
I,
and
=< = 1.
(9) effects
By
54, Vol.
if
an
involu-
tion on lx
and only if
1.
By Theorem 10,
Cor. 2,
any symmetry
the
1.
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
having the
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.
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
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
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
the fundamental theorem (Theorem 13) about the group of displacein the following
more
precise
form
THEOREM
the identity.
23.
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
AC
and
the angle
ABC and A'B' C' are congruent if the & CAB are congruent respectively and the angle & C'A'B'.
X
Proof.
A to A'
AC
is
and
con-
AB
to
AB
and since
AC is
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
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-
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
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)
This
may
THEOREM
line
I,
25.
If A^ and
A2
ellipse
are any two points of an ellipse, any in a point between A and A meets ihb i Z
,
by
Let
z
,
and
let
be a
it.
variable
point
on
and
l
be the
I
points
**
in
which
is
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
Then, by the
hypothesis that
is
between
and A
A CA
1
C^
Hence, by Theorem
6,
Chap.
II,
But the
points C.,
L x Q2
,
L x Q8
,
respectively,
by
Hence
the projectivity
is
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 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.
(fig.
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,
Hence
the lines
AD
and
AD
AA
and
AA
2,
which,
according
to the definition in
DJ)^
A^A 2 on the
COROLLARY
sides
2.
of
the line
A^A 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
,
A r By
the
first
X>x
and
Z>
2,
of
I.
THEOREM
A rotation which
transforms a given
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,,
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-^A'^B^C'l
is
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%
(,').
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
EXERCISES
1.
If 21
A OB
is
any
angle,
PR
to
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
are
any two
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
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
given ray
to
28.
There
one
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
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^
separate
P
z
and
the
ordered
triads
P^Q^ and
same
Q^Q
are in the
sense.
inter-
The
involution
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 A^A 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.
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.
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
For our immediate purpose, however, a set of assumptions for Euclidean, plane geometry is needed. To obtain such a set we merely replace
same plane.
Euclidean plane geometry
is
Thus our
I-VI,
vn, ix-xvi.
Assumptions
to
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
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
AS
congruent
to
EF.
is
Proof. This
form a group.
XII. If AB
is
congruent
to
A'B',
and
BC is
to
{ABC} and
Proof.
carries
AC is
B
1
congruent
congruent A'C'.
to
B'C' and
By Theorem
and
28, there is
to
A! and
respectively.
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
BC is
congruent toB'C'.
by the point
reflection
whose
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
having A'B' as
axis.
According to
BC is con;
B'CZ CA to C^A' 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
circles,
G Cz
I}
C'.
Cj,
it
follows,
by Theorem
23, that
D'
=D
AD is
congruent to A'D'. If
C"= C2
are
it follows,
similarly, that
D'=D Z
AD
If
is
congruent to A'D'.
and JF
two points
of a plane a,
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
Proof.
This
is
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
B2
between
A and
the
C. is
Let
Bn be
AB. Then
,
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
8,
Chap. I
provides
in
The
set of
assumptions
I-XVI
is
not categorical.
It
It
can
be be
made
categorical
29.
It
must
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
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
applications of
geometry to
analysis,
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
of a vector
OB
as
its
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
OA is
fixed
and
(7
is
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
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
is
satisfied, for
example, in
affine
The
ratio of
two
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
DEFINITION.
shall
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
from
line,
to Q,
If
1>
is
any
"*'""'
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
}
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
,
C
,
2
,
the numbers
,
p obtained from
do not
Pv P
Pn
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,
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
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
3.
Two
similar in such a
if
C to C'
if
and only
DistQO) _
Dist (A'B')
__ ~
'
4. Relative to a coordinate
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
an ordered point
DEFINITION.
triad,
may
be defined as follows
OPQ
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,
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
continuity which
developed,
and which,
branch
of
like the
theory of
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
of the
theorems about
-
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
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
by
k times
ellipse, therefore,
a(C
a
(
)
2
_ ~
'
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
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
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
the one
69.
The measure
of angles.
The
unit of distance
may be
chosen
arbitrarily, because
any point
abolic metric group into any other point pair. It is otherwise with angles or line pairs, because, for example, an orthogonal line pair can-
pair.
measurement
by choosing
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.
{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
PP, *
meets
is
o-.
And,
= 1,
2,
that P.
Deraendicular to
RP,
The
pair
line
OP
25,
PP, and
the mid-point
between
Theorem
lation
we
re-
{P'P^Jp,
moment,
the
OP
meets the
Since
is
circle.
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
pp
(a
pp
^'^-tJ_''
Let II denote the rotation
leaves
fixed
Ti'-
and transforms
P to P v
OP 4
and
let
transforming
to
P v
The
rotation
H\
as axis followed
OP
as axis, carries the point pair to the point pair OP^. OP^
In like manner
it
follows that
.&"
Now
all
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.
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.
67, (11).
The
number a
to
lie
rr,
where a
= m/2 n
is
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,
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
satisfy a condition of
the form
^ /3 <
7T,
Or~7T</3^7T.
OP m
2"
do not include
all
rays with
as center, the
definition just given does not determine the measures of all angles.
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
is
no
of the circle
(3)
is
an integer,
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
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
lines play
and must be regarded as exceptional in the statement of of theorems; and another large class of theorems of
order relations)
disappears
J.OO
For the
of reals,
so, of
present, therefore,
but shall
(
make
S',
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
;
of projective geometry,
also valid in
there
effect
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
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
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
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
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.
S, a line of S'
which
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.
one-dimensional forms.
DEFINITION.
of S' is said to
Any
element
elements
of
be complex.
Any
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',
and contain
all
form of
S'.
An
element or one-dimen-
sional
form
of S'
which
is
form
of S' is said to
be imaginary.
transforms each real element of
S' into a real
DEFINITION.
said to be real
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
the
4~
first
rtot-i
i^r^~/i
r\
r\
QeidrtT^4-"r*l
I'
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
following section, putting in explicitly, in notation and language, the distinction between elements of S and S'.
It is also convenient in
many
(#2 ,
67-69
y^ and
y^
are
x i~
2)
+ (^i- y
a a)
is
V(^- a^ + ^-y/
to (xz , yz ). Extensions
made when
needed,
C and
2 (72
in a real
,
Euclidean plane.
= ,
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
&
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.
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.
of the last section, contains the double points of the absolute invo*
JT
and
I,
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
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
,
.
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
and
In this
therefore,
Type
I.
cannot have
to
FIG. 56
%Q
and also
of
PQ
2
2,
and
common
center.
Hence
of the involution
its
is
an ordinary
point.
Let a denote
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.
let
in
which
a meets
C*.
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
make
it
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
of a
Two
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
of
is
all circles
point as center
circles is the set
of all
tangent
to
a given
of
a given point ;
of
all
else
I, the pencil
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
double points.
DEFINITION.
The
line
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
if
which the
meet
is
The double
Any
circle of
the
pencil is said to
points
THEOREM 31. The radical axis of two common to them which are not on
imaginary points
circles
The
the cirlimiting points of the pencil which they determine are real if cles meet and in if they meet in
imaginary
Jll
U VjU JJL>JCJ.a..l.>
JT -LUi.-Pl Oil
THEOREM
of
circles
circles
32.
a pencil
of Type
and
complete quadrilateral.
The
sides
this
quadrilateral are
pencil.
la
and
the line
of centers of
the
Proof.
Let
and
z (fig.
common
57*) be the points other than Jx and Ja and let B l and B 2 be the points
of intersection of
1^ respectively.
Whether A, and 1
An
are
real;
line
B^BZ
Aa
with regard
the pencil,
any
circle of
is real.
=B B
:
it
conall
tains
the
centers
<
of
I., 1
L. Z
of
FIG. 57
Hence
and
Z>
is
the
line
through
and
The points
JE?
being diagonal points of the complete quadrangle A 1A ZI^IZ are evidently the double points of the involution in which the
z
pencil of circles
meets
&,
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
,
B^ B
A^A 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
THEOREM
circle
34.
If a
line
circle
Cz
meets the
in a pair of points
PQ
1
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
T be
and
TP Q
Z
at T.
The pencil
of circles
hence the
of the
if
first
statement
we
The
to
t
line
at
FIG. 58
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
(7
One such
circle,
by
from
2
.
P
2
by
JJ
and
orthogonal to
As
a corollary
1.
we have
The
set
COROLLARY
is the
of all
circles
orthogonal
to
a pencil of Type
COROLLARY
the
to
2.
Let
C2
be
circle,
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.
corresponds to
~by
inverse
reciprocal radii.
We
study of inversions in a
later chapter.
EXERCISES
1.
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.
OT
OX OF is
power
of
with respect
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
any point
it,
of the circle,
and A^ and
OA
5.
OA Z = (OC)
it.
2
.
is
one
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,
The
sircles is called
the
interior,
similitude. similitude.
7.
is
circles
of
Three
centers of
are not collinear determine by pairs six vertices of a complete quadrilateral having
its
diagonal triangle.
Generalize to the
If a circle
at
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
reflection are
Hence
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
/
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).
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
JliU VJJj.UJJCj.tt.-LX
pairs,
In the case
of
measure of point
We
have, in fact,
whenever
Z^
2,
by
substituting
From
(15)
we
define
2
m(^)=clogB(y2 ,v ),
^w+^w^fty
whenever
l
l^,
z,
are concurrent.
is
a multiple-
valued function,
we must
specify
which value
is
chosen; and we
must
Making use
Z ro
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.
we
shall suppose
a and
/3
subjected
This
a
where
= cos (6 + 2 mr),
TT,
= sin (6 4- 2 nm),
=6S2
and n
of
is
an
^=(0,
where
i
1,
Z
t
<>)
and
1
2
Ia =(0,
1,
-i),
=V
1.
Now
if
and
meet
(58,
Vol. I)
to, a.
-4- fl.
S.\
-4-
i (a.
8.
a.B.\
/^ and
cos 6
l
/3
and
= a/ -a /S = cos
2
2
= sin 6,
where
=#
2,
then a
= cos
and
QZ
+ 2 mr.
Hence
Here
In
again,
=a
/3
and
@=2aft
a2 +
2
= l.
fact,
= cos 2 0.
Thus
(17)
B(^ ,v = + ^
2
2)
Hence
(18)
where 2
is real
so that
S2 <
TT.
Hence,
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*,
from which
it
follows
by
(19) that
IT.
Indeed, (19)
is necessarily a
of a pair of lines
and not
of
an angle, because
same point
also
at infinity.
and
2-Tr,
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
EXERCISES
1.
Jf
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
If
Av A2 As A4
,
,
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 (OA^A^) or not, being an arbitrary point of the circle and being a sense-class on the circle.
{
5.
If a,
1),
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 (^a)
sin (bz a)
sin (c 2 a)
_-
sin (c2 i)
The
may
be represented in
the form
(z
Xa, y
0.
+
,
o\
is
A/3),
where a and
ft
are fixed
and A >
There
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
the distance
between
(x, y)
Two
*>
A5,
% % + AJ8)
if
an <
*o (x
Vo
and
if
Ao', y^
A/?')
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
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.
sides meet in
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
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
meet
I,
are concurrent.
This
is
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'.
conjugates in
an arbitrary
is
we
ly the perpendiculars
to the
to the lines
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
current textbooks do not often classify theorems on the basis of the ( 34) and the assumptions which, are neces-
(17). Some
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
abbreviate the proofs of theorems in the literature by a free use of imaginary elements (cf. 78).
EXERCISES
1.
(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
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.
and of the
circle of
pairs
circle is called
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.
its
orthocenter.
5.
lie
The
on the nine-point
CHAPTER V*
ORDINAL AND METRIC PROPERTIES OF CONICS
74. One-dimensional projectivities.
of one-
dimensional
projectivifcies
has a great
many
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
a one-dimensional
line at infinity as
as center
and the
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
is
reflection
with
as center is
effects
ative
all
and hence
all
commuton (72 an
involution
belongs to
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
latter sort
form a
Thus
the theorems of
can
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
Let us
first
some
of the ordinal
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
each pair
If an involution
alters sense,
no pair
An
that
is
separate each
other.
DEFINITION.
If
A, B, C,
point
AB
and
CD
is
called an interior
AB
and
CD
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
that there are no tangents through an interior point. may be stated as follows
:
THEOREM
fliutually
exterior.
the
1.
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
of the conic.
# Cf .
Now
rangle
let
If 0' is
to
ABCD
whose
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.
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,
ABCDj^OPO'P'.
and C do not separate B and D, does not separate the pair PP'. That is,
Since
it
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
is
In
of
78, Vol.
I, it
projectivity
is
a product
two
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
is,
made, a product
two opposite
involutions.
a,
just
THEOREM
involutions,
involutions.
4.
direct projectivity is
and an
opposite projectivity is
is
a product of
three opposite
An
opposite projectivity
involution.
of projectivities
is
of the
product of
have, as
two involutions
Hence we
COROLLAKY
COROLLARY
1.
Any
line in the
an
interior point,
and
definition
and
point of contact
is
EXERCISES
1.
Zoo
What
to
How
cular points? What transformations of the Euclidean group are harmonic OB. L> to the absolute involution?
2.
Show that
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
of a conic
and
RQ
A 1 and B v followed by another involution, and if T- 1 (J. 1 ) =A ^A 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 j^A 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 ^A 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
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
on the
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
we have just proved, but in order to have the result plane we give a proof which is entirely general.
THEOKEM
6.
Any
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 A^A 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
and
J?
of the line
A C which
have
and
their ends.
We
segment &
of the line
Q
1
AC
A A BZ con-
Let
3
S be any point of
A
and
let
2
or.
The
Let
BA
2
meet
it
in
and
.t
A 2 C meet
1
*
it
in
Bv so
FIG. 60
we have
Since
the
l
{C A B.A J.
Z
,
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^B^A^. Hence iptic, we have {BZ B^A ^A 2}. and A^ separate A 2 and G and hence B is interior to the conic.
I}
THEOREM
1
7.
Any two
If the line E^EZ 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 E^E^
Proof.
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
a2
be
0.
By
a2
+ +
6c
>
are opposite,
(4)
<
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
82, Vol. I)
c
is
x xz
:
1,
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,
if
(4) is satisfied
if (3) is satisfied
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
points on it
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.
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
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
,
C2
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.
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
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
3,
gives
By
Cor. 2,
Theorem
3,
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
conic.
2.
COROLLARY
Two
is
elliptic,
common
pair.
The center
of
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
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
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
form
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
P(P-&A, Pj^B)
of
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
8.
An
involution in a
is
orthogonal or there
under the arnne group.* such that every pair of conjuone and only one orthogonal pair of
lines.
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.
one
entirely interior to the other. 12. For any angle, 4. ABC, there
is
I,
I'
of orthog-
B -which
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
4 ABC.
13.
The asymptotes
of
an
any pair
of conju-
gate diameters.
14.
The bisectors
ABC
meet in four
26.
points,
one
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,
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
known
or of
(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
is
not a perfect correspondence between the linear constructions of projective geometry and the Euclidean constructions by means of a ruler.
The
constructions of projective
geometry are
(a) to join
(5)
two points by a
(projective) line
any two
lines.
(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
may
be thought
is
of as carried out
length
not limited.
(3')
The
operation
(1)
and
(2),
together
C such
that the
point pair
AC
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
AA
is
evidently parallel to
BC.
Thus
and
They obviously
;
quadratic constructions
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
do not provide
reader
VII
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.
* 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
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.
Given three points of a hyperbola and a line parallel to each asymptote, hyperbola with a line parallel to one of
number
of tangents to a conic
given by
five of its
points
also
any number
by
11. Construct
any number
points.
12. Construct
lines
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
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
and the
prove some of
DEFINITION.
Assumptions A, E, P. We shall now the most elementary theorems about the relation beS'.
points, lines, or planes are said to be conjugate relative to a real one-dimensional form of the first or second
Two imaginary
if
of
an involution
to
As an example
The
conic
and the
it.
ellip-
78]
183
to
the involution on
<?
whose
axis
is
/.
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
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
are on
By
duality
line.
we have
on a real
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
imaginary
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
THEOREM
12.
The
elliptic involution
From any
an involution
whose double
THEOREM
point
13.
If A^A Z and
B^^
nary points on
A^B Z
and
lines.
Proof. By hypothesis the lines A^A 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
.
AJBZ and A
B^.
By Theorem
12 each
of
The complete quadrilateral whose pairs of opposite vertices are A^Ay BJ32 and C^C^ is analogous to the quadrilateral considered in 71 whose vertices were IJ,2 and the limiting points of two orthog,
(cf. fig.
57).
With
such quadrilaterals
we have
A^A Z BJ3Z C^C^
, ,
THEOREM
of
a,
14. Let
lie
the
pairs of opposite
vertices
complete quadrilateral. If
A^
and
B^
are real
B
,
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
real.
In the second
illy
,/?
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),
Hence the
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
,
,
THEOREM
%l
15.
complex
line
least one
point.
Proof.
^+^+^,=
lis
0.
may
u
f
+ iu'l)
be
-f (u(
x*
>
here
Q,
u'J, etc.
are real.
real, to
This equation
is
equivalent,
if
xa x v x3
,
Q required to
EXERCISES
1.
A
A
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
is real.
The
conies under consideration therefore fall into tsvo classes, the real
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
;
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
of
I.
By
Cor. 3,
Theorem
9,
there
is
a pair of
separate
and
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
seen to be equivalent to the condition that the equaframe of reference in TT shall be expressible with
If
the line
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
is
called
an imaginary
The
76,
ellipse.
7,
15,
37,
may be summarized
group
the conies
THEOREM
imaginary
Under
the affine
classes,
Any
parabolas, hyperbolas, real ellipses, two conies of the same class are equivalent.
conies
must be
characterized
by
their
I2
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
circles
be
and
respectively.
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^,
Cv
a
.
<72
and
Hence
T2
C? to
19.
K\
The product
T2T
carries
C2
THEOREM
Proof.
Any
Let
C 2 and
be two parabolas
and
let
CM
and
let
2 1 ((7
)=
(7f.
Let
K
;
Let
and C
its
point of contact
also let
Kx
tangent to
2
,
and
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
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
JP meets
it
(7 2
in
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
B (J/
K^K
2,
C C2 }=
1
B (1
EXERCISE
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
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
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^
But
since
Jx and I2
are
by Theorem 14
if
F^FZ and
F[F' Z
is
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
is
In case
C2
is
an imaginary
ellipse,
both
are neces-
2 In case C
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,
,
C2
and B^ are conjugate imaginaries. Let the polars of Z\, v F[, z relative to
(7
be denoted by dv dz ,
d',
A[ respectively.
;
real
and
b.
A* and
conic
(7
a being called the major, or principal, axis and & the Each of the points Fv fz F[, F' is called a z
,
A A
I}
2,
B^,
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.
It and J2
?o
Jx and Jg
If
we
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
P would be a focus.
Hence
THEOREM
22.
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
,
The
form a range
47, Vol.
I).
through any point P not on the sides form an involution among the pairs
OA n Ba
which
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
and
PFf
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.
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
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.
aud axes of
by the same
letters as in this
The notation has been assigned so that for an ellipse the points
^
2
and
where J^ and
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
orc l;>
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
A L
ao
and
AnL
respectively.
Aa
is
l
of
as
L and L^
are conjugate
line
is
real.
is real,
Hence
its
pole, F,
is
real.
Hence
F to A*,
pass through
/ and /
a
2,
perpendicular. Conversely, if the pairs of conjugate lines at any point are orthogonal, the tangents
F are
1^
through
must contain
F are
imaginary,
F is F
of
interior to
G'
and hence
on d are
exterior.
The tangent
of
at
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.
may be summarized as
is
follows
THEOREM
is
25.
real
lines.
and
is
the only
a,
conjugate
The focus of
parabola at which
the
real
and
lies
on
the axis.
The focus
is the
unique point
all
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
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
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.
is
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
P meets
directrix.
&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
respectively,
and
F is
t
tangent meets two fixed tangents in points a focus, the measure of &PFQ, is constant.
and Q
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
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
on a
circle.
circumscribed to a
15.
The
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
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
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
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
eccentricity
is
unity.
points.
a hyperbola
if
meets d in real
In this case
Dist (FD)
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
circle is
if
and
be held
be moved so as to increase constant, and Dist (PF)/ Dist (FD) approaches zero.
FD
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.
we have
they have
Two
and only if
On comparing
eccentricity
is
this
it is
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
64) perpendicular to the principal axis meets the conic in two points, Q^Q^ The number Dist (Q^Q^) is evidently
2 , fig.
is
a fixed
DEFINITION.
The number p
2
Dist (Q^Q^)
.
is
(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
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
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-
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
ellipse,
respectively.
Then the
line
X
of
through
OA Y parallel
OA
parallel to
OB
in a point
of the conic.
If x
x
6
is
a cos
0,
I sin 0.
called the eccentric anomaly of the point P. 11. Relative to a nonhomogeneous coordinate
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.
may
be written
_
13. Relative to
a*
+
,
any pair
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.
Ex. 16)
83.
Synoptic remarks on conic sections. An inspection of the one that it would not be practical to include
known
how the
rest
may
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
I, V, VIII, X. second large class contains those theorems which belong to the
affine
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
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
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
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-
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
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
itrary line
through
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
The line
PQ is
.
at a point meets C 2 in one and the axis and the point P the center of an
C2
into
2
.
C 2 The
center of curvature
is
transformed
the pairs of
C2 in which
C2
logonal lines at
LI.
meet C
at
any point
of a conic
;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
W.
14.
The
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
jyclopadie der
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
(/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
now
easily verifiable,
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,
A:
or
A2
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.
The
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.
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
equation
of
the
a oi a'o
fl
'i ~"~
^o
(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
by
a Q1 x Q xl
-f-
^o
(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
;
fundamental conic
which
f ;
(9)
where
A^
is
a.j
in A.
The
coefficients
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.
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
otf+atf-a^O
Hence we have
is
THEOREM
32.
Any
~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
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
u^
^=0.
is
and hence there would be a tangent to (7 2 Hence a pair of necessary and sufficient
Cz
is
on no point
wx = = Substituting x
and
(2)
#=
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)
(or a
simple computation)
5]
ALGEBRAIC THEOEY
205
Hence we have
THEOREM
33.
The imaginary
for which
AM >
and
the real ones are those
and & U
-A>
0,
these conditions
are satisfied
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
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
equation
(8)
satisfied
are taken
by
Hence
the points
(Aqp
A A
ol ,
02 ).
A 01 AQZ
,
must reduce
(14)
It is to
(8) to
\/
i7 h W
A
I-
/Yi
rtn'
II
/\
13 *
"21 *A
*,
*4b
4t
be noted that
n
l
&"
because
(x'
,
if
F would
transform
all
points
x(, x' 2)
and hence
Am
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
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
formed by a collineation to
n (10)
f)\
-v.2 X 1
2 rz X
ft \J,
r>aaa
+-.A
points
if
and only
if
AH ^
0.
(6)
represents a
and only
22
A ^0,
i{
for
i=
0, 1, 2.
= A n = A =A =
0.
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),
(8)
transforms
all
If tin's line be
transformed to z1 =
obviously
becomes
(18)
as*
0.
A degenerate
may
point conic
is
two
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
Hence
is
A=
The complete
may now
ables
THEOREM
may
v '/
be
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).
Suppose
, , ,
we regard
,
(a 00 a u a 22 a ov a 1Q
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
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=
<
is
transformed into
itself
by
The only
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
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)
+ -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 a^a 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,
(22)
(23)
2/
(24)
(25)
To know
this it is
merely necessary
represent couics of the four types respectively. The criteria to determine in which class a given conic
C2
if
belongs
may
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
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/
we may take
X2
*
2
-J-
= =
r\
U,
(27) (28)
(29)
+l=0,
2
?/
a2
0,
"-! =0,
=0, **=0.
(30)
(31)
The
any equation in nonhomogeneotis coordinates. Summarizing these results we have the following algebraic theorem
THEOREM
reduced
35.
Any
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
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 homogeneous
(.4 00 ,
A ov A
0!1 ).
4. If
A 00
0,
the translation x
z
^j y
=y
A^
-^
^00
_ _ ax + 2 ax
5. If
and
A OQ 5*
0,
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
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& +
1
*,&,-(),
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-
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
-+* 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.
>
(43) to
and
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,
and will be left to the reader to be considered in connection with the exercises below and those at the
no special
difficulty,
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
.
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
)*+^oo=0?
and %/
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
The
conies
is
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
,
2122'
=
In
a!
&
:
Uo
1L -\- GL
These
'invariant.
are
If
the
transformations which
[1,
0,
0]
and
6.
c.
d.
c
i
a
i
cn
a.
(47)
is
The
same
is
of course the
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
ellipse, real or
imaginary according as a n
>
or
an
^ 0.
Thus we have
<+<+<=
0, 0, 0,
0.
<-<-< =
w2
conies is dual to projective classification of degenerate line that of degenerate point conies, and therefore yields the following
The
three cases
(1)
two
= 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
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
<+<=
ul
0,
+ u* = 0, <-< = 0,
^=0,
(58)
(59)
^ =0,
2
<=
0,
(60)
<=0.
EXERCISES
1.
The two
(61)
A 00 p 2 -
(a u
a22 ) p
+a=
0.
The
is fa.,
n..-^ 2 4-
4nA.
2.
In case a
and
A 00 ^
0,
3.
(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).
The theorems on
79)
may be
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.
THEOREM
36.
the involutions
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
(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,
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
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.
We
direct
;
of
is
and that
if
two are
From
this the
is
valid in
we have
a,
projective
self-polar triangle
else
and
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
Proof.
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
Let
Q and
and
distinct
from
and C on the
line
A C,
same meaning
tion of
let 0, R, R', JR" have the theorem (fig. 68). Also let
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 B,
and
0'
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.
were any self-polar triangle not satisfying the conby Theorem 36, be one for which
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
an
ordered plane
is
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
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
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
z
.
Thus we have
THEOREM
is either the
38. DEFINITION.
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
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
ABC
the
PQ, and by
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.
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
Any
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
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
and
to
and
B respectively.
lielative to
DEFINITION,
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
OX
OY.
In case
2.
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
and only
&AOX+4.AOY=4.AOZ
Since the direct similarity transformations leaving a point invariant form a group, the operation of multiplication must be associative, i.e.
OZ,
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
a commutative group by
Hence the
operation of multiplication
commutative,
i.e.
OX-OT=OY-OX.
The
fact that the
group of translations
is
group
OY + OX- OZ.
in Chap. VI,
1
.'
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
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
1,
any vector
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
= Dist (OA},
by
i.
Then by
Any
is
is
xf,
where
a?
is
vector parallel to
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.
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
,
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
of this correspondence
which we
line
shall
have
real
complex
which have
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
and
id
isting
theorems obtained by
re the articles
>ciety,
by
Vol.
I, p.
;ain
if.
an
line
s
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
tollowmg form
protective line
and
let
and
J\
lie,
L an
elliptic absolute
involution
given.
in
which
of
to
TT to
a point of
Q i} L distinct from Q x
to
P to
:
to Q*,,
and
This correspondence
I
such that
i.e.
leaving
Qv
invariant,
to
form
b,
x'
= ax -f
a 3=
0,
TT,
there
corresponds
direct
similarity
transformation of
and
conversely.
arises,
What
group of transformations
I,
i.e.
to the set
= ax -}-b
ex
-f-
The transformation
(3)
of TT corresponding to
x'
= I/a?
'
P to
= ^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
P"
Hence
71),
if
P"
an inversion. Hence
the transformation of
TT
corresponding to x' = 1/x is the product with P P as axis and the inversion L
circle
through
P
1
with
P
Q
as center.
Now
(1)
i.e.
any transformation
and
(3).
to a product of a dilation
has been seen to correspond to a direct similarity transformation, and a displacement. A displacement has
IV
to be a product of
and a
dilation will
now
or four inversions
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
is
^A
is
an inversion,
I
2
follows
IJjj.
If
is
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
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)
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
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
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
real reflections, and, conversely, any transformation of the Euclidean plane of this type corresponds to a, projectimty of the
line.
complex
92.
The
at
infinity
which
are products
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
the
real, is
The
first
is
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,
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
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
is
C invariant. By
n = A 2n+1 A2n
where
Aj,
(i
reflection.
Let
= 1, 2, A be
-,
-f 1) is
an orthogonal
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
is
THEOREM
10.
and
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
UAH"
is
Proof.
Let A, B,
C be
then
HAHwhere
H (C)
invariant.
If
n^v-A,,
A
,
AB
HAH- = A A
1
An AA
A^,
is
and
is
thus a product
of
an odd number
last
of
or inversions.
reflection or
theorem
1
an orthogonal
line
A are
1.
carried
by
Since
HAH"
is
an inversion or an
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
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
<7
must be a
is is
z
.
Since there
;,o,
no
',
TT/
/^,i
points on
I*
it
line.
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
point in common with the line pair a real point, I meets it in a single point and therefore
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
and
common and
and
It
are parallel.
was remarked in
the
fine print at
the end)
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^),
Under
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
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
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
Ex. 4
and
to
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.
point of
points
C2
A A'
and BE'
to a circle
with
as center
if
and only
ordered
BE', and
4.
(2) the
(1) triads
if
is
A A'
and.
OAB
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
=
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
If
is
is
any point
of the plane,
Dist (0A")
(cf.
Dist (O'X')
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
_ ~
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-
=
d'
1.
The
a b
c
at-}- I
d
fixed. If c
where
circle
is real
a, b,
c,
reduces to a
11.
The
circles orthogonal to z
ct
0, this
d
&
ft
(c
+ +
d(3)it
z2
+5+ + d+
aa
ca.
where a and
12.
f3
are real.
The
circles
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
formed by the
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
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
circles at Q, or
If C\ is
any
circle,
}
w
m
a line meeting
in a point Q,
and
m
1
mm
2
is
and
C*,
of
mjn^
is
z
and
wij
mf
The measure
of a line pair
mjm
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
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
denoted
x
by
m^
and
if
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
of lines l^.2
In case II
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.
1
1}
the line
goes
from
through
is
the tangent to Cl at Q.
line
1
Any
point
meets
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
with Im 2
is
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
THEOREM
'.h
If three
in a point
way
thai
and
P,
Q and
the other
an angle
Proof.
;h of
The
pair of circles
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
es.
But
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
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
as center
.
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
/.
Since
A l Bl
it
is
parallel to
AB,
and hence
is
the mid-point
Al
of the pair
CB
and
to BC',
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
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
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
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
by an inversion
carried
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.
is
an involution
if
and only
if
the circles
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
any
of the triangles
2 8
formed by pairs
a point 0,
2
,
of its vertices
Let three
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
Thus
far
we have
The
definition of
an inver-
sion given in
2
Euclidean plane
to a circle
(7
,
i.
e.
two points A^
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
point of L.
The
circular point
is
is
carried to
72
07^
C2
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
IB
and
IB
The
points
and A,
conies,
they
From
line
,
this construction
it
is
points, except
/x
of the line
A^,
A 2 fa and
vice versa.
Hence an
3,
and vice
versa.
More-
minimal
if
lines is
is
such
that
B
2
a variable
point of
<7
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
given in
The
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.
is
not contain-
(4)
<7
i
2
,
rr/fi
i
r
7\T
L-^iJ
r 2 L
iJ>
and
(6)
Hence
(7)
But corresponding
lines of these
two pencils
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
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
projectivities (5)
and
(6)
we have
Hence
only
if
is
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.
circles
which are
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
lines
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 A^A^ 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
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
/,,
and a complex
can be established in a particularly interesting way if the onedimensional form be taken as the pencil of lines on one of the circular
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.
vertices
of a complete quadri-
of homologous
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
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',
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
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
3. If
A A',
circle
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
except in that
to
.
it
transforms
la
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
The
is
set of points
on a real
circle, or
on a
real line
together
with
I*,,
An inversion is either
or
imaginary
etc.
an orthogonal line
92.
Circular transformations,
are defined as in
The
set of
figures to
which they
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
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
circles
geometry
For exam-
the propositions that three noncollinear ordinary points determine a circle and that two ordinary points determine a line combine into
THEOREM
19.
In
the inversion
7
77.
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
there
is
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
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 ]
rZT, !
Since there
is
circle
of the
set
[E]
through
of
must be a pencil
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,
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
line.
The
direct
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
= +
(cf.
78).
For
this reason
DEFINITION.
Two
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
99.
EXERCISES
1.
science.*
2.
Work
71,
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
>n
plane.
e plane
By
leaving out
ations on
10.
any
circle
which
by
inversion.
Hence
3y circular transformations.
circle
on
by
circular transformations,
and order
relations
is
on chains
are unaltered
by
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
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.
C2
C2
there,
COROLLARY
tlie
1.
points
to
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.
COROLLARY
2.
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.
B (AB, CD) =1 +
st
iy
and
iy',
&.
cr!
1J 1
ViAincr TP.nl
From
this definition
it is
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
Hence we have
circular transformation carries non-neutral throws into similarly sensed throws, and a, nondirect circular trans24.
THEOREM
A direct
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 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'
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
),
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 A^A_^ = 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
,
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
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
may
be expressed in terms
of the
THEOEEM
25.
Any
direct
.,./,
..,
-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.
we have
and only one
THEOREM
26.
Any
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
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
no
line
through
and no
falls
with
is
as center.
or a dilation of period
two
a point reflection,
transformation
under one
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
and
is
of
period two.
DEFINITION.
if it
said to be parabolic
if it
;
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
COROLLARY.
An involution on
a complex line
is
both hyperbolic
and
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
is
Pm
is
x v the projectivity
may
be written
(11)
x'
= k (x
Zj).
e ie ,
The projectivity is
loxodromic
if
where 6
is real,
and
may
be written.
(12)
or, in case the double point
rL- = _L_ +
at,
is
=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
pencil of chains.
3.
The
projectivities (10)
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
if
Elliptische
4.
nates,
I,
Abschnitt
II.
of view the r
and 6 in Ex.
meeting the
through the
orie-in at
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
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
by
C. Segre
and
C. Juel referred to
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
96 are
(15)
=a
-4-
These trans-
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'
space.
correlation
is
or
called
an anti-
The theory
of antiprojectivities
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
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
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,
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
.
two
28, Vol. i.
8,
t Von Staudt, Geometric der Lage (Ntirnberg, 1847), of a real line as a transformation having this property.
We
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.
COROLLARY
space
1.
Any
collineation or correlation in
a real protective
is projective.
Proof.
in such a
way
-ive
that the correspondence between any two homologous 33-35, Vol. I). Hence, according to the theo(cf.
projective.
and
circles into
Proof.
TT,
minus a point R,
as a Euclid-
ean plane
let II
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
Hence
Since
H'H
H=
a transformation, say H", of the inversion group in H'" 1 !!", is also in the inversion group.
is
TT'.
COROLLARY
is either
3.
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
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
of a circle in
TT
(x
+ 7/
4- 2
ap +
2 ajj
+ a = 0.
%
DEFINITION.
degenerate
an
ll x ,
ordinary point to the circular points at infinity or a pair of lines where 1M is the line at infinity.
circle,
(17)
=
=
/
The condition aQ
clearly
find
means
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=
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
It and J2
respectively.
There
is
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)
TT,
and
the points of
S 8 on (18)
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
(/3 Q , /3 1} /? 2 , /3 g )
such that
A=
The
circles (real or
a v aJ>
av
(*
>
! 2 > 3)
j=<>
(1, 0, 0, 0),
fundamental
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
the equation
(\aQ + tfj
(x*
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.
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.
Theorem
THEOREM
projective
30.
The
to
the
geometry of the quadric (18). COROLLARY. The projective geometry of the real quadric (18)
is
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).
:
lines contains
Q, of
I.
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
The equation
of a circle in TT
The
points (|
^, | 2
3)
which correspond
plane
now
(21)
#=
circles
+#+&
(1, 0, 0, 0), (0, 1, 0, 0),
and the
(0, 0, 1, 0),
and
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
^/^o, ^2/^0'
3/^0
are regarded as
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
A
C
circle
is
linearly dependent on
of
circles Cf,
if it is
member
some
C; +1
,
+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
rest.
The
dependent on
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.
first
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
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
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
we need
.
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
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
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
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
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
and
I'.
It is
denoted by {Z/'},and
31.
and
I'
re/lections {Oca}
a)
{PTT}, where
is the
plane on
V.
P and
and
the
I,
plane on
let
and
V.
and
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
and
P respectively
COROLLARY. A.
two,
line,
reflection is
and any
the points
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,
a, ft, y,
8 the respec-
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
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
on each 5. A
of a cnmnlex:
linp. aunli
whi^Ti is
is
an nth root
of unity
in case n
3,
this
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. ">
'
^&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
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
line
of
<?
and
104, Vol.
I
I)
ing that
to
I
to a point
of
other than
K must
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.
A
Q
a and
a way that
transformed into
itself.
THEOREM
on
the
33.
A
fff
projective collineation of
same
n.'ntJ.
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>,
invariant, all
collineation
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
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
as before,
{wX>
leaves
all
m mJ
i
T" 1
points 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
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
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.
P,
Q,
respectively,
let
a', I
1 ,
c',
p', q
r'
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
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?
to
P, Q,
R respectively.
That there
R*
invariant,
R$ because F leaves
FF'
all lines of
rr'=FT.
By Theorem
where
I,
34,
= {lm}
{rs}
{I'm'}
{r's'},
2
.
m,
r, s
and
The
collineations {rs}
and
{I'm'} are
commutative
for the
same
reason, that
F and
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
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
Hence we have
THEOREM
surface
is
Any
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.
Theorem
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
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.
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
0.
Hence
where
is
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
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
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
of a
=0
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
+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).
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.
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
01^, and
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
.
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
Moreover,
effects a projective
minimal lines
("ml parallel
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
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
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>
Px
's
(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
I of TT
p Q with
m, m
m containing the m
If
with
A point of
points of
circle of
TT
to the point
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
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
TT
space.
Let
be a quadric surface
(fig.
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
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
From
this
it
is
evident that
if
as
X I^
and
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
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
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
K =K
2 2
.
The
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
O x J2 Hence
.
corresponds under
2
to a non-
TT.
An
.
inversion in
TT
minimal
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
.
K
2
in
collmeation of
effected
by
a harmonic
2
.
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
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
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.
WWWOL^OJWUO,
^^
DEFINITION.
The
set of all
ordered pairs
called
oo, is
(x, y} where x and y are a complex function plane, or complex variables, or the com(x, y]
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.
,
of
K'^; y>yd>
and the ideal points
a.re
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
2 7) (
^+
fy^o + 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 + ^i =
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
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,
conditions,
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
, ,
:
where
1.
For the
1}
2,
f 8 ) given by these
#=# + # +
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
of
or
become
infinite
(if
zjz
0); and,
moreover, two
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
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
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
"The
of the
American Mathematical
106. Projectivities
of
XX
(1914),
p.
185.
general.
one-dimensional forms
in
The
theorems
of
among
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
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
*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
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
jvn><,
iwyrvv
(.*
prujvwvbt
ttn/t// ut/t/y
t,u
<//M>
new
to
definition only if it is
I.
22, Vol.
To prove
this
theorem
it is sufficient
show
that a sequence of
of the first degree
perspectivities beginning
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,
where
is
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
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
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 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
is a
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
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
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
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
\W vw
'
effects
tlie
complex,
and
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
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
I'
and I"
on
C 2 Thus on C z
.
But
(cf.
77, Vol. I)
l
l
is
the axis of
I'
center of I".
similar
(i
2, 3, 4)
passes
which Hence
is
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
at points of the
double lines
existent) of F.
If
,
is
any other
2
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
is
as axes.
-
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^
.
Hence
By
TT
definition,
if TT is
form a
TT
flat pencil.
tangent to
let
be the line of
TT.
on
and q the
on
In case
is
2
,
i.e.
the
pencil of lines on IT
In case v
is
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\
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
fully determined
by one
the latter
is
78, Vol.
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
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
,
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,
,
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
J.7UUJ.
juci/
i_/
ueuuuc
i/ue
uuii^i
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
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
It occurs
when a a 2
are
met by a
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
both
and
in
meet pq.
EXERCISES
1.
The
lines
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.
be lines of
Z
C
Z
meets
tliis
7i 2
gate to
P P
,
containing
,
1?
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.
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
interchanges
and
mq and
two
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.
and A'
in
B and B
1
.
By Theorem 35
which
there
carries
is
and
B to
B'.
By Theorem 39
and
and and
m are polar with respect to Q m both meet the line AA', and
line BB'.
and
If there
of polar lines
I',
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.
14 v* rta
ot-*
nvt
o-nfc
Virvf-.l-i
rt/"vn
TTI/TQ^"^
i".f\
THEOREM
4tl.
J.J
a,
simple nexagon
'in
face in such a
way
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,
By
the last
pair of
like
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 C^A 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
By Theorem 38
In view
of
may be
such a
COROLLARY
If a simple hexagon
of
lines
is
inscribed in a quadric in
way
on a
which meet
the
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 a^a^ 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.
on a
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
have a point in common or that they an involution. Thus the theorem on involutions may
three
THEOREM
{L^I^}, {L 2
Z/2/1/,,
42.
2 },
LS
involutions
{L S
3 }
is
an
involution
in
of points
L^MV
the
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,
the theory of the parabolic metric group in the plane. Corresponding theo114 rems in the Euclidean geometry of three dimensions will be found in
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
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.
If
simple hexagon are met by three pairs of lines a^a 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
a^a 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
common
regulus.
two
lines
and
are
met by two
pairs of opposite sides of a simple hexagon. If sides containing them, and is the mid-point of
and
it.
C is
5.
Let
A 1 B Z C A Z B 1 CZ
1
be a simnle hexagon
A. T
BZ
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!"
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
definitions
and notations
collineatioii of S' is
'
(34)
tf
=z
xi
=
i>
^g=
of
The frame of reference is such that the points nates. The transformation changes each point
where the
a's
and
(a
/3's
-*0
if
a^-i/3,,
-*',,
-*
8 ).
by the
,
real line
+ X0 ^ + X^,
a z + Xft
a.
+ X^
3)
and
are the
of
78,
goes to
as real.
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
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
first
or second degrees,
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
said to
THEOREM
second kind.
Proof.
43.
Any imaginary
be an imaginary
of the
Let
line.
(
It
points, else it
would be a
In the
first
I,
real line
70).
Hence
it
contains one or
I,
no real
point.
case let
one of
and
conjugate to P.
The
line
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
would be on a
real plane,
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
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
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
:
is
an
elliptic
THEOREM
of an
44.
An
imaginary
line
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
below.
lines of the
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
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.
of an
imaginary line
correI
sponds to
its
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.
'
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
as a double point.
Let
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
is
expressible
in this form.
110.
The
principle of transference.
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
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
:
[e]
these elements,
and a
G.
fet
\e'~\
of
"be
theorems
which
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
[e]
and
are
denned
is
such as to make
[e']
and
G'.
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
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
it
seems
desirable to
In case the set of theorems [T] is arranged deductively, as explained I, it becomes a mathematical science. The
of the
problem
is
that of
determining,
if
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
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
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
of
an isomorphism
of
the groups in question. The theorems should be classified according to the 83. principle laid down in
*5.
A
R
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
"
laid
CHAPTER
VII
DEFINITION.
Let
TT^
The
set of points of
is
and
7r n is
called the
of this space.
The plane
7r m
;
and the
all
to be ideal or at infinity
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
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
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.
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
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
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
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
',. 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
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
ABC
if
and
Only
7.
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
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
the
ordinary points of
tr.
As
corollaries
we have
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
it
may
The
48
i
is
such that
plane
IT',
a plane
IT
to
49
are
two measure
ABC,
DEF
7?i
by an
affine
colliner^on to
A'
C',
D'E'F'
respectively,
(ABC}
'
m (DEF}
rp.sTilf-.
= m (A'B'C'} ~
m (D'E'F']
'
This
defends on the
p.rmv.
DEFINITION.
lent
if
Two
if
and only
AB C may be
is
triad
A"B"C" which
to A'B' C'.
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
a cor-
measure,
i.e.
measure
is
now
triad in
of
any
parallel plane,
We
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
2, 3, 4)
in a coordinate system in
(0, 1, 0),
which
0,
P, Q,
are (0, 0,
(0, 0,
if
1)
respectively.
if
-^
Two
equivalent
and only
they have the same measure. In real affine (A^A^A^) is called the volume of the
A^AyA^
OPQB
and
is
denoted by v
is
(A^A^A^.
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
.
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
OR
triads
OPQ, OQK,
ORP
systems of vectors. The ordered point are taken as imits of measure for the respectriads.
tively parallel
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
the
planes y
and
EXERCISES
1.
The measures
of ordered tetrads of
x'
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>
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
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
at infinity
and
101)
is
pair
on the
x-,
y-,
and
z-axes
f2
xv
i/2
T/J,
z2
Zj,
a;
and those
OP^P Z on
i
the planes
0,
0, z
ll\ 1/2
9/~' ~2
oA
zi
zi
2
x _ -,'"
v
^2
xi
T
2
2/i
'11
v/2
These numbers
/3/x
yv
= 0. P P2 = Vect P{P 2
X
Any two
points
PiP 2
of the line
PiP2
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
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
,
P P4
', 8
'
Vect (PS'P4')
if
P P2 P P
I;
,
3,
P\PZ
>
P\P%>
common and
Vect (PjPj)
then
*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
A AV
.,
A-^A Z ,
.,
A nA
of a simple polygon
A^A^y,
An
in points
B Bv
,
Bn
respectively,
AB A&
A^
(
AzBi
A n Bn A Bn A A
,
lt
A-^A Z
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
no three
which are
(503)
this
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.
221.
113. The parabolic metric group. Orthogonal lines and planes. be an arbitrary but fixed polar system in the DEFINITION. Let
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
transformations.
Two
figures conjugate
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
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
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
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
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
perpendicular
is
is
the
through Ca meeting the two given lines; this line unique and not minimal.
obviously
meet in a point
U.
114. Orthogonal
plane reflections.
is
DEFINITION.
homology
of
period two
whose
center, P,
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
to
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>,
in a
THEOREM
An
is
a transforma-
By
a direct generalization of
:
Theorems 3 and
4,
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
a non-
the
T=
and
if cr is the plane perpendicular to
TT
I
T=
(3)
{irP}
{o-S}.
I
translation parallel
to
a minimal line
is
a product of four
THEOREM
where
9.
A
,
product A^A,,.^
x
.
of orthogonal plane
reflec-
form
A,'(
Al_i
A^T
or
T'A^A^
Ai,
h! 2
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 A^A,,^ translations. In case
reduce
to
the identity.
Proof.
Let
' t
(i
1, 2,
whose plane
plane through
parallel to the
.
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
A A
_A/ A'
!
A/T J.^1.
"he product
s
A^A,^
x,
since
reflections
A|-, it is left
invariant
to the identity.
three orthogonal
y
plane rejections
J.
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,
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
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
the poles
their
axes
Hence
and
at infinity
its axis,
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
A^A^
two in
reflection.
,
A A A
2,
are parallel
we have
by Theorem 8
(1)
that
AA
2
i.e.
Theorem 8
(2)
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 {X^a} 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
Two
figures
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
congruent congruent
with
&,
then F^
symmetric with
&t
then
is
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
whose planes
reflections
whose planes
through
0.
Let
be the line
of intersection of X,
I
and
of
let
X be
a plane containing
12
'
If
is
nonminimal, by the
corollar
Theorem 10 there
{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
at
Now let
and
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
COROLLARY.
Proof.
An
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
the rotation.
isotropic or
If
the axis
is
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
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
by
and
and
From
this the
DEFINITION.
dicular planes
reflection or
The product
is
called
a half turn.
is its
axis
and
I'
300
[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
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
It also leaves
two on the
uujjuo.
jL-LoiAuc
line of intersection
10
a,
iiuinujLug^ ui
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,
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,
be any displacement leaving TT invariant, an and T the translation carrying to F (0). Then
13,
T" 1 F is
a rotation. Moreover,
it
can leave
it is
TT
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
it effects,
and therefore
also
effects,
symmetry
VUSJI VOOWItt/
r = TP
As
and
T=
P'T',
invariant.
be the translation
carrying
to
(0).
T~ r
x
is
a rotation, P.
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
all parabolic transformations and involutions leaving the point at infinity of p invariant.
posed of
EXERCISES
1.
Two
if
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
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.
*
of
an
(in fact,
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
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
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
affine
geometry
is
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
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
set of
theorems
is
selected,
Euclidean geometry
ii
a self-conjugate subgroup of displacemei defines a relation called congruence having the properties stated
* Cf the remarks
.
article
by Enriques, Encyc
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
satisfies these
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
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
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
[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
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,
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
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
transformation of the Euclidean group which leaves a line pointis an orthogonal reflection in a plane
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
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
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,
and an
elliptic
can be proved as in 116 that in case of a real space polar system the parabolic metric geometry satisfies
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,
coefficients
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
'
affine
considered in
A = 1 and
which
A = 1.
The equations
ad
' 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
homogeneous coordinates
cap8
2
are
axl
+ Hx% +
=0,
0.
#o
^<
= v/ ai
xz
= Vbxz
*
to
(10)
as?-!-!
+<=
<),
= 0.
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
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
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
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)
+ <,
(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,
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
22
13
2l
23
31
12
18
22
23
32
83
its
we have
A = p=
2
8
.
DEFINITION.
is
1 denoted by A"
A
.
its
(95,
Vol.
I) is
A
is
equal
A linear
a zzy
transformation,
(17)
y'
2'
= =
V+ +
81
+a
-f-
28 2,
8S 2,
a^y
(a n a., 2 a a3 )
at
:
is
orthogonal,
is
said to be orthogonal.
THEOREM
can
17.
be written
(18)
y'
==
= p (a u x + a p (a x +
zl
l2
22
+a +a
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
y'
COROLLARY 2. liomology (19) is commutative with tion leaving the origin invariant.
Since an orthogonal matrix
is
any
collinea-
(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
^e determinant
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
of
(17)
must
(21)
The determinant
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
of
(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
line.
The
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
(23) is satisfied.
satisfied.
cofactor of
where
i 3=
3=
Ic
3=
i.
By
-K +
which vanishes. The cofactor
of
aM
+a +1
. B)
=
an element a~ (i
/) of
D^ is
and by
1.
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
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
Let
be an orthogonal
X
plai
of fixed points of 2.
orthogonal transformation for which other fixed points than the origin.
it
A=
By
is
2=
From 2 A
=A
follow
THEOREM
An
A=l
is
COROLLARY
a symmetry.
1.
An
A=
1 is
Any
=1
is
a product of an orthog-
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
the
=l
and
A =1
is
such
x=
Q,
y=
0, 2
x/==y>
^^
z>
= x>
(1, 0, 0) to
(0,
1, 0)
(0,
0, 0)
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
u?
absolutely invariant.
2.
if
w|
4-
2 ws
Two
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
+ 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
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
ACB. Hence
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
which
OP
and
OQ
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
OPQ,
is
The
definition of the
measure
of the vol-
ume
of a tetrahedron
may be
taken from
OPQR is such
112, with the proviso that that the lines OP, OQ, OR are mutually
OR
The
from
definition of the
69.
measure
of angle
may
literally
Since, however,
effected
by
/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
=
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
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,
S X < TT and S 6 z < TT. The numbers which we have been defining in
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
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
= (xv yv zj,
which
are
% = (x
PZ=(XZ,
2/ 3 ,
%), P<
= (XI,
0,
y,, 24 ),
no two
is
of
rectangular, and
0),
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
to
as the
direction cosines of
OP r
If r
(/-
/J 2 ),
PiP2
r 3.
The equation
The
OP l
is
V+
4.
ci
2/i2/
+ +
zi2
= k+ ^8?, +
yz
is
/3//!
Yz t
-8
on the
line C
i
Va2 + ^ +
5. If
f
Pz
\,
Qj
is
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 ^o
~~"
-^^^t
\s i
-t
X/iSt C
-i"
COS
>O /
-t
Lf*n*
6.
m(P 1 P8 P 8 P 4 ) =
and
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
I,
a vector
OK
OP t Pz and
(OP2 )
.
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
of planes ay.
120.
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,
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
follows that the quadric surface is nondegenerate, which contains two proper or improper reguli.
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
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
at infinity in a point
and a
respectively.
are perpendicular,
Lx
Lx
and
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,
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
which
THEOREM
circle.
21.
Any
2
perspective with a
Proof.
Let
(7
having a
and
same cone.
COROLLARY.
point.
Any
from a
EXERCISES
1.
The equation
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
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
This
of
is
analogous to the
of
work
in Chap. V.
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.
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
,
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
[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
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
line reflections.
Proof.
we
shall denote
by A,
is
Theorem
22.
a product of a rotation
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
Let
be the
to
p a and
Q' to r n
These
;
p*,
is
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
p and
r be the lines
PR
and
RRn
respectively.
Then
p and
Moreover, {pp x }
{rr^} effects
the
inverse
of
the
-
transformation
by A.
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)
and
m, is
not minimal.
I
A
is
is
a rotation
about the
(b)
common
intersecting perpendicular of
I
and m.
and
is
minimal.
an isotropic
of
to the point in
and
It
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.
The The
lines
I
and
and
lines
m do not m have a
intersect.
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
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
A*,,
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
77%.,.
i.e.
a product of a non-
isotropic rotation
by a translation
screw motion.
The
THEOREM
24.
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,
And A
cannot be an
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,
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
common
AAA
3
effects
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-
Thus we have
THEOEEM
26.
In
any
dis-
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
,
of
of
satisfies
By
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
A = {qq
A
vector
is
{mm
{mm
{pp'}
= {qq }
1
{pp'}.
The analogy
striking.
of this process
with that
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
,
AgAj^
= AA
and
a
Hence
to
any group
of transformations
See
108 and,
to.
by H. Wiener which
The resolution
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.
and a
are perpendicular to
at P.
Hence
{aa'}'{ll'}
Ais
{IV}
= {&&'}
= 'P
and
p,
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
P and
A
IT is
the set of
of period
two such
that
A = PA
a rotation.
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
which
is
also
an exercise (Ex.
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
a half twist, and orthogonal line reflection is a special case of is a product of two orthogonal line reflections whose
EXERCISES
1.
If
the three
common
site
the mid-points edges of a simple hexagon are also the lines joining
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
line reflections
perpendicular.
5.
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
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
(27)
A = P2
P
or of
.P t
If
where
P x and P 2
is
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
(28)
A = PH
is
where P
P or of can be chosen arbitrarily. For any P and exists a rotation or translation P' and a half twist H'
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
distinct
The main
properties of displace-
s3
Any
in case
A is a translation.
i.e.
ment
is
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
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
[CHAP.VII
The
P such
that
A=TP = PT
A
is
where
is
is
/
t
parallel to
1 2
Let
and
be
two
points of
and
TT
1 2
measure
(
of 4.
0(and not
of the
two
sense-classes
positive.
z,
=
,
Dist (AjB a )
that
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
is
the
same
for
any choice
and
B A
A
for which 9 i=
and,
is
vectors
A^A
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
.
z,
1 2
and the second a unique point B through A v perpendicular to a and such that
tan
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
numbers which will specify the vectors A^A 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
first of
these articles
articles
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
same
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
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'
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
A v Az A8
,
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
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
;
through a point
or
all
meet on a
line or coincide.
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
and
If
A is
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
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,
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
N.
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
lines join-
and
I'.
The The
affine collineation
P (Ex.
11,
122)
to that of H'.
16.
correlation
P (Ex.
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).
we
By
invariant,*
90
it is
amounts
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
way
and
its
points at infinity.
Let OX,
Y,
OZ
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
Let (aQ a^ # 2
,
3)
that
if
R R
is
ordinary,
~a
aQ =
'
~a
ao
a
'
~~
'
ao
and
if
is
at infinity,
'
0.
= 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
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.
= (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,
by the following:
/*
ft=
,
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
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
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
of
rotation corresponding to
x'
1
Xa Xg
,
(33)
y
z<
8 z,
This
is
and
,
its
determinant
+1,
(2)
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)
,
,
/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)
z.
If
P(x,
(35)
y, z)
and P=(x,
y, z),
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,
,
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
z.
It is also to
(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).
we have
a, a.
(39)
A+
l8-o
^
+<
.
5 ==
3a
Ql
y
.4
'
(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
terms of the
a^.'s
:
may
be obtained by
**r\
wnQ*Q
4: OCn GCn
From
(41)
: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 /
).
satisfies
namely
where A, B, C stand
for matrices.
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
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
the matrices
0\
/O
000
\0
taking the roles of
/I
0\
II
of the
and
010
form
O/
\0
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
of
view
of matrices.
For
matrix, the transposed of S is - S. posed matrices of the two given matrices product, the transposed of
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 +
for a
in terms of
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
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
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
leaving a point invariant may be regarded as a subgroup of the collineations of a sphere having this point as center. Let us consider
4-
4-
x2
4-
x3
102.
-1
rf>
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
x.x'
x.xl 4- OUG' 4-
xxl
= 0.
tutmg
(47) in (4$)
we
nave, as in
no,
we have
The matrix
must
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
invariant.
Hence, by
is
102,
by one
of type (51).
A rotation is
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
zero.
Hence the
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 (o^a, 2 (a2 3
+ - a aj
;
2)
2 a
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
,
(/3
/^
/32 , /33 )
7fy
I
'0
'3
-(V
7s
(V
i-i
O/
11
124.
EXERCISE
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)
( ao
ia i) ^o
a ( s
ia 2> *i
M=~
127.
( as
'
aa) ^o
+ ( 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
The same
two
hppn sppn
such matrices.
itself
Hence the
of
set of all
such matrices
is
carried into
'100
1
010 001
0010'
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,
j
(52)
-1
,
*
^^ n
..
-j
n v
f\r
Z**
JL
j
It
-i -1
satisfy the associative
and
commutative laws
satisfy the
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
^o
-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
number system
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
by Dickson
hi the
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
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
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'
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^
e, 2
laws
Although these matrices satisfy the associative and distributrv of addition and multiplication and the commutative law c
it is
if
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
=/.
of
two-rowed matrices
where
a, /3, 7,
is
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
W
*v
/v>
i,
j,
i\
/'
_V:TI/' V-i
which represent the
identity,
tv
, /yi'
wr
__ ^"
__
/>>
tv.
,1,1 _
__
,__,_
Ml w
a?
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
126.
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.
W lilt; II we
unseu.
JLU
g j.^;^ uu.
a,
O^UUIJ.U.CKIJJ. uuu.oiii.u.ub.i.uu
aao LLUYV
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,
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.
all
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
An arbitrary rotation P
fully
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^
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
z>
The point
of
is
on the axis
of
P and
is
P v P P
2,
8.
If the line
OR
meets
S2
in
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.
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
: ,
OR2
meets
2
.
The
line
m' and
m"
.
through
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.
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
of
R%
of
R%
in the plane
Om.
are derivable
from Assumptions A, E, P.
rotations represented
by
with
as origin to a
P x (r) is opposite to P
124
is
2
(r).
The representation
by points given in
identical
S z is
real,
If
$2
is
a ruled quadric
and C* a
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
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
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,
contains
all
is called
exterior,
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
(57)
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
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
of
homo(40),
^o
/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:
(59)
@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
a^ a2 a^
,
Q,
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,
or,
in other
words,
that
A" = A'A. By
In view
of (59), the
elements
of the first
Hence
all
the elements of
B'AB
are zeros.
row of Hence
AB
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,
_
o
ttl
where
A*
K_
\a s
.
a* a2
-a a ~
2 *
as a~
-a\
and
B'AA= B'
+ < + a\ +
2 8 ).
Hence
(62)
Since
follows that
(63)
Hence
(64) '
v
ffi
^=
=ai/
<x 2
-t
130]
BIQUATERNIOKS
347
(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
by means
(66)
=
a's
(ff
+a +
z
l
<r
t
^)
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
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",
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
,
T2
N,
etc.
as
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
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
of tl
73, Vol. I)
if
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
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.
/(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
,
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
,
lines in [Z]
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
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 A^A^AS 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^ ^^A^), and A a point such that the ratio A^AJA^A^ 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
Az = T
*
On
this
cf .
pp. 914-916.
CHAPTER
VIII
NON-EUCLIDEAN GEOMETRIES
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
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
it a
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,
Let us at
first
ordered (A,
JS, S, J
On
this basis
we have
'
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
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(
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
and
(2)
that
F
/
is
between
A
its
and B.
Let
be a line exterior to
points
\
s^^Vc
/
-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
and B.
many
lines
not meet
Proof.
a.
By
Any line
of the hyperbolic
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
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
NON-EUCLIDEAN GEOMETRIES
[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
<
if
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
-'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
whose center
is
an ordinary point
called a displacement.
Two
figures
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
(
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
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,
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
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.
of elliptic involutions
on the
It is
shown
in
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 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,
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,
;
DEFINITION.
Two
lines (or
two
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.
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
is
an ordinary point
is
called a displacement.
Two
figures
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
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
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
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
to
of
a ray having
is.
as end.
The point
I) is interior to
C'
because
Assumption XI,
the displacements form a group. Assumption XII the hyperbolic geometry by the argument used in
may be proved
for
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
Euclidean
circles
71)
5.
THEOEEM
DEFINITION.
circle
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
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 {P^ABCQa}.
1
Let
B^ B B
Z,
&,
be points
lt
a,
such that
AB
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
= bx
Qx
.
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
EXERCISES
!<!.
,
,.,,,,,
1.
If
'
'
"'"*
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
.
meet in
,,;,,. ;|^||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
,
is exterior,
$ 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,
effects
a parabolic trans-
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
We
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
MM'. Then
is
the product of
as orthogonal line reflection with axis by that with OM' as azis carries
S,
OM
M,
T to
S',
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
from S' to PQ, S'M is congruent to 8M. For this reason cr and r are
called equidistantial curves of
PQ.
point
A
ment
(see
absolute
a product of
two orthogonal
line reflections
whose
of the displacement.
Hence the
of
disp
leaves invariant
the third
And by
as
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
0.
conic interior to the absolute and having cor of the third order with it is called a Jioroeycle.
DEFINITION.
The
of
all
path
ci
134=.
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
any transformation
z
of
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
and
the
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
.
hyperbolic geometry.
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
.
of the inversion
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
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
,
hyperbolic plane
correspond
K'2 invariant.
THEOREM
In
the correspondence
to
circles
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
,
and
(7
invariant
line reflections
whose
and
axes are on 0.
to the displacements
symmetries
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
to 0, the
argument
just
made
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
COROLLAKY
COROLLARY
Two
If
circles
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
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.
conjugate
THEOREM
8,
In
the correspondence
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,
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
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
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
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
COROLLARY
sponds
to
1.
In
the correspondence
T a
circle
touching
corre-
a horocycle of
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,
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.
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
I
in
VI, VII,
X~ XVI of the
are true,
29 and 66
CoKOLLARY. Assumption
plane geometry.
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
tCf. Vol.
p. 202,
of
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
it
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
of the differential
geometry (1854).
(141-143
below),
though
it is
of elliptic
geometry.**
The proof
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
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
title
"The
Science Absolute
German
translation
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||
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.
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
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.
The measure
of angles
may
be defined
defi-
nitions
69 without modification. If we represent the absolute and an arbitrary point by a Euclidean circle C" and
and theorems
of
its
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
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
The transformation
as a direct
measure invariant
93),
and
bolic angular
measure invariant.
Z
at the center of
<7
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.
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
at A,
B, and
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
*
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.
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.
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
,
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
= Dist (.R4),
t
because
$ (AB, A m B )= B (BA Bx A m ),
a)
and that
if
A, B,
{ABC},
Dist (AB)
because
K(AB, A m Bx
it is
).
evident from the properties of the collineations Moreover, transforming a conic into itself that a necessary and sufficient condition for the congruence of
CD
is
&(AB, A x B m ) =
where
(CD,
,
0,-D,,),
A x B^
,
and Cn
Dx are chosen
analogously.
of
Hence
and
a necessary
is
and
congruence
AB
CD
Dist(^J?)= Dist(CZ>).
distance function denned above
(
Hence the
is
67).
-=
logB(0P, 0..).
Let us consider
=:
)
"'oo
'
^'Ol^'O'^'l
'
^02^1)^2
-\_1_
a x x
n
-\'
a x
x*
'
-\_1_
a x
x^
'
_1_
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
\, are
,
2)
The
\Y + pZ = (Xy + /!,
and the points in which
by the values
this line
\y
fi#v
tyz + /w 2 ),
are determined
F)
+2
Z)-f(Y, Y}f(Z, Z]
A*,
/(^ ^)
,
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) =
'
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,
of these
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
(13)
f(X,X)~C,
is
where
a constant.
Since f(X,
X]
is
two
sets
xn \
Some
manifestly necessary in order to apply the processes of differential calculus to formulas in homogeneous coordinates. The particular relation
coordinates
f(X, X)
among
any
now
familiar notation,!
ds__
dt ~AJ=O
Pist(X,X+AX)
A
2
X+ AX)
= L
= Z i2
A-O A*
by
(8).
2
2
)/ (X,
Since
,
f(Y+ Y'
2
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
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
(15)
= F ^U
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
(16) ^ '
Let us
now
and choose
(17)
If
C=
47
xl
is
%l
4 72 # 2
4 72
real
(18)
=f(dX, dX}
= dx? + dx%
If
4 ffdx~Q.
v
we
substitute
'-^
>
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
differential geometry,
E= G and ^=0.
l/47
it
2
The curvature
of this surface
loc. cit.,
(cf.
Clebsch-Lindemami,
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.
=e
and
c)
in the surface
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.
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
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
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
when
the
hyperbolic geometry.
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
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
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
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
).
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
may
= I,
so that
= ~i log
is
The discussion
if
in
138
to be a bilinear
that /(X,
X) =
is
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
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.
from a
fixed line.
10. If
A, B,
C are
Dist (A,
=
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
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
may
be taken to be
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.
The
pendicularity, distance, etc. follow the pattern of the hyperbolic and parabolic geometries, and the same method may be used, as in those
the parabolic and hyperbolic geometries. Through a given point not on a given line
to
I
there
is
no
line parallel
geometry.
of
There
is,
parallelism to
elliptic
many
more
of the properties of
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
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
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,
is
based on Assumptions
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
Clifford parallel to n.
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
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
Any
displace-
as a product of
of Clifford
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
its
lines,
conjugate under the group of direct projective collineations transforming the sphere into itself.
and two
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
between a single elliptic plane and a double each point of the first corresponds to a pair
elliptic plane, in
which
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
we have seen
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
geometry
c
ill
UU
lUIUl U&CU Ul
4.
J.VXUICUVCA,
J.A
VYG
oou
in the formulas of
j and
- log -J 2 w^i +
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*
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
by
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
F do not
c
e.
the
geometry according to which the distances and angles are measured changes from elliptic through parabolic to hyperbolic.
For example,
if
F were
sum
TT
of the angles
were
e,
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.
of
is
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
of It* invariant
is
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
and
A'
.
(51) respectively,
.
(27)
A = B'A'BA = B'B
all
A'A,
2
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~
~^~
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
'-
>
of the last
matrix in
and
A = A'A,
n
and hence
A" = A
for the
The formulas
plication of quaternions,
's are, by 127, the same and the formulas for the
are given
by
Now let \
(31)
\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
Let
The
K (7 + 7^ + 7 y +
2 x
(S
+ B^
+ 7i + T ;/ +
;
= 7o 7 ~
/
7 3'=
(36)
,'
_i_
,'X
/'^
_ X'v.
-4-
elliptic displacements,
may
2 2
be subjected to
2 . 8
+
of (35)
+ ccl + a* = /3 + {3? + /3 + /3
2
By means
(38)
7<A+
new parameters
yi
i+
The formulas
of
the
are found
by substituting
and
r (
/3's.
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
EXERCISE
The
elliptic
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
126
it is
seen that
any
colline-
ation leaving one regulus on the absolute invariant has the matrix
/
(ta z
p.
A=
and any
-ft
^1032
ft ft
"ft
ft
^ft-ft
ift
Hence any displacement has
ft
ft
-ft
a matrix
if
(40)
it
A' and
are
.BA^B'B-A'A.
of
two displacements
BA
and B'A'
is
a displace-
ment^"^" such
and
that
...
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).
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
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
7?= 7?=
7^7,
For
(U
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,
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
^=1
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
/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
systems
of real biquaternions.
CHAPTER IX
THEOREMS ON SENSE AND SEPARATION
147. Plan
of
the
chapter.
The theorems
and
of
definitions
of
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
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
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
in
to
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
and the
interior of a
conic
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
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-
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.
of the lines
DE
and
and
AB
By
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
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 u^AAu.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
section.
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
common
2.
of
P v P2
Pn
finite in
I.
of
JP
such that
P v P z>
. ,
on the same
side of
*3.
the
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
XVII
for a
convex region.
6.
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.
segments joining
Proof.
to
the points
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
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
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
of a
Theorem
ABC.
It is also
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,
the segments
of points of R joining
to
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-
DAB, and on
ABC.
ing
of intersection of a
BCD
is
with a
line
joining
CAD
hi the set
this it
is
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
Proof.
The
points of
R on
TT
and
(3)
of
the
definition
of a planar
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.
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
R.
By
taining no point of R.
of
By Theorem
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
that (1)
I.
\vhich
is
one point.
If
and
are
and
are separated
exist
are points of R.
[0']
all
Hence there
and [P
],
0'
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
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
^4
as
vertices
consisting
only
all
of points of
R and containing
segment 0(02
.
points
/
of
the
Since
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
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
R.
The proof
space,
6.
above
to
using
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
DEFINITION.
Two
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
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
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
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
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
,
point of
Proof.
R.
If
is
any point
/,
of a,
and
a line on
Q coplanar with
R
I
points both of
containing
contains
ing
Q
of
contains a segment of
.
both of R 1 and of R 2
Hence Q
is
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,,
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.
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
;
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.
Ex.
7,
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
common
with the boundary of a convex region. a Tf a segment consists of boundary points of a given
T
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.
The theorems
We
26 and
shall
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
b together (by
Theorem
10)
The
line c
meets a and
b in the points
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,
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
line.
The boundaries
the
7
7 7 7
Eegion
Eegion
I,
II,
a, a,
/3, /3,
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
boundary
of
is
one
and
/3
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-
a,
ft,
and
7, together
and
ft,
form
boundary of the triangular region. COROLLARY 1. On any point coplanar with but not in a given
is
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
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
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
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,
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
Dualize.
152.
in
to space.
26.
The
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^,
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.
Three planes a^ a z a^ meet by pairs these lines bounds two planar convex
,
az and as form four spatial convex bounded by the three lines and
,
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
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
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.
are found
by considering succes-
26.
is
2
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.
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"
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,
families
transformations, continuous
groups,
etc,
"We
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
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
and so that
if
the pair
T'T"
is
T Tf
is
varies.
The
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]
:
P= F
is
(C),
t,
P = F (V),
is
and
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
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
THEOREM
Proof.
lt
15.
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
2,
{T^T^T^
T
T^T^,
carried
by a
projective
,
T,
val
projectivity
respectively.
which
carries
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
[P] con-
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
The
proof
precisely the
same
as that of
lemma
is
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
taining
(Theorem
1),
PQ has in
(fig.
common with R
Let
81).
be the
common
1
order
{TT'PT"}.
Let S'
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
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
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
entirely of points of R.
For
if
T and S
QP
as
is
not in R,
is
in R.
404
[CHAP. IX
The boundary
155.
Connected
if
be
a
is
connected
and only
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
set of points
[Y]
is
said to be continuous
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
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
R.
The boundary
150) of a region in a plane (space) separates (149) from the set of all points of the plane (space)
The
notion of con-
DEFINITION.
}f
of all points
on an interval
TQ T^
a line L
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)
S there
one and only one set S such that S = T(T) is at least one T such that S = F(T);
(3)
and
includ-
ing a point of
a segment
or
of
containing T'
md
such that
T is
in
o-
F(T)
is
in R.
The
ihe
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,
DEFINITION.
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].
of
elementary is a con-
The
so as
may be
to include 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
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
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
"
"
may
points
and
(e.g.
S may
and
defined in
13, Vol.
In
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-
and
if
II
TQ
is
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
Any
a
158. Affine
theorems on sense.
about
main
propositions
sense-classes
The group
of
all
projectivities
has a subgroup of direct projectivities for which a 0. This subgroup is self-conjugate, because if a transformation of the group be
>
= 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
30 and
The
generalization consists
for
A
into
sense-class
B(AB)
is
the set of
which a pair
(
jectivities
23).
30 and to w-space in
31.
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-.
^i-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
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
AB
and
are in the
same sense-class
1 1
and only
if
a
I
and
sign, a,
b,
B' respectively.
is
S(AB)
positive or
negative.
30, 31.
Similar
It
is
the plane
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 ).
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 ).
,
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
ABC
according as
S (ABCD)
The
positive or negative.
/
projectivities
= ax
-V-
b of
the Euclidean
The
direct projectivities
f.Ti <
other side.
projectivities
tivities is
From
In like
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
12
a 21
'
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
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 the other
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.
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
from
it
is
of the pair
Thus
that
AB* AB may be
(cf.
other words
23)
S(AB}=S(AB'}, and
is
Hence
it
follows that
if
AB
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
to A'B',
From
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
tinuous deformation (
B may be
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
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
Hence
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
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
{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
Let A, B,
let C'
and B' be
AB
and
84,
CA
p.
respectively.
tary transformations
this to C'BJi';
(cf.
fig.
423)
ABC
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
CAB
an elementary transformation leaves the 30, Hence, by Theorem 26, no odd permutation can be effected by elementary transformations.
27,
30,
By Theorem
If
A', B',
C'
are
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
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
if
S(ABC)^ S(A
BC
f
))
and since
even permutations
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.
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
A)
\C
A]
\A
B)'
product of
any even permutation of ABCD is a these permutations. Hence any even permutation of a
may
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)
ho
no^riori
160,161]
ELEMENTARY TRANSFORMATIONS
two points A"B"of the planes
this
of
413
carried into
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
1
,
C',D' respectively.
re-
of the last
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
it
side
of
the
line
joining
is
called
an elementary
transformation.
The
by
finite
sequences
of
elementary
transformations
is
ABC
called a sense-class
denoted
by S(ABC}.
This definition
is
in
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
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.
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
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,
, , .
point
reflections.
The product
of
two point
a translation.
(1)
+ &,
form
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
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
[CHAP. IX
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
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.
cause
of
unique beof
structure
the body.
A
is
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
it
sense-class.*
On
a Euclidean line,
such, that
to a point
S(AB)
is
is
any point
to a point
B
if
positive.
is
Such
Any
other translation
carries
called negative
is
to D,
S(CD)
negative.
Any
i.e.
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 +
S(PQP1 )
is
The inverse
of a
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
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
The inverse
of a positive
negative.
Any
rotation
is
a product of
two orthogonal
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
and
negative respec-
tively,
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
all choices of
independent
of the choice of A.
The
and
to 0'
and A'
carries 0'
and A'
is
positive
Since S(0'A'OA)
tive.
= S(OAO'A'},
Any
With
in
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
two
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.
of a
EXERCISES
1.
The
collineations
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
P P^
,
are so
chosen that
S(OP P^)
is positive.
By the
definition in
72,
0<m(l^z )<-
or
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
.
164. Sense-classes in
projective
spaces.
It
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,
(
,
a determinant
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
19
and
32.
164,165]
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
of
may be changed by
which
by
1,
multiplying the coordinates of one in a second way that there is only one
The projectivity
of a point (a , n
a ln
2l ,
a 22
in.
a prodirect
projectivities are
on one
= 0,
and those representing opposite
with that considered in
129.
projectivities
on the other
side,
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
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
;
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.
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 ,
for this
S.
of
Assumptions A, E,
Cf.
An
I
-
ABC
points
of a line
P v P
2 2
',
P
P n
carries
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
is
the
resultant of a finite
number
COROLLARY. Let
P vP
z,
P n
be
I
Two
one
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
Pv P%,
>,
P n
Jo
le
3
Pv P
z,
Pn can
made
166.
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.
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-
D into A', B
1
,
C',
Proof. It can be
trad
ABCD
FIG. 83
dered tetrad
quadrangle elementary
To complete the
proof
of
A^A A^
Z
let
A^,
Let
2,
s,
and
6
let
C and
l
{A^B^D^AJ.
let
A C
2
Ca
X>
B C
3,
D
:
(fig.
83)
By Theorem 7, Chap. II, it follows that no two of the pairs of points A^A^, A 2 A & A^A & 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
,
.
n=
1
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.
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-
by elementary transformations.
Let
C,
That the
may
and
be
seen
as
follows:
the
vertices be
let
denoted by A, B,
D,
the line
DE
ABC in a point F. This point is not on a side of the triangle ABC. Let
A
1
lines
FB
A r &**'
FC
to Bi x>
_,
Ji
and AB.
Let
FIG 84
A^ be a point in the order {BA^A'C} B l the point in which the line (fig. 84) and
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
BCADE. Thus
(ABGDE\ \BCADE)
can be effected
by elementary transformations.
is
It is easily verifiable
It can
be proved by the same methods as in Theorems 18 and 19 no four of which are coplanar can be carried
an exercise
to the reader.
When
1
this
THEOREM
if
24.
In a
S(ABCDE}-=S(A'B'C D'E'}
and only
The
D,
E into
A',
1 ,
6", D',
respectively.
of
Assumptions A, E,
and the
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
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
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
it
a triangle
WP.
slifl.ll
ABC
decom-
fvria.Ticrnlar rficrirms
wTiiV.li
rlfinnt.fi
hv the
168,169]
425
I,
notation of
is
Any
contained in a
infinity
convex
region, say
line at
of
not meeting Region I), which contains the boundary So the determination of sense for Region I
boundary and
also to a portion of
Region II
Let the sense of
ABC
positive.
The
any
segment
aries of
AB (fig.
16), is
common
I'.
to the
bound-
I and II to
I'.
and hence
I'
is
point
common
II.
and
II,
If C' is
AB
in
in Region
Hence, according to
is
S(BAC'}
positive
Region
Hence S(BAC)
Regions II and
a repetition
IV
have in
common
of -this
argument S(CAB}
with respect to
Region IV.
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
projective plane.
S(ABCDE]
(cf.
164)
be designated as right-handed,
class
is
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
down above
and,
by
167,
is
unique
Any
by
definition
set of all
an
(155), the an
infinite set of
convex regions. In
like
may
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
is
called
among
its
points
is
Two
to
a line
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
the line at
infinity.
Hence the
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)
the two sense-classes on any line are interis the point at infinity
projective plane can be carried into any other oriented line of the projective plane by a direct collineation.
By
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
/*<->.
it is
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
and
of
B ( ~^
A (+ and^
Since
A (+)
and
with respect to
are separated
by
and
B^
86
with respect to
S (+) are
Aw
m
J?
and
(cf.
,
Ex.
1,
26,
and
fig.
85).
of 8 ( -> are
A^
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.
line
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
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
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
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
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,
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
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
in
which
The correspondence T
is
on
S2
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.
is
defined in
23
for
and extended
28 will
The
definition of angle in
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
among the
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
rems in
this chapter.
third
way
indicated in
Theorems
34, 35,
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)
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
b, c
of
a pencil,
of replac-
and such
sequences denoted
between
and
The
is
transformable by
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
be
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
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
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
; ;
12
AyA^A & 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
segment complemen;
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>,
and
a',
V,
c'
are
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
angles) into
and
called a sense-class
THEOREM
A, B,
then
C,
34.
If
if
is
the
center
b,
c,
of a pencil of rays
and
S(ab)=S(cd)
and only if
S(OAB)= 8 (OCD).
THEOREM
pencil,
35.
If a and
S (ab)i=S (ba).
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
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.
ab
f<M
and
vertex as
class to
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
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
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
COROLLARY.
the orders {ab}
two rays
{ba}.
Any
three
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
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
same
DEFINITION.
lying in the
set of all
is
same plane
The common
end
is
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
to be opposite,
and the
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
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
direc-
This
justified
the notion
of
betweenness
to
directions is
THEOREM
*
If a and
J
I
(3
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
,
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 ^\> ^(cr^a-^. and <r a cr 2 <r s three segments in the directions d^ d z d 3 respec-
>
S(d^dz d^
is
which
S(Ws)THEOREM
43.
If
<r^
cr
2,
cr s
cr
s
is
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
ham
of
the
same
sense.
We now
take from
2123
and on account
Theorem 44 we have
at
once
COROLLARY
applied
to
1.
The Theorems of
the directions
of a pencil instead of
the
points of
line.
COROLLARY
2.
Two
Let
DEFINITION.
pencil;
a-
and
o2
by an elementary transformation
,
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
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
of
in the sense-class
It is
may be
restated, without
material change, in
all
common
Let a,
''ury
origin
The point
is called
;he center of
b, c
the bundle.
By
an elemen-
transformation
say
e is a,
ays,
by a ray
a'
of tlie
plane containing
md
nto
between a and
The
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,
and a'b'c' are two ordered triads of noncommon origin O and A, J3, C, A', B C' are
1
',
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
S (a'b'c'}
to
S(abc} or S(a'l'c')
=
S (acb).
THEOREM
nd
If
a, b, c
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)
to
points JL I} JC 2
S(^xz x s
A, 3,
= 8(x?F}
the interior
if
and only if
b, c
.A
4,
JL
5
If
respectively,
and
the
region
is
of
the triangle
ABC, F may
to
le
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],
= 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
_Z?,
C,
of the bundle
ABCD.
common end
is
DEFINITION.
called
a bundle of segments.
The
the segments of a
bundle
is
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 $(cr,cr * 1 O'
*
<8
Theorems 48-52 to the corresponding theorems for a bundle of segments presents no difficulty. *173. One- and two-sided regions. discussion of the order relaof
The generalization
which
is
closely analogous
both to
168
and to
Let
169
may
The
be the
containing
and contained in R,
B>
and
let a,
/3,
C
R
respec-
tively.
If 0' is
0' to
the segments of
joining
is
and
S(ap)
i
is
like
#()
A
aud
<% P
like
#(
V3
),
be
fo'/M
ti(a"ft").
region for
which a given
be two-sided.
sense-class at one
)int is
ded.
Any
other region
region
is
is
said to
A convex
gion.
two-sided.
protective plane
is
a one-sided
be any point of a three-dimensional region R. Let A, B, be the vertices of a tetrahedral region T containing and ntained in R, and let a, /3, y, B be the segments in R joining
Let
A, B, C,
If 0' is
respectively.
Then S(a/3y) = S(j3a$) = /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
is
two-sided.
172 can be
yarded as
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
dch
ABCD
is
.tions is
called a sense-class
53.
THEOREM
THEOREM
ere
a',
54.
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
line.
difference
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-
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,
&
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
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.
= 0,
> 0.
to represent
l)
is
From
the algebraic
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
D
if
is
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.
to
a point
1
A',
from A,
is opposite or direct
according as
A and A
are
the
and
COROLLARY COROLLARY
1.
2.
is direct.
COROLLARY
3.
The
is
Since an elation
the
a product of
it
COROLLARY
Since
An
elation is direct.
(
line
reflection
101)
is
homologies,
COROLLARY
5.
THEOREM
is
57.
A A
collineation leaving
three
skew
lines
invariant
direct.
Proof.
Denote the
lines
by
(
^,
2,
is effected
by
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
&,
any
l s
line meeting
2,
and
1
points on
is
F = H.
and
invariant.
Hence
H~ T
all
direct,
is direct. 1.
COROLLARY
lines
Any
collineation
leaving
all
invariant
is direct.
Proof.
Such a
collineation
of
leaves
invariant
three
skew
lines
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
5,
above, and
Proof.
The two
homology
56,
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
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
line
AB
r/37)
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 '
'hick
doubly oriented
not meeting
line,
a collineation
T(l).
itself
AB, and V
AB
is
transformed into
The by F
julus containing
lonically separated by
ll'
is
if
is
I.
leaves both
AB
and
invariant.
a/3,
on m. Hence by
74,
AF
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&
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.
are direct
Theorem
56.
Hence F
is direct.
COROLLAKY
Any
collineation
is direct.
sense-classes
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
Any
which transforms each of two skew a direct projectivity on each is direct. collineation which transforms each of two
effects
an
is
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
fundamental.
THEOREM
59.
If
the collineations
carrying a figure
into itself
all
under
the
group of
and any
be
a figure of
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
442
Let
[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
which
carried
An
an F' into an
F is
opposite.
we have
The figures conjugate
to
COROLLARY.
collineations
under
the
group of direct
non-
corollaries of
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
and complexes
(cf.
178).
Let us
now suppose
S(ABCD)
(cf,
in
162).
Any
is
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
177,178]
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
col-
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
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
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
is
opposite.
lines
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,
reguli, congruences,
and complexes.
is
right-
left-handed.
In the
first
be said
to be
be
left-handed.
THEOREM
60.
conelliptic linear
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
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
Hence
F'
is
if
F'
is
any colline-
{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.
COROLLARY
1.
The triads of
lines
of an
elliptic linear
congruence
For any
collineation.
triad
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.
said to be right-handed
it is
if
every triad
right-handed
otherwise
said to be left-handed.
(
A
is
109)
deter-
it is
mined by
142)
is
said to be right-handed or
they
belong
i.s
into a
system
a
carrying
system
right-handed
Clifford
parallels
into
THEOREM
111.
The
collincu lions
linear
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
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*
(106,
Vol. I)
changes
larly
and
I" leaves
invariant. Let
I"
I
F' he
I'
a collineation, simileaves
obtained,
The product
direct.
FFT
and
and
C invariant.
and T" are
invariant, and
Any
to
collineation leaving
C and
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
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
[CHAP.IX
The
article
Vol. II, graphierte Voiiesungen iiber nicht-Euclidische Geometrie, Chap. I, Gottingen, 1890.
EXERCISES
1.
(107, Vol.
I) as right-handed
and
left-
handed.
2.
For two
lines
f p and p
let
O X) =
is
PuP'o,
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,
42
^42
is
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,
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
equivalent to
F under
direct transformations.
Prom
is
a continuous
179,180]
ELEMENTARY TRANSFORMATION
62.
l y
447
THKORKM
plane
If
lf
l
z
copla,iar 'with
l.J.^
and such
that
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.
and
l^
(fig.
86) in points
A v A AA
y
,
taining
l
z
of intersection of
and
l^.
If
is
:
in a,
an elation with
as
as center,
points,
plane of
into
fixed
and
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
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,
which
By Theorem 62
by a
direct collineation.
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
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
DEFINITION.
perspective
1 ,
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
1
1}
l
l
z,
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
THEOREM
63.
If a doubly
,
.
oriented line
1
2
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
C respectively
which meet
and
The
triad of planes
or
l Q
0>
IJM Q ,
/
m m
z
fl
Ijn
(fig.
I 2 9
ljti Q
In the
first
case
87) let
m = 7.
l^i=a,l^ = ft,
Ijn^
= 7.
l
s
Lm
m^
respectively, a^ yQ
7j
the planes
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.,
i^i7i)'
2,
an(^ s ^ uce
^/3272 ).
But
by construction
#2/3^7,,)
C2
tive
ie.
m^
In
doubly perspecw/
()
.
tive
with
the
second
case
with
an(^
r
(A^B^
wi^1
Since
i^i7i)
^ n i-s
m = (AHC,
TOJ
is
(^a^2^2'
az i'iP^'
a/37),
and
FIG.
doubly perspective
with
m m =
o
,
fi\Y^)'i
an d since
wz^
is
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
is
THEOREM
(ACB, a
Proof.
2
65.
doubly
oriented
line
(ABC,
afii]
to
is
similarly
and
oppositely oriented
(ABC,
aB
and
Let
Q,
is
on
/ 2
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
of
which meets
let
l
l}
and
Let
,
lt
respectively and a2 A,, 72 the planes ^Om 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>
?
^i'
m V m
l
^2
,
72
l^
respectively
,
A2
2>
C2
be the points
,
respectively.
notation
it
is
is
(ACB,
a/3j).
(ABC,
(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
[CHAP IX
m = (ASC,
line of the
and
mn = (ACB,
a/3y),
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
,
on
TO.
By Theorem 64
is
1
TO.
By
TO or TO.
But
this
means that
n is
(ABC,
(ACS,
66.
ayft).
THEOEEM
larly
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)
or
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
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'
(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.
oriented line
is
either in
or in K'.
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
it is
lines
which are
similarly
if it
carries
it
carries every
doubly
Any
collineation
which
is
carries a
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
The theory
of sense-classes
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
and left-handedfigures
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
different
way
so as to apply to the
geometry of
Tl-i
all
continuous transforma-
tions instead of
merely
to the projective
Any
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
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
boundary into
itself.
If the sense-classes
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
/ 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
its
oriented curve,
negative sense-class
Any
is
one-to-one, reciprocal,
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
oriented
2-cell.
The
two
classes such
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
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
of
either
can be
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
them
invariant.
In the
first
is
not.
3-cell associated
with one
surfaces is called
an oriented
3-cell.
The oriented
3-cells of
a projective space
fall
into
two
classes
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
it is
not.
projective space
associated with
one of
its
classes of oriented
3-cells is called
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
;
rw/~*-i/-\/->-f-iTTr
rtAr\-rr\
f\4-
ntr
ot-*
-i
a vmrvtm/^
rol rvmr
"I
Q rt\
The three-dimensional
propositions
outlined
all
been proved as yet, but are (in form) one- and two-dimensional ones.
Let us note that an ordered triad
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.
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
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.
A BCD
BCD
BCD.
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^
-,
'
n
to
points
A A
, .
is called
If
A^=A n
is
said to be
otherwise
it is
is
said to
be open.
The
If
on a
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.
A A
I}
g}
>.-,
A^A 2
-
.,
A n _^A n
respectively.
-
If
denoted by A^A Z
A n _^\
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 A^A 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
I//
JJUblWO
U-0
(//(/>
C/M-yti
yi yi
^/
M,
simple polyyon
A^A 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 A^A Q
before
A nA
same as
is
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
two
points.
A class
each
of
which
is
transformable by a
number
of
elementary
ABCis
ABC
and
is
denoted by S(ABC).
There exists a one-to-one and reciprocal correspond-
THEOREM. 68.
line such that
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
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
Bi+1
respectively.
of the
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
PP
1
<.;r
ineorem
&&.
the
COROLLARY
" broken line " interval "
1.
words
substituted
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.
is
A, A n
is
An
The order
Suppose we
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
:
joining
to
the
to
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,
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
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.
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
of a simple
THEOREM
set
of points [P]
is
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
(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
an end
of
two
edges.
Let us
now
(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
may
be regarded as edges.
that
xxxin
(i9i:n. p. 37.
Now
let e
P
l
one
of its ends.
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
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
each _Z?(/=e,
+1
P _J
n
is
(1).
an end
of
e's.
Hence
all
it
satisfies
Condition
By
Condition
must include
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.
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.
-
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
on
polygon
of the
second kind.
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 A^A 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
or
A^A Z
which contains P,
point of the edge
T^,
or
any
all
A^A &
By
we
obtain a
T^
2,
Tn
each contained in
Tk
AkJrV any point of the broken line A^A^ and contains no point of the polygon A^AZ
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.
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
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
of
intersection of the
lines
l
lf
/
2
and
to
have no
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
,
^i>
B all
in the
set of
lines
of their points
-=
~ 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
lines.
The number
is
al
n (n
1),
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
,
1
l^,
2
The number
a:,
satisfies
n(n
1}
Proof.
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)
-+
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
regions, each of
which
is
the set of
points
3).
common
to
k+:
and
By Theorem
8 each
of
two con-
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
^
-+
-
1 convex regions.
2>
z,
by
or any one of
points determine
a1
or
a-
...
a1
1
;
a2
or
any one of
them by a 2
2.
COROLLARY
line
If
the
lines
l
lf
1
2
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,
Proof.
1
2,
,
The given
n
which we
shall call
I,
meets the
lines
l
lt
in at least
two
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
Its
lj
Let
l{
and
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
are separated
therefore not a
boundary point
of the
aa
it
the proof of
COROLLARY
not in the a
2
3.
Any
an a 2
.
to
a point
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
,
,
The theorem
is
a direct consequence of
;
151 in case
i.e.
nS.
it
we assume
n=k
and prove
it
for
= k + 1.
1}
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
segment may be
R.
taken as vertices of the polygon p which by hypothesis bounds Thus we may denote this polygon by A^A Z A AjA i} where
{
and
which
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
AA
Z
a
or
i}
A^.;
,
if
it
contains
AA^
it
it
AA
A^^,
l
and since
;
con-
tains
AA
{
AAA
{
AAj
it
since it
AA^ AJ _ Aj _
S 2,
if
,
it
contains
t,
A i+l A
and
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
the polygon A^AjAj^ and those on the edge AA^ are in one of the two
all points of
manner
R on any
nnp
-nnint. -Frirm
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,
boundary
of
must
i
"be
>
A^A^
of the
polygon A^A 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 A^Aj 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
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 A^A Z In like manner the polygon is the boundary of A^A5 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
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.
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
z,
',
determine a
of
a' points,
a[
linear
a[ planar
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
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
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
cases.
of
row being
element
row and
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
's
(Theorem
each row
of
con-
of the
a1
'
a let
and
We
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
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
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 )
(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 >
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
, ,
of
symbol
'
(x^,
.
xz
xa^) which
is
the
sum
is
-,
p
2
>
tt
Let us
>
now
inquire what
(%!>
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
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 %>
'>
(cf.
151)
(5)
(mod. 2)
+tf4 =(X
(4)
We
column
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^.
in
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
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.
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
number
all
of solutions in a
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
it
follows that
in
if
we add
all
the
0.
(H^
which
all
On the other hand, if we add any subset of sum will be a solution in which not all the
the columns of
the
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
subset.
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
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.
l^
l
z
n is
a polygon,
it
(H
a ),
but
it
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
of
,
n,
2,
we have a
But since
by
is
edges are a
of a set of lines l^
l
,
Two
is
is
or is not a
first
sum
of a subset of
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
[CHAP.IX
columns
-
of
>,
H
c
3
2,
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 ,
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
:
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
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
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].
[P] can be joined by a broken line conand, since every point of p is on the
any point
is
in [P].
Consider
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
Ak
[P], and
of
hence
Q would
is
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
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
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
to
<x,
point of the
a point of
point
the polygon.
COKOLLAEY
on
1.
Any
of
can
be joined 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,
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
,
T
is
are in [Q],
and vice
versa.
THEOREM
bol
76.
.,
Any
)
simple polygon
not the
(x^ xz
aja
which
sum
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
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
p with which
^,
l
-
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
of an
UJJJL>
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
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
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
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
we can show
that
common
with
By
p z can be
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
COROLLARY
polygon
2.
p
in
is
meeting
an even
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
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
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
,
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
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
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
2,
s
,
n.
l
n
,
is
l
interior to a triangular
1
2
n,
boundary point
Let af,
seen, all
,
<x 's 2
's
are in R.
its
-,
R
(r
or on
I,
boundary
that
is
is,
of
an
a,
may be
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
containing
contains
474
not have
[CHAF.IX
B on
its
boundary, any
*
cfl
having a point in
common with
is
T' is one of
?,,?.
1
on
of
.,
R
a
and
its
boundary
81.
is
identical
with the
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]
. .
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.
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
and
polyhedron analogous to
matrix,
the element
77^
is
or 1 according as a?
is
not or
is
an end of
aj.
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
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
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
THEOREM
82.
The points of space which are not upon any one of a 7rn fall into a finite number a of convex s
,
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,
planes
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.
all concurrent,
any
line through
I of one
the
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
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
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
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
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
.
of
a,
a1
(1).
Hence
the
The matrices
lt
and
8.
The
relations
among
may
,
we
shall call
H^ H
H^foJ),
*
= 1,
2,
y = 1,
of aj.
2,
or is not
an end
according as a\
is
H = W),
2
= 1,
is
or
-,
oTjj
-,
and
of #?.
is
H.
*
=
;
= 1,
is
2,
.,
a2
y = l,
2,
or
aj.
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
, ,
=
'
if
and only
sets of
if
(#
lS
xa ^
is
sum
of
two
polygons (modulo
2) will be
478
[CHAP. IX
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
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,
-,,).
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
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 ).
8,
there
is
set
of
linear
1,
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
The discussion
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
152,
tf
3
it is
= 4,
a 1==12, a g
= 16,
=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
,
In case
w>4,
the columns of
fall into
;
(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 .
2
.
's
which have no
a 1 in
Any column
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
s
.
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
met by 7rn Hence the row of H 2 just two 1's in columns of Class
.
's
determined by
TT^
7r
2,
.,
irn _ v
of
every a
ir^
of irn is
2, .,
two
of the planes
7r
7r B _j
an edge
of
?r n .
No
of
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
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,
lJ
deter-
7T
.,
a 2 'a determined
by
TT^
7T
2,
ir n
,r
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
for
TT^
7r
ir n
and aQ
of
is the
7r
2,
same
7r n
as the difference
of
al
for
Wj,
and
a
for
7r
x,
7r
2>
tion of
7rn
increases o^
Since a1
=r
for
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
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
of poly-
(regardTT
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.
set
a
'
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
number
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.
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
is eit/ier
oaa or even.
sum, modulo
2,
of the boundaries
of
t
a. The
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
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
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
sum
of
two
for
solutions of
(H 2 )
is
a solution of
(H 2 ).
is
The
set of polyhedra
sum
of the
symbols
2, of j? x
two
polyhedra p t and
pz
is
called
the the
sum, modulo
and
As
polygons,
p is a set p and pz
1
of polyhedra
.
whose
faces are in
or in
<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
Since any
a set of
as
1 of the
columns
of
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
linearly dependent on
the
a& -,
1
.
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,
and
polyhedra which
bound convex
Proof.
edges,
Let
vrn
be a
set of
and
"fl-
expressible
as a
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
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.
. ,
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
we may
a *,
the
a*,
-,*.
of the
is also
sum
boundaries of
because the
sum
of all the
columns of
is
(0, 0,
-,
0).
There
is
no other
of
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
THEOREM
polyhedron.
Any
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
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.
segment, interval, broken line, polygon, two-dimensional convex region, or three-dimensional convex region associated with a senseclass
among
its
points
line,
interval,
broken
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
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
Salu. to
uv
tively related to s
To
we may think of
is
an.
oriented segment as
is
marked with
which
at the
end which
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
on the edge A 1A 2 are said to agree in the order {A^ABA^ 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
segment obtained
the negative of s\ may be indicated
sense-class of si
of the s
's
by
s/
and
call it
,
to the points
a,
call
a
ft
by means
of a
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,
sj.
and
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
Let
(ic 1 ,
xz
-,
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
a^
The
ments.
sense-class of
an oriented polygon agrees with a definite sensesides and thus determines a set of oriented seg(x^,
The symbol
xz
is
0i )
may
also be regarded as a
symbol
Each
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
sides
a{ and
a,..
The
course be zero.
linear equation
Hence the symbol (a? 1 must satisfy the ) whose coefficients are given by the row of E x cor,
If s {
a
and
,
are oppositely
C{
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.
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
be
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
186,
it
follows that
(E^ in integers represents one or more directed simple polygons. The situation here differs from that described in
any solution
side may enter into more than one polygon and the same polygon may be counted any number of
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
is
o^
tf
+l. In view
of
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
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
two columns
else
the elements
cancel.
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
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
a 2 in one
determined by l^
to this triangular
The elements corresponding to edges interio region must all cancel, because they cannot canct
2,
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
,
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
Hence in every
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
pp
sf,
,
where
E p
2,
.,
is
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
to
is
the columns of
2
Hence the
T//S
Since any two having an edge in points not on p can be joined by a broken line not meeting p
correspond-
(Theorem
:
If
they are
all
obvious that p
= 2h
THEOREM
the
93.
is expressible
form
(8)
*P=$<*1
e i is
where each
+1
may
or
1.
This theorem
the matrix
E2
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
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
is
also
all
(7).
Aside
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
's
in
it
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
, ^
Hence we have
the
THEOREM form
(9)
94.
The symbol
is expressible in
?=%'& i=l
e { is
where
or -f
or
1.
The
regions
R referred
to
in Theorem 75.
DEFINITION.
By
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(OA^A 2 ) ^(O'A^AJ (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
OA
S,
etc.
sense-class in
R,
which we shall
call positive,
determines a
i.e.
positive sense-class in
R'
containing R,
the
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
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-
polygons
which have
in
A m and A A B
vertices
l
Bm
A and A and
the points of the edge A^A 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 A^A 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
,
R.
of the a?'s
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 A^A 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 (A^AA^ and for
so that
of
l
.
y^
the boundary of
bols
(aj l}
R2
is
S (A^AA^. Hence
(y^ yz ,
-,
the
sum
of the
two sym-
x2
xa ) and
y a)
I}
A^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
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
. . , ,
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
94,
it
follows (as
is
obvious
is
that
even.
the
convex region
The argument
modification to
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
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
-
segments
si
1
,
Sz
1
,
,
sai :
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
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
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
the
a?'s
197).
DEFINITION.
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
and one
for
which
this
assignment of sense-classes
not possible
is
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
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
a z or
1.
By
reference to
195 we find
is
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
if
C and
si of
points
points
and
OCD. Hence
Hence the
that
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
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.
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
of the
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
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
197.
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
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
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.
= 1,
2,
.,
al j
= 1, 2,
to,
.,
az
and
e| is 1,
1,
is
positively related
negatively related
to, or
of sf.
E2
each column
be a symbol
by changing some
2
,
to -1's in such a
-
way
l!
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.
197),
it is
CE } \ g/
V
.7
y j jj e?..
UJ
(i
\
=l
-*-j
2,
~*)
')
=1
a} l/*
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
?
if
fc
=l
and
and
if
is
#A =
and
eik
1.
Hence
there are as
many terms
equal to
+1
s}.
If neither s
s\
2
fc
nor
its
is
in
if
sg
on
boundary,
the
Mi
term
of this
t
equation
is zero, for in
first
case xk
=
,
and
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^
}
197.
= 1,
2,
o;
3
and
e?.
is
+1,
1,
or
according
as s f
is
boundary
1's
The matrix
by changing
resulting
solu-
to
1's in
the columns of
in such a
and therefore
This
is
possible
by Theorem
96.
exercise, the
E8
from the
of
H
E
x
The sum
of
of the
columns
is
contains one
+1
and one
1.
relation
among
the columns of
because this would imply, on more than one linear relation among the columns
1,
is
x
1,
of solu-
tions of
which
all
the solutions
a1
a +l.
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
'
c
i>
'
>
ck
the
s 's
(10)
e c
1
+ A +"-+*fc a=2
where
I is
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,
*-, +
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
,
same
a3
plane.
Since the
number
#2
(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
The
form
500
e?.
t
[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
by a
s?
solution in
which the
and
s? to
which
Sj 3
s?
and
to
which
to
s? is negatively related
and
if it
includes that
Hence the
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.
sf
1,
by
!,,
s
2
1)
is
a solution
of
With the
3
each
is
and negatively
We thus
T
7r
I}
,
have
98.
7r n
THEOREM
-,
If each of
a,
in
of
its sense-classes to
&
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
is
a generalization of an
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
,
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
;
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
;
H
I,
H, 11
,
H, 33
80
J, 7
K, 3
;
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-
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,
vertex
of,
191
limiting points of pencils of, 159; nine-point, linearly dependent, 250 pencils of, 169, 233 orthogonal, 101
;
power
;
of a point
of,
with
245
direct,
248
U/iu/nv^l Wll u.^ V^iXJJ.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
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
"
asymptotes
of,
of.
73
73;
central,
73;
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,
;
algebraic, 15
De
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;
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
;
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
;
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
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
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
33, 34
limit
projectivity
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
lines, 442,
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
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
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,
;
294
plane, 294
rotation, 299 translation, 317
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
14 circular
transfor-
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
;
;
;
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
:
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
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
_
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
;
; ; ; ;
;
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.
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
; ; ;
;
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
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
; ; ;
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
;
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
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
; ;
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
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,
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
; ;
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
;
Symmetry,
123, 124, 129, 138, 297, 300, 352, 373 plane of, 295 ; with respect to a point, 300
;
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
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,
; ;
negative, 417
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
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; ; ; ; ; ;
Upper
side of a cut, 14
Vacuous statements, 24
Vailati, G., 44
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-
327;
isotropic,
;
negative, 410
positive, 416
317;
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
; ; ;
similar,
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
PROJECTIVE GEOMETRY,
Norbert Wiener expressed
1964 by M.
his
Volumes
and
II
T. Press) as follows:
recommended
to
me two mathe-
summer
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
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
Advanced Study,
He
is
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
geometrical systems, characterized by properly chosen postulate systems, and which, within the framework of such, gives a treat-'
ment
$7.50