QA

C37

GERSTM

Digitized by tine Internet Arcliive

in

2007

witli

IVIicrosoft

funding from

Corporation

littp://www.arGliive.org/details/elementsofnoneucOOcarsuoft

:^

.ungmans' /IDobern ^atbematical Series

THE ELEMENTS OF
NON-EDCLIDEAN PLANE GEOMETRY

AND TRIGONOMETRY

BY

H.

CARSLAW

S.

Sc.D. (Camb.), D.Sc. (Glaso.)

PROFESSOR OF MATHEMATICS IN THE UNIVERSITY OF SYDNEY
FORMERLY FELLOW OF EMMANUEL COLLBOK, CAMBRIDOB

fV/Tff

DIAGRAMS

LONGMANS, G fee^N A "NtF C of*^

'

'"

39

7fc

PATERNOSTER ROW, LONDON

FOURTH AVENUE &

30th

STREET,

NEW YORK

BOMBAY, CALCUTTA, AND MADRAS

1916

O /,

J
\JC

%\
\

PREFACE
In

book

have attempted to treat the Elemenl>8

Geometry and Trigonometry in such
a way as to prove useful to teachers of Elementary Geometry
in schools and colleges.
Recent changes in the teaching of
Geometry in England and America have made it more than
ever necessary that the teachers should have some knowledge
of the hypotheses on which Euclidean Geometry is built, and
this little

of Non-Euclidean Plane

on which Euclid's Theory of

treatment of the Theory
of Parallels leads naturally to a discussion of the Non-Euclidean
Geometries and it is only when the logical possibility of these
Non-Euclidean Geometries is properly understood that a
teacher is entitled to form an independent opinion upon the
teaching of Elementary Geometry.
The first two chapters of this book are devoted to a short
discussion of the most important of the attempts to prove
Euclid's Parallel Postulate, and to a description of the work
of the founders of Non-Euclidean Greometry, Bolyai, Lobatschewsky and Riemann.
In Chapters III.-V. the Non-Euclidean Geometry of Bolyai
and Lobatschewsky, now known as the Hyperbolic Geometry,
The feature of this
is developed in a systematic manner.
treatment is that in Chapter III. no use is made of the Principle
of Continuity, and that both the Geometry and the Trigonometry of the Hyperbolic Plane are built up without the use
of Solid Geometry.
especially of that hypothesis
Parallels

rests.

The

historical

PREFACE

vi

In Chapters VI. -VII. a similar treatment, though in

detail, is

less

given for the Elliptic Geometry.

Chapter VIII. deals with Poincare's representation of the

Non-Euclidean Geometries by the geometry of the families
of circles orthogonal or diametral to a fixed circle.

From

these representations an elementary proof of the impossibility

of proving Euclid's Parallel Postulate can be obtained,

and

they throw fresh light upon the Non-Euclidean Geometries

themselves.

book could never have been written had it not

It was from him that I first
learnt that an elementary treatment of the subject was possible.
Both to his historical work, an English translation of which
I had the privilege of undertaking, and to his article in Enriques'
This

little

been for the work of Bonola.

Questioni

riguardanti

geometria

la

especially

elementare,

in

extended form in the German edition of that work, this

its

book owes a very great deal.

The other writers on the same subject to whom I am most
indebted are Liebmann and Stackel. The treatment of Plane
and to the
Hyperbolic Trigonometry is due to Liebmann
;

second edition of his well-known Nichteuklidische Geometrie,

as well as to his original papers,

me

as they appeared^ I

most

which he has sent to

of

am much indebted. A similar acknow-

ledgment is due to Stackel. When he learnt that I was engaged

on this work, I received from him, in the most generous way,
a set of

papers on the subject,

all his

me

many

and the

of which were

a copy of his
book on Wolfgang and Johann Bolyai, immediately on its
publication, allowed me to make some use of his final account
of the discovery of the Hyperbolic Geometry in reading my
inaccessible to

in Australia

gift of

proofs.

be found in their proper place

would mention here the frequent
use I have made of Halsted's work and^of the Bibliography
also the assistance which I have received
of Sommerville

Other acknowledgments

in the text.

However,

will

PREFACE
from Dr. F.

many

S.

valuable

Macaulay, who read

suggestions

vii
all tlie

proofs

and amendments.

and made
The work

of another of the Editors of this Series, Mr. C. S. Jackson,

has

made my labour

Sydney, Mr. R.

J.

lighter,

and one

of

my

colleagues in

Lyons, has also read a great part of the

final proofs.
II.

Sydney,

CARSLAW.

September, 1914.

NOTE.
The

final proofs of this

book had been corrected, and the

foregoing preface written and set up in type before the

outbreak of the war.

In the course of years v^hetinie\m^ come

when such

co-

operation as I have here acknowledged will again be possible.

H.
Sydney, January, 1916.

S.

C.

CONTENTS
CHAPTER

I.

THE PARALLEL POSTULATE, AND THE WORK

OF SACCHERI, LEGENDRE AND GAUSS.
L

Euclid's treatment of parallels

2.

The

3.

Some problems

4.

Two theorems

5.
6.
7.
8.

9.

10.
11.
12.

------

principle of continuity

of construction

5
-

10

work of Saccheri
work of Legendre
Postulate of Archimedes

independent of the Parallel Postulate

The controversy regarding the Parallel Postulate

The
The
The
The

FAOa

12
15

and the

Parallel Postulate

work of Gauss
Gauss and Schweikart
Gauss and Taurinus
Gauss and Schumacher

CHAPTER

18

19
21

23
26

II.

THE WORK OF BOLYAI, LOBATSCHEWSKY, AND RIEMANN,

THE FOUNDERS OF THE NON-EUCLIDEAN GEOMETRIES.
14.

John Bolyai, and his father Wolfgang

Bolyai's Appendix

15.

Bolyai's later years

16.

The work of Lobatschewsky

17.

Lobatschewsky' s Principles of Qeomeiry

Gauss, Bolyai, and Lobatschewsky

13.

18.

86 19-20.

The work

of

Riemann

27

28

30
32
33
36
38

CONTENTS

CHAPTER

III.

THE HYPERBOLIC PLANE GEOMETRY.

...

21. Lobatschewsky's treatment of parallels

22. Hubert's Axiom of Parallels
23-25. Some theorems on parallels
26. Properties of the figure formed by two parallel rays through
two given points and the segment joining these points

27.

The angle of

parallelism

28.

Saccheri's Quadrilateral

The quadrilateral with two right angles

30. The quadrilateral with three right angles
31. The sum of the angles of a triangle
29.

common

....
-

Not-intersecting lines have a

33.

Parallel lines are asymptotic

34.

The

36.

The correspondence between a right-angled

36.

The

perpendicular

The

52
54

58
triangle

and a

69

63

Proper and Improper Points

40.

52

56

series of associated right-angled triangles

The perpendiculars

51

two not-intersecting lines is

and on each side of this the

quadrilateral with three right angles

39.

47

53

32.

37-38.

40
42
43

50

shortest distance between

their common perpendicular,
lines continually diverge

FAQG

66

to the sides of a triangle at their middle

points are concurrent

68

Parallel Constructions

71

Given p, to find n{p)

44. Construction of a common parallel to two given straight
lines in one plane
46. Given n{p), to find p
-'
46-47. Corresponding points 48. The Limiting-Curve or Horocycle
49. The Equidistant-Curve
Equivalent polygons
50. The Measurement of Area.
51. Equivalent triangles
62-53. The areas of triangles and polygons

71

41-43.

....

CHAPTER

74
76
77

80
82
84
85
88

IV.

THE HYPERBOLIC PLANE TRIGONOMETRY.

Some theorems on concentric limiting-curves
The equation of the limiting-curve

64-56.

67.

91

97

CONTENTS

xi

59.

The hjrperbolic functions of complementary segments

The equations connecting the sides and angles of a

60.

Corresponding equations for an oblique-angled triangle

58.

right-

angled triangle

100
-

61. The measurement of angles

103

104

62.

The trigonometrical functions, and the fundamental equation

63.

The

of hyperbolic trigonometry

105

relations connecting the sides

and angles

of a triangle

108

(cant.)

64. The angle of parallelism

65.

PAGB
98

109

The Euclidean formulae hold true

in infinitesimal

geometry

on the hyperbolic plane

109

CHAPTER

V.

MEASUREMENTS OF LENGTH AND AREA, WITH THE

AID OF THE INFINITESIMAL CALCULUS.
The element of arc
The element of arc
68. The element of arc
66.

in Cartesian coordinates

112

67.

in polar coordinates

114

in limiting-curve coordinates

"

69. Applications of these formulae

70. The element of area in limiting-curve coordinates
71. The element of area in Cartesian coordinates
72. The element of area in polar coordinates
73.

Th6 area of a

triangle

116
118

119

122

123

and of a quadrilateral with three

right angles

124

CHAPTER

VI.

THE ELLIPTIC PLANE GEOMETRY.

74.

Plane geometry when the straight

of a line

line is

not

infinite

75.

The pole

76.

All lines have the

77.

The two elliptic geometries

.
.
The sum of the angles of a triangle
Saccheri's quadrilateral, and the quadrilateral with three

78.
79.

same length

127

129

right angles

127

131

132

I34

CONTENTS

xii

CHAPTER

VII.

THE ELLIPTIC PLANE TRIGONOMETRY.

80-83.

Gerard and Mansion's treatment of the Non-Euclidean

trigonometrical formulae
84.
86.

86.

136

The function <p(x) is continuous

The functional equation <p{x + y) + <p{x -y) = 2<l>{x)<f){y)
The function 0(a;) and the cosine

143
-

87. The formula cos v = cos T cos T

K
K
Ic
h

88. Thefonnula tanT=cos

146

145
146

Atanr

149

89.

The other trigonometrical formulae

160

90.

Further results

163

CHAPTER

VIII.

THE CONSISTENCY OF THE NON-EUCLIDEAN GEOMETRIES AND THE IMPOSSIBILITY OF PROVING

THE PARALLEL POSTULATE.
91.

method of proving that the Non-Euclidean Geometries

are consistent

92-93. Poincar^'s representations of the

153

Non-Euclidean Geo-

metries

The system of circles passing through a fixed point,

and the Euclidean Geometry
97-101. The system of circles orthogonal to a fixed circle, and

154

94-96.

the hyperbolic geometry

The impossibility of proving the Parallel Postulate .103. The system of circles cutting a fixed circle diametrally, and
the Elliptic Geometry
104. Is the Euclidean Geometry true ?

102.

156
160

170
171

174

Index of Names of Authors

176

Subject Index

177

NO:^r-EUCLIDEAX GEOMETRY.
CHAPTER

I.

THE PARALLEL POSTULATE, AND THE WORK OF

SACCHERI, LEGENDRE AND GAUSS.
1. By the term Non-Euclidean Geometry we understand
a system of Geometry built up without the aid of the Euclidean
Parallel Hypothesis, while it contains an assumption as to
parallels incompatible with that of Euclid.
The discovery that such Non-Euclidean Geometries are
logically possible was a result of the attempts to deduce
Euclid's Parallel Hypothesis from the other assumptions
which form the foundation of his Elements of Geometry.
It will be remembered that he defines Parallel Lines as follows
:

Parallel straight lines are straight lines which, being in the

same plane and being produced indefinitely in both directions,

do not meet one another in either direction*

Then

in I. 27

he proves that

// a straight line falling on two straight lines make the alternate angles equal to one another, the straight lines will be parallel
to one another.

And

in I.

28 that

If a straight line falling on two straight lines make the exterior

angle equal to the interior and opposite angle on the same side,

*Here and in other places where the text of Euclid's Elements is

quoted, the rendering in Heath's Edition (Cambridge, 1908) is adopted.
This most important treatise will be cited below as Heath's Euclid.
N.-E.G,
A

NON-EUCLIDEAN GEOMETRY

or the interior angles on the

same

[ch

i.

side equal to two right angles,

the straight lines will be parallel to one another.

In order to prove

namely

(I.

tlie

two propositions,

converse of

tliese

on parallel

straight lines

29), that

straight line falling

makes

the

alternate angles equal to one another, the exterior angle equal to

the interior

same

and opposite

angle,

and

the interior angles

on

the

side equal to two right angles,

he found

it

parallel lines,

necessary to introduce

the hypothesis as to

which he enunciates as follows

If a straight line falling on two straight lines make the interior

angles on the same side less than two right angles, the two straight
lines, if produced indefinitely, meet on that side on ivhich are
the angles less than the two right angles.
This hypothesis we shall refer to as Euclid's Parallel PostuIt is true that in some of the MSS. it finds a place among
In others it is one of the Postulates, and it
the Axioms.
seems to belong more properly to that group. No use is made
of it in the earlier propositions of the First Book. Accordingly
these would find a place in the Non-Euclidean Geometries,
which differ only from the Euclidean in substituting for
his Parallel Postulate another incompatible with it. Other
theorems of the Euclidean Geometry will belong to the NonEuclidean, if they have been proved, or can be proved, without
the aid of the Parallel Postulate, and if these geometries
adopt the other assumptions, explicit and implicit, made by
late.

Euclid.

2. It is not within the scope of this book to discuss the

modern treatment of the assumptions on which the Euclidean
and Non-Euclidean Geometries are based. We shall deal
simply with the assumption regarding parallels. But it is right
to mention that the idea of motion or displacement, which
forms part of the method of superposition, itself involves an
axiom. The fourth proposition of Euclid's First Book now
finds a place among the Axioms of Congruence, and upon this
group of axioms the idea of motion is founded. Apparently
Euclid recognised that the use of the method of superposition
was a blot upon the Elements. He adopted it only in 1. 4, and

EUCLID'S ASSUMPTIONS

1, 2]

from employing it in other places, where it would

have shortened the demonstration.
Again, Postulate I., which asserts the possibility of drawing
a straight line from any one point to any other, must be held
refrained

to declare that the straight line so drawn is unique, and that

two straight lines cannot enclose a space. And Postulate II.,
which asserts the possibility of producing a finite straight line
continuously in a straight line, must also be held to assert that
the produced part in either direction is unique in other words,
that two straight lines cannot have a common segment.
;

But the following more fundamental and distinct assumpmade by Euclid, without including them among the
axioms or postulates
tions are

(i)

That a

straight line is infinite.

This property of the straight line is required in the proof of

16.
The theorem that the exterior angle is greater than
either of the interior and opposite angles does not hold in the
Non-Euclidean Geometry in which the straight line is regarded
as endless, returning upon itself, but not infinite.
I.

(ii)

let

a,

Let A, B,

and
ABC, and not passing

be three points, not lying in a straight line,

be a straight line lying in the plane

through any of the points A, B, or C. Then, if a passes through

a point of the segment AB, it must also pass through a point of the
segment BC, or of the segment AC {Pasch's Axiom).

From this axiom it can be deduced that a ray passing through

an angular point, say A, of the triangle ABC, and Ijdng in the
region bounded by AB and AC, must cut the segment BC.
(iii) Further, in the very first proposition of the First Book
of the Elements the vertex of the required equilateral triangle
is determined by the intersection of two circles.
It is assumed
that these circles intersect.
similar assumption is made in
I. 22 in the construction of a triangle when the sides are given.
The first proposition is used in the fundamental constructions
of 1. 2 and I. 9-11.

Again, in I. 12, in order to be sure that the circle with a

given centre will intersect the given straight line, Euclid makes
the circle pass through a point on the side opposite to that in
which the centre lies. And in some of the propositions of
Book III. assumptions are made with regard to the intersection of the circles employed in the demonstration. Indeed

NON-EUCLIBEAN GEOMETRY

[ch.

i.

right through the Elements constructions are effected by means

of straight lines and circles drawn in accordance with Postulates

Such straight lines and circles determine by their interI. -III.

section other points in addition to those given
and these
points are used to determine new lines, and so on. The existence of these points of intersection must be postulated or
proved, in the same way as the existence of the other straight
Unes and circles in the construction has been postulated or
;

proved.
The Principle of Continuity, as it is called, is introduced to
fill this gap.
It can be stated in different ways, but probably
the simplest is that which Dedekind originally adopted in discussing the idea of the irrational number. His treatment of
the irrational number depends upon the following geometrical

axiom

of a straight line can he separated into two

such that every point of the first class is to the left of every
point of the other class, then there exists one, and only one, point
which brings about this division of all the points into two classes,
this section of the line into two parts*

//

all the points

classes,

The assumption of
than an axiom by which we assign

This statement does not admit of proof.

this property is nothing less
its

continuity to the straight

line.

Postulate of DedeJcind, stated for the linear segment, can

be readily applied to any angle, (the elements in this case
being the rays from the vertex), and to a circular arc. By
this means demonstrations can be obtained of the theorems
as to th intersection of a straight line and a circle, and of
a circle with another circle, assumed by Euclid in the pro-

The

above mentioned.")* The idea of continuity was

adopted by Euclid without remark. What was involved in the
assumption and the nature of the irrational number were
unknown to the mathematicians of his time.
This Postulate of Dedekind also carries with it the important
positions

* Dedekind,
schweig, 1892)

Stetigkeit
;

und

irrationale

English translation by

Zahhn,

Beman

p. 11

(2nd ed., Braun-

(Chicago, 1901).

t This question is treated fully in the article by Vitali in Enriques'

volume, Queationi rigiiardaiiii la geomeiria elementare (Bologna, 1900)
German translation under the title, Fragen der Elementargeometrie,
See also Heath's Euclid, vol. i. p. 234.
vol. i. p. 129 (Leipzig, 1911).

THE PRINCIPLE OF CONTINUITY

2,3]

Postulate of Archimedes, which will be frequently referred to in

the following pages
:

// two segments are given, there is always some muUijple of

the one ivhich is greater than the other.*

3. An interesting discovery, arising out of the recent study

of the Foundations of Geometry, is that a great part of Elementary Geometry can be

built up without the Principle of

Continuity. In place of the construction of Euclid I. 2, the
proof of which depends upon this Principle, the following
Postulate "] is made
:

// A, B are two points on a straight line a, and if A' is a point

upon the same or another straight line a', then we can always find
on the straight line a', on a given ray from A', one and only one
point

B',

such that the segment

AB

is

congruent

to the

segment

A'B'.

In other words, we assume that we can always set off a

given length on a given line, from a given point upon it,
towards a given side. By the term ray is meant the half-line
starting from a given point.
With this assumption, for Euclid's constructions for the
bisector of a given angle (I. 9), for the middle point of a given
straight line (I. 10), for the perpendicular to a given straight
line from a point upon it (I. 11), and outside it (I. 12), and,
finally, for an angle equal to a given angle (I. 23)
all of which,
in the Elements, depend upon the Principle of Continuity
we may substitute the following constructions, which are
independent, both of that Principle and of the Parallel

Postulate. J

* For the proof of the Postulate of A rchiinedes on the assumption of

Dedekind's Postulate, see Vitali's article named alwve, 3. Another
treatment of this question will be found in Hill)ert's GrHudtayen der
Geometrte, 3rcl ed. 8.
An English translation of the first edition was
made by Townsund (Chicago, 1902). The Postulate of Archimedes
stated above for linear segments is adopted also for angles, areas, and
volumes.
tCf. Hilbert, he.

The

cit.

3rd ed.

5,

Axioms

of Congruence.

constructions in Problems 1, 2, 3 and 5 are given by Halsted

in his book, Rational Geometry (2nd ed. UX)7).
Those for Problems 4
and 6 in the text are independent of the Parallel Postulate, and
replace those given by Halsted, in which the Euclidean Hypothesis is
J

assumed.

NON-EUCLIDEAN GEOMETRY

Problem

Construction.

[CH.

1.

To bisect a given angle.

On one of the lines bounding the given angle A take

any two points B, C.

the other bounding line take AB' = AB and
Join BC and B'C.
Let them intersect at D.
Then AD is the desired bisector.

On

AC'=AC.

The triangles BAG' and B'AC are congruent.

Proof.
Therefore Z.ACB'= Z.AC'B and ^DBC=^DB'C'.
It follows that the triangles BDC and B'DC are congruent, since

BC = B'C'.
DB' = DB.

Therefore
Finally the triangles
the given angle.

Problem

2.

BAD

and B'AD are congruent, and

To draw a perpendicular to a given straight

Construction.

Let

Take any other

AB

be the given straight

straight line

AC

AD

bisects

line.

lin?.

through

A.

Upon AB

take

AD ^ AC.

Join CD.

and

let

the

On AB take AF = AG, and on the ray

take AH = AD.
Join FH.
Then FH is perpendicular to AB.

AG

Bisect ;^CAD (by Problem

at G.
bisector cut

1),

CD

From the triangles ACG and ADG, we have Z.AGD equal to

Proof.
a right angle.
Also the triangles AGD and AFH are congruent.
Therefore Z.AFH=: Z.AGD = 1 right angle.

PROBLEMS OF CONSTRUCTION

3]

Problem

At a given point on a given straight line to erect the

3.

perpendicular.

Let A be the given point and

Comtruction.
the perpendicular ZOY (by Problem

Draw

OY = OZ,

Take

the given straight line.

meeting

BC

in O.

join AY and AZ.

through A to X.

Produce YA

and

Bisect Z-XAZ by AD (by Problem

Then AD is the perpendicular

I).

to

BC

By the construction, the

Proof.
angles OAZ and GAY are congruent.

tri-

through

BC
2),

A,

Therefore

L ZAO = L Y AG

But

LD^Z=L\^D.

=
Therefore

AD

is

Z.XAC.

perpendicular

Fig.
t(j

3.

BC.

Pkoblem 4. From a given point

draw the perpendicular to the line.

outside a given straight line to

CoiiMrucflon.
Let A, B be two points on the given line, and
point outside it.
Join AC and BC.
On the segment AB take a point D, and
(by Problem 3) draw the perpendicular at

to

the

AB.

By

Pasch's Axiom, this lino nnist cut

AC or
Let it cut AC, and let the point of
intersection be E.
Produce ED through D to F, so that

BC

either

DE = DF.
Join AF and produce AF to G, such
that AG = AC.
Join CG, and let it be cut by AB, or AB
produced, at H.
Then CH is the required perpendicular.
I'roni the construction, the triProof.
F'o- *
angles ADE and ADF are congruent, so
that AB bisects /.CAG.
It follows that tlie triangles ACH and AGH are congruent, and that
Z.AHC is a right angle.

Problem 5. At a given point on a giYen straight line to

angle equal to a given angle.

make an

Let A be the point on the given line a. (Cf. Fig. 5.)

be the given (acute) angle.
From a point E on one of the lines bounding the angle, draw (by
Problem 4) the perpendicular EF to the other bounding line.
Canstniction.

Let

NON-EUCLIDEAN GEOMETRY
On Aa take AC=DF.
At C erect the perpendicular Cc
Make BC = EF, and join AB.

Aa

to

(by Problem

[CH.

I.

3).

C a
Fig.

5.

By the construction, the triangles

Proof.
congruent.
Therefore
L BAG = Z. E D F.
Problem

To bisect a given

6.

Let

finite

DEF

and

ABC

are

straight line.

AB

be the given segment.

draw the perpendicular Bfo to AB (by

Construction.

At B
Problem

3).

Upon B/) take any point C and join AC.

At B make Z.AB"E= Z.BAC (by Problem
Let the

5).

BE

iLADB

Bisect

Problem

line

cut AC at D.
by the line cutting

AB

at

F (by

1).

Then F

is

the middle point of AB.

From the construction it follows that

Proof.
the triangles ADF and DBF are congruent.

Fig.

0.

AF=FB.

Thus

Noli.
This construction has to be slightly modified for the Elliptic
Geometry. The point C must lie between B and the pole of AB.
[Cf. 78.]

4. Two
Postulate.
1.

The

Theorems

independent of the

Parallel

'perpendicular to the base of any triangle through its

is also perpendicular to the line joining the middle

middle point
'points

Let

of the two sides.

ABC

be any triangle, and

let

F and E be

tlie

middle

points of the sides AB and AC.

Join F and E
and draw A A', BB', and CC' perpendicular
to EF from A, B, and C.
Let H be the middle point of BC, and K the middle point
of B'C.
;

Join HK.

TWO THEOREMS

3,4]

We

prove that HK is perpendicular to BC and EF.

the triangles AFA' and BFB', which are congruent,
have AA' = BB'.
shall

From
we

= CC'.

Similarly

AA'

Therefore

BB' = CC'.

Join BK and KC.

In the triangles BB'K and CC'K

we have
BB'

= CC',

and the angles

B'K

at B'

= C'K,

and

C are equal.

Therefore the triangles are congruent, and BK = CK.

Again, in the triangles BHK and CHK, we have the three
sides equal, each to each.
Therefore the triangles are congruent, and

L BHK = L CHK = a right angle.

Also ^BKH=^CKH.
But, from the triangles BB'K and CC'K, we have

iBKB'=^CKC'.
Therefore l HKB' = ^ HKC' = a right angle.
Thus HK is perpendicular to both BC and EF.
2. The locus of the middle points of the segments joining a set
of points ABC... on one straight line and a set A'B'C'... on
another straight line is a straight line, provided that AB = A'B',

BC =

B'C', etc.

Via.

8.

Let M, N, and P be the middle points of AA', BB', and CC'.

Join BM and produce it to B", so that BM = M B".
Join B"A' and B"B'.

NON-EUCLIDEAN GEOMETRY

10

[ch.

r.

sides of the triangle BB'B" are bisected at M and N.

Therefore the line bisecting B'B" at right angles is also
perpendicular to MN.

The

But

this line bisects

Now

z.

B'A'B", since A'B'

= A'B".
= B'C'.

produce A'B" to C", so that B"C" = BC

Join C' C",

MC" and MC.

MAC and MA'C" are congruent, and it follows

that MC and MC" are in one straight line.
Since A'C' = A'C", the line bisecting C'C" at right angles
coincides with the line bisecting B'B" at right angles.
Therefore MN and MP are perpendicular to the same
The

triangles

straight line.

Therefore MNP are coUinear.

Proceeding to the points A,
corresponding result, and in this

B,

D, A',

B',

D'

way our theorem

we have a
is

proved.

5. From the Commentary of Proclus * it is known that not

long after Euclid's own time his Parallel Postulate was the
The questions in dispute remained
subject of controversy.
unsolved till the nineteenth century, though many mathematicians of eminence devoted much time and thought to
Three separate problems found a place
their investigation.
in this discussion
:

deduced from the other

assumptions on which Euclid's Geometry is based ?
(i)

Can the

Parallel Postulate be

(ii) If not, is it an assumption demanded by the facts of

experience, so that the system of propositions deduced from
the fundamental assumptions will describe the space in which

we

live

And

finally, are both it and assumptions incompatible

with it consistent with the other assumptions, so that the
adoption of the Euclidean Hypothesis can be regarded as an
arbitrary specialisation of a more general system, accepted not
because it is more true than the others, but because the
Geometry founded upon it is simpler and more convenient ?
(iii)

There can be
that the
negative.
* Of.

libmm

first

The

little

doubt that Euclid himself was convinced

of these questions must be answered in the

place he assigned to the Parallel Postulate and

Friedlein, Prodi Diadochi in primiim Eudidis elementorum

Also Heath's Eudid, vol. i.
comvientarii (Leipzig, 1873).

Introduction, chapter

iv.

EUCLID'S PARALLEL POSTULATE

4,5]

his refusal to use it earlier

him

it

than

I.

11

29 are evidence that with

had only the value of an hypothesis.

It

seems at

least

very probable that he realised the advantage of proving

without that postulate such theorems as could be established
independently
just as he refrained from using the method
of superposition, when other methods were available and
sufficient for the demonstration.
But the followers of Euclid were not so clear sighted. Fruitless attempts to prove the Parallel Postulate lasted Well into
the nineteenth century. /Indeed it will be surprising if the
use of the vicious directio7i-theory of parallels, advocated at
present in some influential quarters in England, does not raise
another crop of so-called demonstrations the work of those
who are ignorant of the real foundations on which the Theory
;

of Parallels rests.

The assumption involved in the second question had also

an effect on the duration of the controversy. Had it not been
for the mistake which identified Geometry
the logical doctrine
with Geometry the experimental science the Parallel
Postulate would not so long have been regarded as a blemish
upon the body of Geometry. However, it is now admitted
that Geometry is a subject in which the assertions are that
such and such consequences follow from such and such premises. Whether entities such as the premises describe actually
exist is another matter. Whenever we think of Geometry as

a representation of the properties of the external world in

which we live, we are thinking of a branch of Applied Mathematics.
That the Euclidean Geometry does describe those
properties we know perfectly well. But we also know that it
is not the only system of Geometry which will describe them.
To this point we shall return in the last pages of this book.
In the answer to the third question the solution of the
problem was found. This discovery will always be associated
with the names of Lobatschewsky and Bolyai. They were the
first to state publicly, and to estabhsh rigorously, that a consistent system of Geometry can be built upon the assumptions,
explicit and implicit, of Euclid, when his Parallel Postulate is
omitted, and another, incompatible with it, put in its place.
The geometrical system constructed upon these foundations is
as consistent as that of Euclid.
Not only so, by a proper
choice of a parameter entering into it, this system can be made
to describe and agree with the external relations of things.

NON-EUCLIDEAN GEOMETRY

[ch.

i.

This discovery, wMch was made about 1823-1830, does not

detract from the value of Euclid's work.
The Euclidean
Geometry is not to be replaced by the Non-Euclidean Geome-

The latter have thrown light upon the true nature of

Geometry as a science. They have also shown that Euclid's
Theory of Parallels, far from being a blot upon his work, is
" When
one of his greatest triumphs. In the words of- Heath
we consider the countless successive attempts made through
more than twenty centuries to prove the Postulate, many of
them by geometers of ability, we cannot but admire the genius
tries.

of the man who concluded that such a hypothesis, which he

found necessary to the validity of his whole system of geometry,

^was really indemonstrable." *

6. The Work of Saccheri (1667-1733).

The history of these attempts to prove the Parallel Postulate
does not lie within the scope of this work.f But we must
refer to one or two of the most important contributions to that
discussion from their bearing on the rise and development of

r^
I

\
\

J
\

''

the Non-Euclidean Geometries.

Saccheri, a Jesuit and Professor of Mathematics at the
University of Pavia, was the first to contemplate the possibility of hypotheses other than that of Euclid, and to work
out the consequences of these hypotheses. Indeed it required
only one forward step, at the critical stage of his memoir, and
the discovery of Lobatschewsky and Bolyai would have been
anticipated by one hundred years. Nor was that step taken
by his immediate successors. His work seems to have been
quickly forgotten. It had fallen completely into oblivion when
the attention of the distinguished Italian mathematician
Beltrami was called to it towards the end of the nineteenth
century.
His Note entitled " un precursore italiano di Legendre
e di Lobatschewsky " % convinced the scientific world of the
importance of Saccheri's work, and of the fact that theorems,
which had been ascribed to Legendre, Lobatschewsky, and
Bolyai, had been discovered by him many years earlier.
*

Heath's

Eiiclid, vol.

i.

p. 202.

tCf. Bonola, La. geometria iioneudidea (Bologna, 1906); English

translation (Chicago, 1912).
In quoting this work, we shall refer to
the English translation.

XHend. Ace. Lincei

(4),

t.

v.

pp. 441-448 (1889).

SACCHERI

5, 6]

13

EucUdes db omni ncevo vindicatus

It was published in 1733, the last
year of his life. Much of it has been incorporated in the
elementary treatment of the Non-Euclidean Geometries. A
great deal more would be found therein were it not for the fact
that he makes very frequent use of the Principle of Continuity.
It must not be forgotten that Saccheri was convinced of the
He discussed the contruth of the Euclidean Hypothesis.
tradictory assumptions with a definite purpose not, like
Bolyai and Lobatschewsky, to establish their logical possibilit)
but in order that he might detect
the contradiction which he was persuaded
must follow from them. In other words,
he was employing the reductio ad absurdum
argument.
The fundamental figure of Saccheri is
the two right-angled isosceles quadrilateral
^
ABDC, in which the angles at A and B are ^
right angles, and the sides AC and BD equal.
It is easy to show by congruence theorems that the angles
at C and D are equal.
[Cf. 28.]
On the Euclidean Hypothesis they are both right angles.
Thus, if it is assumed that they are both obtuse, or both acute,
Saccheri's

is

now

little

book

easily accessible.*

r"

the Parallel Postulate is implicitly denied.

Saccheri discussed these three hypotheses under the

names

The Hypothesis of the Right Angle

z. C = z. D = a right angle.
The Hypothesis of the Obtuse Angle ... LC = LD=an obtuse
. . ,

angle.

The Hypoth'jsis of

the

Acute Angle

...

LC=LD=an

acute

angle.

He showed

that

According as the Hypothesis of the Right Angle, the Obtuse

Angle, or the Acute Angle is found to be true, the sum of the
angles cf any triangle will be respectively equal
or less than tivo right angles.

to,

greater than,

Also that

If the sum of the angles of a single triangle

than, or less than two right angles, then this
*Cf. Engel

u. Stackel,

auf Gaiiss, pp. 31-136

is

equal

to,

sum mil

greater

be equal

Die Theorie der ParcUlellinien von Euclid

(Leipzig, 1895).

bis

NON-EUCLIDEAN GEOMETRY

14
to,

[ch.i.

greater than, or less than two right angles in every other

triangle.

Again, he showed that

The Parallel Postulate follows from, the Hypothesis of the Right

Angle, and from the Hypothesis of the Obtuse Angle.

He was
Angle

thus able to rule out the Hypothesis of the Obtuse

since, if the Parallel Postulate is adopted, the

sum

of

the angles of a triangle is two right angles, and the Hypothesis

of the Obtuse Angle is contradicted. It should be remarked
that he assumes in this argument that the straight lijae is
infinite.
When that assumption is dropped, the Hypothesis
of the Obtuse Angle remains possible.
As we have already mentioned, Saccheri's aim was to show
that both the Hypothesis of the Acute Angle and that of the
Obtuse Angle must be false. He hoped to establish this by
deducing from these hypotheses some result, which itself
would contradict that from which it was derived, or be inconsistent with a previous proposition. So, having demolished
the Hypothesis of the Obtuse Angle, he turned to that of the
Acute Angle. In the system built upon this Hypothesis,
after a series of propositions, which are really propositions in
the Geometry of Lobatschewsky and Bolyai, he believed that
he had found one which was inconsistent with those preceding
He concluded from this that the Hypothesis of the Acute
it.
so that the Hypothesis of the
Angle was also impossible
Right Angle alone remained, and the Parallel Postulate must
;

be true.
In his belief that he had discovered a contradiction in
the sequence of theorems derived from the Hypothesis of the
Acute Angle, Saccheri was wrong. He was led astray by the
prejudice of his time in favour of the Euclidean Geometry as
the only possible geometrical system. How near he came to
the discovery of the Geometry of Lobatschewsky and Bolyai
will be clear from the following description of the argument
contained in his Theorems 30 to 32
He is dealing with the pencil of rays proceeding from a
point A on the same side of the perpendicular from A to a given
line h, and in the same plane as that perpendicular and the line.
:

considers the rays starting from the perpendicular AB

right angles to AB.
In addition to the last ray AX, he shows that, on the hypo-

He

and ending with the ray AX at

SACCHERI AND LEGENDRE

6,7]

15

Acute Angle, there are an infinite number of rays

which have a common perpendicular with the line h. These
rays obviously cannot intersect the line b.
There is no last ray of this set, although the length of the
common perpendicular decreases without limit but there is
thesis of the

a lower limit to the set.

Also, proceeding from the line AB, we have a set of rays which
but there
intersect the line 6. There is no last ray of this set
is an upper limit to the set.
The upper limit of the one set and the lower limit of the other,
he showed to be one and the same ray.
Thus, there is one ray, the line a^, which divides the pencil
of rays into two parts, such that all he rays on the one side
and all
of the line a^, beginning with AB, intersect the line b
the rays on the other side of the line a^, beginning with the
line AX, perpendicular to the line AB, do not intersect 6.
The line a^ is the boundary between the two sets of rays, and
is asymptotic to b.
The result which Saccheri obtained is made rigorous by the
introduction of the Postulate of Dedekind. According to that
postulate a division of the two classes such as is described above
carries with it the existence of a ray separating the one set of
lines from the other.
This ray, which neither intersects 6 nor has with it a common
perpendicular, is the right-handed (or left-handed) parallel of
Bolyai and Lobatschewsky to the given line.
;

7.

The Work of Legendre

(1752-1833).

The contribution

Like
of Legendre must also be noticed.
Saccheri, he attempted to establish the truth of Euclid's
Postulate by examining in turn the Hypothesis of the Obtuse
Angle, the Hypothesis of the Right Angle, and the Hypothesis
of the Acute Angle. In his work these hypotheses entered as
assumptions regarding the sum of the angles of a triangle.
If the sum of the angles of a triangle is equal to two right
angles, the Parallel Postulate follows ; at any rate, if we
assume, as Euclid did, the Postulate of Archimedes.*
Legendre thus turned his attention to the other two cases.
He gave more than one rigorous proof that the sum of the
angles of a triangle could not be greater than two right angles.
*Cf. Heath's Euclid, vol.

i,

pp. 218-9.

NON-EUCLIDEAN GEOMETRY

16

In these proofs the infinity of the

them

is

as follows

line is

assumed.

[ch

One

i.

of

Let the sum of the angles of the triangle ABC be tt + ol, and
A be the smallest angle.
Bisect BC at D and produce AD to E, making DE = AD.
Join BE.

let

Then from the triangles ADC

and BDE, we have

C^

lCad=l bed,
^ACD=^DBE.
Thus the sum of the angles of
the triangle AEB is also equal to
TT + oc, and one of the angles

BAD

AEB

or

is

__
/\

than or

less

equal to | ^CAB.
Apply the same process to the triangle ABE, and we obtain
a new triangle in which one of the angles is less than or equal
to

lCAB,

while the

sum

is

again

-tt

+ ol.

way after n operations we obtain a triwhich the sum of the angles is tt + ol, and one of the

Proceeding in this
angle, in

angles

is less

than or equal to

^CAB.

But we can choose n so large that 2"a>^CAB, by the

Postulate of Archimedes.
It follows that the sum of two of the angles of this triangle
is greater than two right angles, which is impossible (when the
length of the straight line is infinite).
Thus, we have Legendre's First Theorem that
The sum of

the angles of

a triangle cannot he greater than two

right angles.

Legendre also showed that

If the

sum

of the angles of a single triangle is equal to two

sum of the angles of every triangle is equal

right angles, then the

to

two right angles.

From

these theorems

it

follows that

If the sum of the angles of a single triangle is less than two

right angles, then the sum of the angles of every triangle is less
than two right angles.

LEGENDRE

7]

All these results

had been obtained many years

17
earlier

by

Saccheri.

Legendre made various attempts to prove that the sum cannot be less than two right angles, even in a single triangle
but these efforts all failed, as we now know they were bound to
do.
He published several so-called proofs in the successive
editions of his text-book of geometry, the Moments de GeomHrie.
All contained some assumption equivalent to the hypothesis
which they were meant to establish.
For example, in one he assumes that there cannot be an
absolute unit of length * an alternative hypothesis already
noted by Lambert (1728-1777).t
In a second he assumes that from any point whatever, taken
within an angle, we can always draw a straight line which
will cut the two lines bounding the angle.
In a third he shows that the Parallel Postulate would be
true, if a circle can always be drawn through any three points
not in a straight line.
In another [cf. p. 279, llth Ed.] he argues somewhat as follows
A straight line divides a plane in which it lies into two congruent parts. Thus two rays from a point enclosing an angle
less than two right angles contain an area less than half the
plane.
If an infinite straight line lies wholly in the region
bounded by these two rays, it would follow that the area of
half the plane can be enclosed within an area itself less than
half the plane.
Bertrand's well-known " proof " (1778) of the Parallel Postulate { and another similar to it to be found in Crelle's Journal
They
(1834) fail for the sani^ reason as does Legendre's.
depend upon a comparison of infinite areas. But a process of
reasoning which is sound for finite magnitudes need not be
valid in the case of infinite magnitudes. If it is to be extended
to such a field, the legitimacy of the extension must be proved.
Lobatschewsky himself dealt with these proofs, and pointed out
the weakness in the argument. First of all, the idea of congruence, as applied to finite areas, is used in dealing with infinite regions, without any exact statement of its meaning in
this connection. Further
and here it seems best to quote his
;

* See below, p. 90.

Also Bonola,

loc. cit. 20.

t Cf. Engel u. Stackel, loc. cit. p. 200.

jCf. Frankland, Theories of Parallelism, p. 26 (Cambridge, 1910).
N.-B.O.
B

NON-EUCLIDEAN GEOMETRY

18

[ch.i.

" when we are dealing with areas extending to

we must in this case, as in all other parts of mathematics, understand by the ratio of two of these infinitely great
numbers, the limit to which this tends when the numerator and

own words

infinity,

denominator of the fraction continually increase." *

It is not a little surprising that at the present day mathematicians of distinction have been found quoting Bertrand's
argument with approval.
"j"

8. Both Legendre and Saccheri, in their discussion of

these hypotheses, make use of the axiom that the length of the
straight line is infinite, and they also assume the Postulate of
Archimedes. Hilbert | showed that the Euclidean Geometry
could be built up without the Postulate of Archimedes. Dehn
investigated what effect the rejection of the Postulate of
Archimedes would have on the results obtained by Saccheri
and Legendre. He found that the sum of the angles of a triangle can be greater than two right angles in this case. In
other words, the Hypothesis of the Obtuse Angle is possible.
Again, he showed that without the Postulate of Archimedes
we can deduce from the angle-sum in a single triangle being
two right angles, that the angle-sum in every triangle is two
right angles.
But his most important discovery was that,
when the Postulate of Archimedes is rejected, the Parallel
Postulate does not follow from the sum of the angles of a
triangle being equal to two right angles.
He proved that
there is a Non-Archimedean Geometry in which the angle-sum
in every triangle is two right angles, and the Parallel Postulate
does not hold.
His discovery has been referred to in this place because it
shows that the Euclidean Hypothesis is superior to the others,
which have been suggested as equivalent to it. Upon the
Euclidean Hypothesis, without the aid of the Postulate of
Archimedes, the Euclidean Geometry can be based. If we
* Cf. Lohatsehewsky, Nein Principles of Geometry icith a Complete
Theory of Parallels, Engel's translation, p. 71, in Engcl u. Stiickel's
Urkunden zur Geschichte der nichieuklidischen Geometrie, I. (Leipzig,
1898).

tCf. Frankland, The Mathematical Gazette, vol. vii. p. 136 (1913)

Nature, Sept. 7, 1911, and Oct. 5, 1911.
p. 332 (1914)

and
:J:

Cf

loc. cit.

chapter

iii.

Cf. Math. Ann. vol.

liii.

p.

404 (1900),

GAUSS AND BOLYAI

7,8,9]

19

substitute for it the assumption that the sum of the angles of a

triangle is two right angles
or that the locus of the points
equidistant from a straight line is another straight line
One of these is the
different geometries can be created.

Euclidean Geometry, in which only one parallel can be drawn

to a straight line from a point outside

Dehn calls
number of

parallels

The

9.

Though

it.

Another

is

what

the Semi-Euclidean Geometry, in which an infinite

Work
Bolyai

can be drawn.*

of Gauss (1777-1855).
were

and Lobatschewsky

the first to
publicly announce the discovery of the possibility of a NonEuclidean Geometry and to explain its content, the great

German mathematician Gauss had also independently, and

some years earlier, come to the same conclusion. His results
had not been published, when he received from Wolfgang
Bolyai, early in 1832, a copy of the famous Appendix, the
work of his son John.
This little book reached Gauss on February 14, 1832. On
the same day he wrote to Gerling, with whom he had been
frequently in correspondence on mathematical subjects f
"... Further, let me add that I have received this day a
little book from Hungary on the Non-Euclidean Geometry.
In
I find all my own ideas and RESULTS, developed with
it
remarkable elegance, although in a form so concise as to offer
considerable difficulty to anyone not familiar with the subject.
The author is a very young Austrian officer, the son of a friend
:

of my youth, with whom, in 1798, I have often discussed these

However at that time
matters.
ideas were still only
slightly developed and far from the completeness which they
have now received, through the independent investigation of

my

young man. I regard this young geometer v. Bolyai as a

genius of the highest order. ..."
The letter in which Gauss replied to Wolfgang Bolyai three
weeks later is better known, but deserves quotation from the
this

throws upon his own work J

If I commenced by saying that I am unable to praise
this work (by John), you would certainly be surprised for a
moment. But I cannot say otherwise. To praise it would be to
light it

"

* Cf. Halsted, Science, N.S. vol. xiv. pp. 705-717 (1901).

tCf. Gauss, Werke, vol.

ij:

Gauss, Werke, vol.

viii. p.

viii. p.

220.

220.

NON-EUCLIDEAN GEOMETRY

20

Lch.i.

praise myself.

Indeed the whole contents of the work, the path

taken by your son, the results to which he is led, coincide
almost entirely with my meditations, which have occupied my
mind partly for the last thirty or thirty-five years. So I
remained quite stupefied. So far as my own work is concerned,
of which up till now I have put little on paper, my intention
was not to let it be published during my lifetime. Indeed the
majority of people have not clear ideas upon the questions of
which we are speaking, and I have found very few people who
could regard with any special interest what I communicated to
them on this subject. To be able to take such an interest one
must have felt very keenly what precisely is lacking, and about
that most men have very confused ideas. On the other hand,

was my idea to write all this down later, so that at least it

should not perish with me. It is therefore a pleasant surprise
for me that I am spared this trouble, and I am very glad that
it is just the son of my old friend who takes the precedence of
me in such a remarkable manner. ..."
Wolfgang sent a copy of this letter to his son with the

it

remark

" Gauss's answer with regard to your work is very satisfactory, and redounds to the honour of our country and nation.
A good friend says, That's very satisfactory " *
John Bolyai was the reverse of pleased. That he would
be disappointed at the news that Gauss had already reached
the same conclusions as himself was natural. But his chagrin
led him to doubc whether Gauss had really made these discoveries independently of his work. He conceived the absurd
idea that his father must have sent his papers to Gauss some
time earlier (they had been in his hands for several years),
and that Gauss had made use of them, jealous of being beaten
by this young Hungarian. In this he relied upon a remark
made by Gauss in 1804, in a letter to his father, when both
of them were trying to demonstrate the Parallel Postulate.
Wolfgang had sent him what he thought was a rigorous proof,
and Gauss replied that his demonstration was invalid, and that
he would try as clearly as possible to bring to light thestumbling!

*Cf. Stackel, "Die Entdeckung der nichteuklidisehen Geometrie

durch Johann Bolyai," Math. u. Naturwiaseyiachaftliche Berichte aws
Ungarn, Bd. xvii. p. 17 (1901). Also by the same author in Engel u.
Stackel's Urkunden zur Geschichte dfr nichleiildidiachen Geometrie, II.,
Wolfgang u. Johann Bolyai, vol. i. p. 72 (Leipzig, 1913).

GAUSS AND SCHWEIKART

9.10]

21

That this was not unlike the

obstacle which so far had baffled his own efforts. " However,
I am always hopeful," he added, "that some day, and that
in my own lifetime, a way over this obstacle will be revealed." *
block which he found therein.

Though John Bolyai afterwards saw how groundless his

had treated him
badly in this matter and it does seem unfortunate that Gauss
did not more effectively use his great influence to rescue from
ill-merited neglect the notable work of the two comparatively
unknown young mathematicians, Bolyai and Lobatschewsky.
Not till years after they had passed away did the scientific
world realise the immense value of their discoveries.
suspicions were, he always held that Gauss
;

10. Bolyai's discovery was made in 1823, and first published in 1832. Far away in Kasan, Lobatschewsky
one of
the Professors of Mathematics in the local University not

than 1829, and probably as early as 1826, had also discovered this new Geometry, of which the Euclidean was a

later

special case.

Thus

it is

interesting to trace, so far as

we

Gauss's attitude to the Theory of Parallels at that time.

chief available authorities are

some

letters of his

and some notes found among

survive,

which

can.

The
still

his papers.f

In the early years of the nineteenth centiiry he shared the

belief that a proof of the Euclidean Hypothesis might
possibly be found. But in 1817 we find him writing to Olbers

common

as follows
" Wachter has published a little paper on the First Principles of Geometry,' of which you will probably get a copy
through Lindenau. Although he has got nearer the root of
the matter than his predecessors, his proof is no more rigorous
than any of the others. I am becoming more and more
convinced that the necessity of our geometry cannot be
:

proved ..." J
In 1819 he learnt from Gerling in Marburg that one of his
colleagues, Schweikart
a Professor of Law, but formerly a
keen student of Mathematics had informed him that he was
practically certain that Euclid's Postulate could not be proved
without some hypothesis or other
and that it seemed to him

Gauss, Werlce,

vol. viii. p. 160.

t See Gauss, Werke, vol.

J Gauss, Werke, vol.

viii.

viii. p.

177.

NON-EUCLIDEAN GEOMETRY

22

[ch.i.

not improbable that our geometry was only a special case of a

more general one. At the same time Gerling sent him, at
Schweikart's request, a Memorandum, which the latter had
given him, desiring to know Gauss's opinion upon it.
This Memorandum is as follows *
:

" Marburg, December, 1818.

" There are two kinds of geometry a geometry in the

the Euclidean
and an astral geometry.
" Triangles in the latter have the property that the sum of
their three angles is not equal to two right angles.
" This being assumed, we can prove rigorously
strict sense

(6)

That the sum of the three angles of a triangle is less

than two right angles
That the sum becomes always less, the greater the area

(c)

That the altitude of an

(a)

of the triangle

isoscele(' right-angled triangle

continually grows, as the sides increase, but it can
never become greater than a certain length, which I
call the Constant.

" Squares have, therefore, the following form (Fig. 11)

" If this Constant were for us the radius of the earth (so
that every line drawn in the universe from one fixed star to
another, distant 90 from the first, would be a tangent to the
surface of the earth), it would be infinitely great in comparison
with the spaces which occur in daily life.
*

Gauss, Werkt, vol.

viii. p.

180.

GAUSS. SCHWEIKART,

10, 11]

" The Euclidean geometry holds

the Constant is infinite. Only in
three angles of every triangle are
and this can easily be proved, as

Constant

is

AND TAURINUS

23

only on the assumption that

this case is it true that the

equal to two right angles

soon as we admit that the
;

infinite."

This document is of peculiar importance, as it is in all

probability the earliest statement of the Non-Euclidean Geometry.
From a passage in a letter of Gerling's,* we learn that
Schweikart made his discovery when in Charkow. As he left
that place for Marburg in 1816, he seems by that date to have
advanced further than the stage which Gauss had reached
in 1817, according to the letter quoted above.
To Gerling, Gauss replied as follows f
"... Schweikart 's Memorandum has given me the greatest
pleasure, and I beg you to convey to him my hearty congratulations upon it. It could almost have been written by
myself. (Es ist mir fast alles aus der Scele geschiieben).
I would only fm'ther add that I have extended the Astral
Geometry so far, that I can fully solve all its problems as soon
as the Constant = C is given, e.g. not only is the DefectJ of the
angles of a plane triangle greater, the greater the area, but it is
exactly proportional to it
so that the area has a limit which
it can never reach
and this limit is the area of the triangle
formed by three lines asymptotic in pairs. ..."
From Bolyai's papers it appears that at this date he was
attempting to prove the truth of the Parallel Postulate. Also
in 1815-17 Lobatschewsky was working on the same traditional
:

lines.

11. The above Memorandum is the only

kart's on the Astral Geometry that is known.

work of Schwei-

Like Gauss, he
seems not to have published any of his researches. However,
at his instigation, and encouraged by Gauss, his nephew
Taurinus devoted himself to the subject. In 1825 he published a Theorie der Parallsllimen, containing a treatment of
Parallels on Non-Euclidean Lines, the rejection of the Hypothesis of the Obtuse Angle, and some investigations resembling
those of Saccheri and Lambert on the Hypothesis of the Acute
Angle. For various reasons he decided that the Hypothesis of
*Cf. Gauss, Werke, vol.
t Gauss, Werke, vol.

viii. p.

viii. p.

238.

181.

J See

p. 54.

24

NON-EUCLIDEAN GEOMETRY

[ch.

i.

the Acute Angle must also be rejected, though he recognised the

logical possibility of the propositions which followed from it.
Again, it is from a letter which Gauss wrote to Taurinus in
1824, before the publication of his book, that we obtain the
fullest information of his views *
" Your kind letter of the 30th October with the accompanying little theorem I have read not without pleasure, all the
more as up till now I have been accustomed to find not even a
trace of real geometrical insight in the majority of the people
who make new investigations upon the so-called Theory of
Parallels.
In criticism of your work I have nothing (or not
much) more to say than that it is incomplete. It is true that
your treatment of the proof that the sum of the angles of a
plane triangle cannot be greater than 180 is still slightly
lacking in geometrical precision. But there is no difficulty in
completing this
and there is no doubt that that impossibility can be established in the strictest possible fashion. The
position is quite different with regard to the second part, that
the sum of the angles cannot be smaller than 180. This is
the real hitch, the obstacle, where all goes to pieces. I imagine
that you have not occupied yourself with this question for
long. It has been before me for over thirty years, and I don't
believe that anyone can have occupied himself more with this
second part than I, even though I have never published anyThe assumption that the sum of the three
thing upon it.
angles is smaller than 180 leads to a peculiar Geometry, quite
For
distinct from our Euclidean, which is quite consistent.
myself I have developed it quite satisfactorily, so that I can
solve every problem in it, with the exception of the determination of a Constant, which there is no means of settling a priori.
The greater we take this Constant, the nearer does the geometry
approach the Euclidean, and when it is given an infinite value
The theorems of that Geometry appear
the two coincide.
almost paradoxical, and to the ignorant, absurd. When considered more carefully and calmly, one finds that they contain
nothing in itself impossible. For example, the three angles of
a triangle can become as small as we please, if only we may take
the sides large enough ; however, the area of a triangle cannot
exceed a definite limit, no matter how great the sides are
taken, nor can it reach that limit. All my attempts to find a
:

* Cf. Gauss, Werke, vol. viii. p. 186.

This letter is reproduced in
facsimile in Engel u. Stackel's Theorie der ParalleUinien (Leipzig, 1895).

GAUSS. TAURINUS,

11, 12]

AND SCHUMACHER

25

an inconsistency, in this Non-Euclidean Geomehave been fruitless. The single thing in it, which is opposed
to our reason, is tliat if it were true, there must exist in space
a linear magnitude, determined in itself (although unknown to
contradiction,

try,

us).

But methinks,

in spite of the meaningless

Word- Wisdom

of the Metaphysicians, we know too little or nothing at all

about the real meaning of space, to stamp anything appearing

unnatural to us as Absolutely Impossible. If the Non-Euclidean

Geometry were the true one, and that Constant were in some
ratio to such magnitudes as we meet in om- measurements on
the earth or in the heavens, then it might be determined a
Thus I have sometimes in jest expressed the
posteriori.
wish, that the Euclidean Geometry were not the true one,
because then we would have a priori an absolute measure.
" I have no fear that a man who has shown himself to me
as possessed of a thinking mathematical head will misunderstand what I have said above. But in every case take it as
a private communication, of which in no wise is any public
use to be made, or any use which might lead to publicity.
Perhaps, if I ever have more leisure than in my present circumstances, I may myself in the future make my investigations

known."

12.

Finally, in 1831, after Bolyai's

Appendix was

in print,

but before a copy had reached him, we find Gauss writing to

Schumacher, who thought he had proved that the sum of the
angles of a triangle must be two right angles, by a method
practically the same as the rotation method of Thibaut, which
so unfortunately has lately received official sanction in England
and crept into our text-books of Elementary Geometry. He
pointed out to him the fallacy upon which that so-called proof
rests.
Then he added *
" In the last few weeks I have commenced to put down a
few of my own meditations which are already to some extent
forty t years old. These I had never put in writing, so that I
have been compelled three or foiir times to think out the
whole question afresh. Nevertheless I did not want it to
perish with me."
:

* Cf.

Gauss, Werke,

vol. viii. p. 213.

t Forty years before the date of this letter Gauss would be just a
little over 14 years old
!

NON-EUCLIDEAN GEOMETRY

26

[ch.

1.

12]

The Notes on Parallels,* found among his papers, probably

belong to this period. Some use of them will be made below in
the formal development of the Geometry of Bolyai and
Lobatschewsky.
However his plans were changed when, in February, 1832,
Bolyai's work reached his handS.
He saw that it was now
unnecessary for him to proceed with this work. The enthusiasm with which he read the Appendix we have already seen.
I have entered at some length into this story, partly because -a,
of its intrinsic interest
partly because of the urifortunate
\
claim made by some mathematicians that to Gauss should
be ascribed the discovery of the Non-Euclidean Geometry ;
/
partly, also, because it has been suggested that the work of
/
Bolyai and Lobatschewsky had been inspired by the investiga(
tions of Gauss.
|
The claim and the suggestion we now know to be unfounded.
The wonderful discovery, which revolutionised the science of
Geometry, must always be associated with the names of Bolyai
and Lobatschewsky, who, independently and without any
knowledge of the work of Gauss, fully developed the new
Geometry. While the glory of the discovery is theirs, we must
not forget the advance which Gauss, and also Schweikart, had
made along the same lines.
;

Cf. Gauss, Werke, vol.

viii. p.

202

also Bonola, loc.

cit.

p. 67.

CHAPTER

II.

THE WORK OF BOLYAI, LOBATSCHEWSKY, AND RIEMANN.

THE FOUNDERS OF THE NON-EUCLIDEAN GEOMETRIES.
13. John Bolyai (1802-1860).
As we have already seen, John Bolyai, a Hungarian officer
in the Austrian army, had in 1823 built up a consistent system
of geometry in which the Parallel Postulate of Euclid was
replaced by another, contradictory to the former. His discovery was published in 1832 as an Appendix to his father's
work
Tentamen juventutem studiosam in elernenta tnatheseos

purae, elementaris ac sublimioris, methodo intuitiva, evidentiaque huic propria, introducendi. This work is usually referred
to as the Tentamen. The title of the Appendix contributed by
the son, and placed at the end of vol. i. of the Tentamen, is
Appendix. Scientiam spatii absolute veram exhibens a veritate
aut falsitate Axiomatis XI Euclidei (a priori lumd unquani
decidenda) independenlem
adjecta ad casum falsitatis, qtiadratura circuli geometrica.
Auctore Johanne Bolyai de eadem,
Geotnetrarum in Exercitu Caesareo Regio Austriaco Castrensium
Capitaneo.
If we omit the title page, a page explaining the notation,
and two pages of errata, the Appendix contains only twentyfour pages.
Bolyai's discovery was made as early as 1823, when he was
but 21 years old. At the time, we find him writing to his
father as follows *
" I have resolved to publish a work on the theory of parallels,
as soon as I shall have put the material in order, and
cir:

my

* Stackel u. Engel, " Gauss die beiden Bolyai und die nichteuklidische
Geometrie, Math. Ann. vol. xlix. p. 155 (1897). Also Stackel, loc. cit.
vol.

i.

p. 85.

NON-EUCLIDEAN GEOMETRY

28

[ch.h.

cumstances allow it. At present I have not yet completed this

work, but the road, which I have followed, has made it almost
certain that the goal will be attained, if that is at all possible
the goal is not yet reached, but I have made such wonderful
discoveries that I have been almost overwhelmed by them, and
it would be the cause of continual regret if they were lost.
When you will see them, my dear father, you too will recognise
it.
In the meantime I can only say this
/ have created a new
universe from nothing. All that I have sent you till now is but
a house of cards compared to a tower. I am as fully persuaded
that it will bring me honour, as if I had already completed the
:

discovery."

Wolfgang suggested that his son should publish his work,

and offered to insert it as an Appendix in the Tentatnen. He
advised him, if he had really succeeded, not to lose time in
letting the fact be known, for two reasons *
" First, because ideas pass easily from one to another, who
can anticipate its publication
and, secondly, there is some
truth in this, that many things have an epoch, in which they
are found at the same time in several places, just as the violets
appear on every side in spring. Also every scientific struggle
:

is

just a serious war, in

arrive.

which

cannot say when peace

Thus we ought to conquer when we are

advantage

is

always to the

first

will

able, since the

comer."

But the publication of the Tentamen was delayed for some

years.
In the meantime the MSS. was placed in his father's
hands, and he called some parts of it in question. His doubts
were partly removed, and the work was inserted in the first
volume, an advance copy of which reached Gauss at Gottingen
The younger Bolyai attached immense
in February, 1832.
importance to the approval of Gauss, at that time the greatest
authority in the world of mathematics. The high praise which
Gauss gave to his work we have already mentioned.

14.

We now

give a short description of the Appendix.

opens with a definition of parallels. // the ray AM is

not cut by the ray BN, situated in the same plane, but is cut by
every ray BP comprised in the angle ABN, this will be denoted
by BNlllAM.
(i)

It

*Stackel, "Die Entdeckung der nichteuklidischen Geometrie durch

Johann Bolyai," Math. u. Naturw. Berichte aus Ungarn, vol. xvii.
p. 14 (1901).

Also

loc. cit. vol.

i.

p. 86.

THE APPENDIX

13,14]

29

In a footnote he adds " pronounced BN asymptotic to AM."

Bolyai always used tlie word -parallel and the symbol
in the sense of equidistant, while he reserved the word
for the new parallels, in the
asymptotic and this symbol
sense in which we shall see Lobatschewsky used the term.
The properties of the new parallels are then established.
II

III

(ii) The properties of the circle and sphere of infinite radius

are obtained. It is shown that the geometry on the sphere of
infinite radius is identical with ordinary plane geometr}'.
(iii)

Spherical

Geometry

is

proved to be independent of the

Parallel Postulate.
(iv) The formulae of the Non-Euclidean Plane Trigonometry
are obtained with the help of the sphere of infinite radius.

(v) Various geometrical problems are solved for the Non"

Euclidean Geometry
e.g. the construction of a " square
whose area shall be the same as that of a given circle,*
;

Bolyai laid particular stress upon the demonstration of the

theorems which can be established without any hypothesis as
to parallels. He speaks of such results as absolutely true, and
they form part of Absolute Geometry or the Absolute Science of
Space. As the title of the Appendix shows, one of his chief

was to build up this science.

In the Appendix he says little about the question of the
impossibility of proving the truth of the Euclidean Parallel
objects

* Of course the Non -Euclidean "square" is not a quadrilateral with

equal sides and all its angles right angles.
A rectangle is impossible \ .
in the Non-Euclidean plane.
The square of Bolyai is simply a regular \ y
quadrilateral. The angles are equal, but their size depends on the r"

sides.

NON-EUCLIDEAN GEOMETRY

30

[ch.ii.

He refers to the point more than once but he

an occasion
postpones fuller treatment till a later occasion
which, so far as the public are concerned, never came. The
Postulate.

last sentences of the

follows

Appendix (Halsted's

translation) are as

" It remains finally, (that the thing may be completed in

every respect), to demonstrate the impossibility (apart from
any supposition), of deciding a priori, whether 2, or some S
(and which one) exists.* This, however, is reserved for a more
suitable occasion."

Bolyai retired from the army in 1833 and lived till

So far as we know he published nothing further, either
in extension of the Appendix or on any other mathematical
subject. From several sources, chiefly notes found among his
papers, we learn that he occupied himself with some of the
problems of the Non-Euclidean Geometry. He carried his
work further in the direction of Solid Geometry. He investigated more fully the relation between the Non-Euclidean
Geometry and Spherical Trigonometry and he pondered the

15.

1860.

question of the possibility or impossibility of proving Euclid's

Hypothesis.
An unpublished version of part of the Appendix exists in
German,! in which he gives clearer expression to his views upon
the last of these topics than is to be found in the corresponding
section of the original. In this version, which dates from 1832,
the first part of 33 reads as follows
" Now I should briefly state the essential result of this
:

what it is in a position to effect

Whether 2 or S actually exists, remains here

theory, and

"

I.

(and, as

the author can prove, for ever) undecided.

" II. Now there is a Plane Trigonometry absolutely true
{i.e. free from every hypothesis), in which, however, (according
to I.), the constant i and its very existence remain wholly
undetermined. With the exception of this unknown everyBut Spherical Trigonometry was
thing is determined.
* Bolyai calls 2 the system of Geometry resting upon Euclid's
Hypothesis; and S the system founded upon his own definition of

parallels.

fCf. Stiickel, " Untersuchungen aus der absoluten Geometrie aus

Johann Bolyai's Nachlass," Math. u. Naturw. Berichte aus Ungam,
vol. xviii. p. 280, 1902.

Also

loc. cit.

vol.

ii.

p. 181.

BOLYAI'S LATER YEARS

14, 15]

'Si

developed absolutely and completely in 26

so that the
ordinary familiar Spherical Trigonometry is not in the least
dependent upon Axiom XI. and is unconditionally true.
;

" III. By means of these two trigonometries and several

subsidiary theorems (to be found in the text of 32) one is able
to solve all the problems of Solid Geometry and Mechanics,
which the so-called Analysis in its present development has in
its power (a statement which requires no further qualification),
and this can be done downright without the help of Axiom XI.
(on which until now everything rested as chief-foundationstone), and the whole theory of space can be treated in the
above-mentioned sense, from now on, with the analytical
methods (rightly praised within suitable limits) of the new
(science).

" Taking now into consideration the demonstration of the

impossibility of deciding between 2 and S (a proof which the
author likewise possesses), the nature of Axiom XI. is at
length fully determined
the intricate problem of parallels
completely solved
and the total eclipse completely dispelled,
which has so unfortunately reigned till the present (for minds
thirsting for the truth), an eclipse which has robbed so many of
their delight in science, and of their strength and time.
" Also, in the author, there lives the perfectly purified conviction (such as he expects too from every thoughtful reader)
that by the elucidation of this subject one of the most important
and brilliant contributions has been made to the real victory
of knowledge, to the education of the intelligence, and consequently to the uplifting of the fortunes of men."
;

His proof of the impossibility of proving the Euclidean

Hypothesis seems to have rested upon the conviction that
the Non-Euclidean Trigonometry would not lead to any contradiction.
The following sentences are to be found among
his papers
" We obtain by the analysis of a system of points on a
plane obviously quite the same formulae as on the sphere
and since continued analysis on the sphere cannot lead to any
contradiction (for Spherical Trigonometry is absolute), it is
therefore clear that in the same way no contradiction could
ever enter into any treatment of the system of points in a
:

plane." *
*Cf. Stiickel,

loc. cit. vol.

i.

p. 121.

NON-EUCLIDEAN GEOMETRY

3(1'

And

[cnn.

lower down in the same passage

" But there still remains the question, whether in some way
or other the considerations of space would not avail for the
establishment of 2."
Indeed, owing to a mistake in his analysis, he thought for
a time that he had actually obtained a proof of the Euclidean
Hypothesis on these lines. But he discovered his error later.
From the fact that at one time he was willing to admit that,
Avith the aid of Solid Geometry, evidence against the logical
consistency of the Non-Euclidean Geometry might be obtained,
we must not imagine that he had failed to grasp the significance
of his earlier work. On the contrary, his argument shows that
he had seen more deeply into the heart of the matter than
Lobatschewsky himself. The latter, as we shall see below,
relied simply upon the formulae for the plane. Even when it
has been established that the Non-Euclidean Plane Geometry
is a perfectly logical and consistent system, the question still
remains, whether, somewhere or other, contradictory results
might not appear in the theorems of Solid Geometry.
This question, raised for the first time by Bolyai, was
settled many years later by Klein,* following upon some
investigations of Cayley. We shall give, in the last chapter
of this book, an elementary and rigorous demonstration
of the logical possibility of the Non-Euclidean Geometry of
Bolyai-Lobatschewsky, and shall show how the same argument
can be applied to the Non-Euclidean Geometry associated with
the

name
16.

of

Riemann.

The

Work

of Lobatschewsky (1793-1856).

Nicolaus Lobatschewsky Professor of Mathematics in the

University of Kasan was a pupil of Bartels, the friend and
fellow-countryman of Gauss. As early as 1815 he was working
at the Theory of Parallels, and in notes of his lectures (18151817), carefully preserved by one of his students, and now in
the Biblioteca Lobatschewskiana of the Kasan PhysicalMathematical Society, no less than three " proofs " of the

From a work on Elementary

Geometry, completed in 1823, but never published, the MSS. of
which was discovered in 1898 in the archives of the University
of Kasan, we know that by that date he had maSe some

Parallel Postulate are to be found.

*Cf.

Ann.

"tJber die sogenannte Nicht-Euklidische Geometrie," Math.

vol. iv. (1871).

LOBATSCHEWSKY

15, 16. 17]

33

advance
for lie says regarding the Parallel Postulate, " a
rigorous proof of this truth has not hitherto been discovered
those which have been given can only be called explanations,
and do not deserve to be considered as mathematical proofs
in the full sense." *
;

Between 1823 and 1826 Lobatschewsky had entered upon

the path which finally led him to his great discovery. It is
known that in 1826 he read a paper to the Physical-Mathematical Society of Kasan, entitled, Exposition succiiicte des
jrrincipes de la geometrie, avec mie demonstration rigoureuse du
theorl'me des paralleles. The MSS. of this work does not survive,
and the last clause in the title is ominous, for it suggests that
he had not yet reached his goal. But in 1829-30 he published
a memoir in Russian, On (lie Principles of Geometry, '\ and in
a footnote to the first page he explains that the work is an
extract from the Exposition succincte.
This memoir and many other works of Lobatschewsky have
come down to us, for, unlike Bolyai, he was- a prolific writer.
He published book after book, hoping to gain for the NonEuclidean Geometry the recognition it deserved a recognition

which

in his lifetime it

wholly failed to receive.

But

his first

published work contains all that is essential to the treatment

of the subject
and fully establishes the truth and value of
his discovery.
Thus, if the year 1826 cannot, with absolute
certainty, be taken as the date at which Lobatschewsky had
solved the problem, there is not the least doubt that his discovery of the Non-Euclidean Geometry was an accomplished
fact in the year 1829.
;

This memoir consists of nearly seventy pages. The

with the ordinary geometrical
notions of surface, line, point, distance, etc. In 8 he introduces his theory of parallels.
This section reads as follows J

17.

earlier sections, 1 to 7, deal

* 1

am

indebted to Dr. D. M. Y. Sommerville for a rendering of the

I. by Vasiliev to the Russian translation of Bonola's La
geometr'ta non-euclidea.
From this Appendix the sentence in the text

Appendix

is

taken.

We

t When Lobatschewsky's works appeared in Russian.

give the
titles in English.
This work is available in German in Eugel's
translation.
See Engel u. Staekel's Urkimden ztir Oeschichte der vichteuk/idischen Geometrie,

jCf. Engel,
N.-E.a.

I.

loc. cit. p.

(Leipzig, 1898).
10.

NON-EUCLIDEAN GEOMETRY

34
"

[ch.ii.

We

have seen that the sum of the angles of a rectilinear

tt.
There still remains the
assumption that it may be equal to tt or less than tt. Each of
these two can be adopted without any contradiction appearing
in the deductions made from it
and thus arise two geometries
the one, the customary, it is that until now owing to its simplicity, agrees fully with all practical measurements
the
other, the imaginary, more general and therefore more diffitriangle cannot be greater than

cult in its calculations, involves the possibility of a relation

between lines and angles.

" If we assume that the sum of the angles in a single rectilinear

is equal to tt, then it will have the same value in all.
On the other hand, if we admit that it is less than tt in a single
triangle, it is easy to show that as the sides increase, the sum of
the angles diminishes.
triangle

"In all cases, therefore, two lines can never intersect, when
they make with a third, angles whose sum is equal to tt. It is
also possible that they do not intersect in the case when this
sum is less than tt, if, in addition, we assume that the sum
of the angles of a triangle is smaller than tt.
'
In relation to a line, all the lines of a plane can therefore
be divided into intersecting and not-intersecting lines.
The
latter will be called parallel, if in the pencil of lines proceeding
from a point they form the limit between the two classes
or, in other words, the boundary between the one and the
;

other.

"

We

imagine the perpendicular a dropped from a point to

and a parallel drawn from the same point to the
same line. We denote the angle between a and the parallel
a given

by

line,

F(a).

It is

easy to show that the angle F(a)

is

equal to

TT

z
the sum of the angles of a triangle is equal
to TT
but, on the other hypothesis, the angle F (a) alters with
o, so that as a increases, it diminishes to zero, and it remains
for

every

line,

when

always

less

than

" To extend the meaning of F(a) to

we shall take

all lines a,

on the

latter

hypothesis,

f{0)

In this

way we can

= ^,

F(-a) = 7r-F(a).

associate with every acute angle A a

LOBATSCHEWSKY'S GEOMETRY

17]

and with every obtuse angle

positive line a,

that

a, sucli

Further
perties

A=
.

parallels,

in

^,

35

A, a negative line

F(a).

both cases, possess the following pro-

" If two lines are parallel, and two planes passing through
intersect, their intersection is a line parallel to both.
" Two lines parallel to a third are parallel to each other.
" When three planes intersect each other in parallel lines, the
sum of the inner plane angles is equal to tt."
In 9 the circle and sphere of infinite radius are introduced
the Limiting-Curve a.nA Limiting-Surface* of the Non-Euclidean

them

Geometry.
In 11 to 15 he deals with the measurement of triangles
and the solution of the problems of parallels.
At the end of 13 are to be found the fundamental equations
connecting the angles and sides of a plane triangle.
and those which follow it, are devoted to the determination, in the Non-Euclidean Geometry, of the lengths of curves,
the areas of surfaces, and the volumes of solids.
After the most important cases have been examined, he adds
a number of pages dealing with definite integrals, which have
only an analytical interest.
From the conclusion I make the following extract, as it is
related to the question already touched upon in the sections
dealing with Bolyai's work the logical consistency of the
(17)

16,

new geometry
' After
relations

we obtained the equations

(17),

between the sides and angles of a

finally given general expressions for the

which express the

triangle, we have
elements of

lines,

and volumes. After this, all that remains in Geometry

becomes Analysis, where the calculations must necessarily
agree with one another, and where there is at no place the
chance of anything new being revealed which is not contained
surfaces,

From them all the relations of the

geometrical magnitudes to each other must be obtained. If
anyone then asserts that somewhere in the argument a contradiction compels us to give up the fundamental assumption,
which we have adopted in this new geometry, this contradiction
can only be hidden in equations (17) themselves. But we

in these first equations.

''

See note on

p. 80.

NON-EUCLIDEAN GEOMETRY

36

[ch.h.

remark that these equations are transformed into the equations

(16) of Spherical Trigonometry by substituting ia, ib, and ic for
the sides a, b, and c. And in ordinary geometry and Spherical
Trigonometry there enter only the relations between lines. It
follows that the ordinary geometry, (Spherical) Trigonometry
and this new geometry must always be in agreement with
one another." *

18. " The writings of Lobatschewsky were brought under

the notice of Gauss as early as 1841, and we gather from his

how much impressed he was with them. Indeed it

almost appears as if he had thrown himself into the study of
Russian that he might be able to read the numerous papers
which he hears this " clear-sighted mathematician " had
Through Gauss the elder Bolyai
published in that tongue.
learnt in 1848 of the Russian's work, and in particular of the
Geometrische Untersuchungen zur Theorie der Parallellinien of
1840. The astonishing news and the volume, which Lobatschewsky had written as a summary of his work, were passed
on from the father to his son. How he received the intelligence
we learn from the following passage in some unpublished Notes
upon Nicolaus Lobatschewsky' s Geometrische Untersuchungen
" Even if in this remarkable work many other methods are
adopted, yet the spirit and the result so closely resemble the
Appendix to the Tentamen matheseos, which appeared in MarosVasarhely in 1832, that one cannot regard it without astonishment. If Gauss was, as he says, immensely surprised, first by
the Appendix and soon after by the remarkable agreement of
the Hungarian and Russian mathematician, not less so am L
" The nature of absolute truth can indeed only be the same
in Maros-Vdsdrhely as in Kamschatka and on the Moon, or, in
and what one reasonable
a word, anywhere in the world
being discovers, that can also quite possibly be discovered by
another."
letters

"i"

*
cf.

is referred to in Lobatsehewsky's other works

(ii)
GeoImaginary Geometry (Liebmann's translation, p. 8)

The same point

(i)

metrische Untersuchimr/en zur Theorie der Parallellinien (Halsted's

translation, p. 163) ; (iii) Pang4om^trie, 8 (quoted by Bonola, he. cit.
p. 93).

+ Cf. Kiirscliak ii. Stackel, ".Tohann Bolyai's Bemerkungen iiber

Nicolaus Lobatsehewsky's Geometrische Untersuchungen zur Theorie
der Parallellinien," Math. m. Naturw. Berichte aus Uiiga^vi, vol. xviii.
Also, Stackel, loc. cit. vol. i. p. 140.
p. 256 (1902)".

17, 18]

GAUSS, BOLYAI,

AND LOBATSCHEWSKY

3?

Then lie goes on to remark that in the world of science discoveries are not unlikely to be made about the same time
but
;

he cannot help wondering whether someone had not brought

his own work to Lobatschewsky's notice
after which the
latter might have attempted to reach the same goal by another
path. And he also makes the absurd suggestion that Lobatschewsky's work might really be due to Gauss himself that
Gauss, unable to endure that anyone should have anticipated
him in this matter, and yet powerless to prevent it, might have
himself written this work under Lobatschewsky's name. Bolyai
was undoubtedly a great genius, but he seems to have been the
possessor of an extraordinarily suspicious nature
The opinion of Gauss on the same work is given in a letter
to Schumacher of 1846 *
"... I have lately had occasion again to go through the
little book ... by Lobatschewsky.
It contains the outlines
of that geometry which must exist, and could quite consistently
exist, if the Euclidean Geometry is not true. A certain Schweikart called such a geometry the Astral
Lobatschewsky calls it
the Imaginary. You are aware that for fifty-four years (since
1792) f I have had the same conviction (with some extension
later, of which I shall not say more here). I have found nothing
really new to myself in Lobatschewsky's work
but the
development is made on other lines than I had followed, and by
Lobatschewsky, indeed, in a most masterful fashion and with
real geometrical spirit. I feel compelled to bring the book under
."
your notice. It will give you exquisite pleasure.
Lobatschewsky died in 1856 and Bolyai four years later
one of them, probably, a disappointed man
the other,
certainly, an embittered one.
Public recognition they had
not gained, and in all likelihood the number of mathematicians
acquainted with their work was extremely small. Had Gauss
only made public reference to their discoveries, instead of
confining himself to praise of their work, cordial and enthusiastic though it was, in conversation and correspondence, the
world would earlier have granted them the laurels they
;

deserved.
A few years after they
of Gauss
*

had passed away the correspondence'

and Schumacher was published, and the numerous

Gauss, Werke,

vol. viii. p. 238.

t Rather an early date, surely, for Gauss was born in 1777.

NON-EUCLIDEAN GEOMETRY

38

[ch. n.

references to the works of Lobatschewsky and Bolyai showed

the mathematicians of that day in what esteem Gauss had held
these two still unknown and obscure names. Soon afterwards,

thanks chiefly to Lobatschewsky 's works, and to the labours of

some well-known French, German, and Italian geometers, the
Non-Euclidean Geometry, which Bolyai and Lobatschewsky
had discovered and developed, began to receive full recognition.
To every student of the Foundations of Geometry their names
and their work are now equally familiar.

Work of Riemann (1826-1866).

development of Non-Euclidean Geometry is due
chiefly to Riemann, another Professor of Mathematics at
Gottingen. His views are to be found in his celebrated memoir
19.

The

The

later

Uber die HypotJiesen welche der Geometrie zu Grunde Uegen. This

paper was read by Riemann to the Philosophical Faculty at
Gottingen in 1854 as his Hahilitationsschrift, before an audience
not composed solely of mathematicians. For this reason it
does not contain much analysis, and the conceptions introduced
are mostly of an intuitive character. The paper itself was not
and the
published till 1866, after the death of the author
developments of the Non-Euclidean Geometry due to it are
mostly the work of later hands.
Riemann regarded the postulate that the straight line is
^adopted by all the other mathematicians who had
infinite
devoted themselves to the study of the Foundations of Geomeas a postulate which was as fit a subject for discussion as
try
the Parallel Postulate. What he held as beyond dispute was
the unboundedness of space. The difference between the in\jinite and unbounded he puts in the following words
"In the extension of space construction to the infinitely
great, we must distinguish between unboundedness and infinite
the latter
the former belongs to the extent relations
extent
to the measure relations. That space is an unbounded threefold jnanifoldness is an assumption which is developed by
every conception of the outer world according to which every
instant the region of real perception is completed, and the
possible positions of a sought object are constructed, and which
by these applications is for ever confirming itself. The unboundedness of space possesses in this way a greater empirical
certainty than any external experience, but its infinite extent
on the other hand, if we
by no means follows from this
;

RIEMANN

IS, 19, 20J

39

assume independence of bodies from

position,

and therefore

must

necessarily be
finite, provided this curvature has ever so small a positive
value." *
ascribe to space constant curvature,

it

for the hypothesis

the more general one that it is
With this assumption the Hypothesis of the
unbounded.
Obtuse Angle need not be rejected. Indeed the argument which
led Saccheri, Legendre, and the others to reject that hypothesis
depended upon the theorem of the external angle (I. 16). In
the proof of this theorem it is assumed that the straight line is

20. Riemann, therefore, substituted

that the straight line

is infinite,

infinite.

The Hypothesis of the Obtuse Angle being available, another

Non-Euclidean Geometry appeared. The importance of this
new Geometry was first brought to light, when the ideas of
the Non-Euclidean Geometry were considered in their bearing
upon Projective Geometry.
A convenient nomenclature was introduced by Klein. f He
called the three geometries Hyperbolic, Elliptic, or Parabolic,
according as the two infinitely distant points on a straight
The first case we meet
line are real, imaginary, or coincident.

the second
Geometry of Lobatschewsky and Bolyai
Geometry of Riemann the third in the Geometry of
These names are now generally adopted, and the
Euclid.

in the

in the

different

Non-Euclidean Geometries

will

be referred to below

by these terms.
It is evident that at this stage the development of the NonEuclidean Geometries passes beyond the confines of Elementary
Geometry. For that reason the Elliptic Geometry will not
receive the same treatment in this book as the simpler Hyperbolic Geometry. Also it should perhaps be pointed out here
the question will meet us again later that the Elliptic Geometry really contains two separate cases, and that probably only
one of these was in the mind of Riemann. The twofold nature

of this

Geometry was discovered by Klein.

taken from Clifford's translation of Riemann's

The surface of a sphere is unbcuvcled
A two-dimensional being moving on the surface of a
it is not infinite.
sphere could walk always on and on witliout being brought to a stop.
tCf. Klein, " Uber die sogenannte Nicht-Euklidische Geometric,"
Math. Ann. vol. iv. p. 577 (1871), and a paper in Math. Ann. vol. vi.
*"
Also Bonola, loc. cit. English translation, App. iv,
*

This quotation

memoir {Nature,

is

vol. viii. 1873).

[cH. in

CHAPTER

III.

THE HYPERBOLIC PLANE GEOMETRY.

21. In this chapter we proceed to the development of the
Plane Geometry of Bolyai and Lobatschewsky -the Hyperbolic
Geometry. We have already seen that we are led to it by the

consideration of the possible values for the sum of the angles

of a triangle, at any rate when the Postulate of Archimedes
This sum cannot be greater than two right angles,
is adopted.
assuming the infinity of the straight line. If it is equal to
two right angles, the Euclidean Geometry follows. If it is

than two right angles, then two parallels can be drawn

through any point to a straight line.
It is instructive to see how Lobatschewsky treats this
question in the Geometrische U ntersuchungen * one of his later
works, written when his ideas on the best presentation of this
fundamental point were finally determined.
" All straight lines in a plane which pass through the same
point," he says in 16, " with reference to a given straight
line, can be divided into two classes, those which cut the line,
and those which do not cut it. That line which forms the
boundary between these two classes is said to be parallel to the
less

given

line.

the point A (Fig. 13) draw the perpepdicular AD

to the line BC, and at A erect the perpendicular AE to the line
AD. In the right angle EAD either all the straight lines going
out from A will meet the line DC, as, for example, AF or some
of them, as the perpendicular AE, will not meet it.
" In the uncertainty whether the perpendicular AE is the
"

From

only line which does not meet DC,

* Geomefrische
1840).

let

us assume that

it is

Untersxichungen zur Theorie der Paralhllinien (Berlin,

English translation by Halsted (Austin, Texas, 1891).

PARALLELS.

21]

LOBATSCHEWSKY

possible tliat there are other lines, such as AG,

cut

DC however

41

which do not

far they are produced.

" In passing from the lines AF, which cut DC, to the lines AG,
which do not cut DC, we must come upon a line AH, parallel
to DC, that is to say, a line on one side of which the lines AG
do not meet the line DC, while, on the other side, all the lines
AF meet DC.

" The angle HAD, between the parallel AH and the perpendicular AD, is called the angle of parallelism, and we shall
denote it by ^{p), p standing for the distance AD."
Lobatschewsky then shows that if the angle of 2}arallelisvi
were a right angle for the point A and this straight line BC,
the sum of the angles in every triangle would have to be two

right angles.

Euclidean Geometry would follow, and the angle

of parallelism would be a right angle for

any point and any

straight line.

On

the angle of parallelism for the point A

angle, hf shows that
the sum of the angles in every triangle would have to be less
than two right angles, and the angle of parallelism for any
point and any straight line would be less than a right angle.

and

the other hand,

this straight line

if

BC were an acute

The assumption n(^) = ^ serves as the foundation

ordinary geometry, and the assumption

for the

IT

^{p)<a

leads to the

to which he gave the name Imaginary Geometry. In it two parallels can be drawn from any point to any
straight line.

new geometry,

In this argument Lobatschewsky

relies

upon the idea

of

NON-EUCLIDEAN GEOMETRY

42

[cH.

:ii.

continuity without stating the assumptions underlying that

term. The same remark applies to the argument of^Bolyai.
Indeed their argument does not prove the existence of the two
parallels.
The existence of the two parallels in this geometry
is an axiom, just as the existence of only one parallel is an
axiom in the Euclidean Geometry.

22. Hubert's

Axiom

of Parallels.

Hilbert makes the matter clearer by definitely inserting in his

treatment of the Hyperbolic Plane Geometry * the following
Axiom of Parallels
:

// b is any straight line and A any point outside it, there are
akvays two rays through A, a^, and ag, which do not form one and
the same straight line, and do not intersect the line b, while every
other ray in the region hounded by a^ and ag, which passes through
A, does intersect the line b.

Fig. 14.

Let BC be the

line h and AH, AK the rays a^ and a2Pasch's Axiom f it follows that no line in the regions
H'AH, K'AK cuts BC (Fig. 14).
Hence the rays a^ (AH) and flg (AK) form the boundary
between the rays through A which cut BC and the rays through
A which do not cut BC.
Through A draw the perpendicular AD to the line b (BC),
and also the perpendicular E'AE to the line AD.
Now E'AE cannot intersect BC, for if it cut BC on one side of
D, it must cut it at a corresponding point on the other.

From

Hilbert,

loc. cit. p. 160.

t Of. p. 3.

Also

Axiom

HILBERT

PARALLELS.

21, 22, 231

43

cannot be parallel to BC, because according to the

the two parallels are not to form one and the same

it

straight line.

Therefore the angles between

We

shall

now show

a^, a^,

and AD must be acute.

that they are equal.

one of them must be the greater.

the greater angle with AD, and at A make

If the angles are unequal,

Let

ttj

make

-DAP=_DAK.
Then AP must cut BC when produced.
t\

\^v^
\P

^^^^/
a.^

<r-

^a,

>

Fio. 15

Let

it cut it at Q.
the other side of D, from the line h cut off DR = DQ and
join AR.
Then the triangles DAQ and DAR are congruent, and AR
makes the same angle with AD as a^, so that AR and a^ must

On

coincide.

But ag does not cut BC therefore the angles which a^, a^

make with AD are not unequal.
Thus we have shown that the perpendicular AD bisects the
;

angle between the parallels ay and a^.

The angle which AD makes with either of these rays is called
the atigle of parallelism for the distance AD, and is denoted,

by ^{p), where AD =p.

and a^ are called the right-handed and
handed parallels from A to the line BC.

after Lobatschewsky,

The rays

a^

left-

23. In the above definition of parallels, the starting point

A of the ray is material. We shall now show that

straight line

maintains

its

property of parallelism at all -^

'

its

points.

NON-EUCLIDEAN GEOMETRY

44

[cH. ni.

In other words, if the ray AH is the right-handed {or lefthanded) parallel through A to the line BC, then it is the righthanded {or left-handed) parallel through any point upon the ray
AH, or HA produced, to the given line.

Let A' be any point upon the ray AH other than A.

A' draw A'D' perpendicular to BC.
In the region bounded by A'D' and A'H draw any ray A'P,
and take Q, any point upon A'P.
Join AQ.
Then AQ produced must cut DC.
It follows from Pasch's Axiom that A'Q must cut D'C.
But A'H does not cut D'C, and A'P is any ray in the region
D'A'H.
Therefore A'H is a parallel through A' to the line BC.

Case

I.

Through

Case II. Let A' be any point upon the ray AH produced
backwards through A.

Draw

A'D' perpendicular to BC.

SOME THEOREMS ON PARALLELS

23,24]

45

In the region bounded by A'D' and A'H draw through A'

A'P, and produce PA' through A' to Q.
Upon A'Q take any point R and join AR.
Then RA produced must intersect DC.
It follows that A'P must intersect D'C.
Therefore, as above, the ray A'H is a parallel through any
point A', on HA produced, to the line BC.
In both cases the parallels are in the same sense or direction as
the original ray (i.e. both right-handed or both left-handed).
We are thus entitled to speak of a line AB as a right-handed (or
left-handed) parallel to another line CD, without reference to
any particular point upon the line AB.

any ray

24. Another property of parallels with which we are

Geometry also holds for the Hyperbolic
Geometry.

familiar in Euclidean

If the line

AB

is parallel to the line

parallel to the line

CD, then

the line

CD

is

AB.

From A draw AC perpendicular to CD. and from C draw

perpendicular to AB.

CE

Fio. 18.

In the region DCE draw any ray CF, and from A draw AG
perpendicular to CF.
It is easy to show that the point G must lie in the region

ECD.
Further, since
angle,

AC>AG.

.^

ACG

is

an acute angle and

z.

AGC

is

a right

NON-EUCLIDEAN GEOMETRY

46

From AC cut
AH on the same
Make

off

AH =AG, and draw HK

side as

[chiti.

perpendicular to

CD.

HAL = ^ GAB.

Then the ray AL must cut CD, and

it follows that
cut AL.
Let HK cut AL at M.
From AB cut off AN =AM, and join GN.
Then the triangles HAM and GAN are congruent.

Thus

AGN =a right angle.

GN and GF coincide, and CF produced

HK must

/.

Therefore
AB.

intersects

in the region between CE and CD, and

does not cut AB.
Therefore CD is parallel to AB, in the same sense as A B is

But CF was any ray

CD

itself

parallel to

25.
proved

CD.*
third important property of parallels

must

also be

If the line (1) is parallel to the line (2) and to the line (3),
the three lines being in the same plane, then the line (2) is also

parallel to (3).

Case

I.

Let the

line (1) lie

between

(2)

and

(3).

(Cf. Fig. 19.)

Fio. 19.

Let A and B be two points upon

(2)

and

(3),

and

let

AB

cut

(1) in C.

Through A

AB and
*

let

any arbitrary

line

AD

be drawn between

(2).

The proof

in the text is

Ne.w Principles of Oeometry

(Engel's translation,

p.

161t)-

adapted from that of Lobatsehewsky in

a Complete Theory of Parallels, 96

loifh

SOME THEOREMS ON PARALLELS

24, 25, 26]

Then

must cut

it

(1),

and on being produced must

47

also cut

(3).

Since this holds for every line such as AD, (2)

is

parallel to

(3).

Case II. Let the

let (2) lie

between

line (1)

be outside both

and

(Fig. 20.)

(1)

(3).

(2)

and

(3),

and

Fio. 20.

not parallel to

If (2) is

random upon
is

(3),

(3),

through any point chosen at

(3) can be drawn which

a line different from

parallel to (2).

This,

by Case

26.

We

I., is

shall

also parallel to (1),

now

which

is

absurd.*

consider the properties of the figure

formed by two parallel rays through two given

points and the segment of which these two points are the ends.
[of.

Fig. 21]

Fio. 21.

speak of two parallel lines as meeting at

In the Hyperbolic Geometry each straight line will
have two points at infinity, one for each direction of parallelism.
With this notation the parallels through A, B may be said to
meet at 12, the common point at infinity on these lines.
It is convenient to

infinity.

*The

proof in the text

loc. cit. p. 72.

is

due to Gauss, and

is

taken from Boiiola,

NON-EUCLIDEAN GEOMETRY

48

12,

[cH.m.

Also, a straight line will be said to pass through this point

when it is parallel to these two lines in the same sense.
1.

If a straight line passes through one of the angular points

12, and through a point inside this figure, it must cut

A, B, or

the opposite side.

(Fig. 21.)

Let P be the point within the figure. Then AP must cut BS2,
by the Axiom of Parallels. Let it cut 8(2 at Q. The line Pl2
must cut one of the sides AB or BQ, of the triangle ABQ, by
Pasch's Axiom. It cannot cut BQ, since it is parallel to Bl2.
Therefore it must cut AB.
2. A straight line in the plane ABI2, not passing through an
angular point, which cuts one of the sides, also cuts one, and only
one, of the remaining sides of this figure.
Let the straight line pass through a point C on AB. Let Ci2
be drawn through C parallel to AI2 and BI2. If the given line
lies in the region bounded by AC and Ci2, it must cut Ai2
and
if it lies in the region bounded by BC and Ci2 it must cut Bl2.
Again, if the line passes through a point D on AI2, and B, D
are joined, it is easy to show that it must cut either AB or Bi2.
We shall now prove some further properties of this figure.
;

3.

The

exterior angle at

or

is greater

than the interior and

opposite angle.

.c

Fio. 22.

Consider the angle at A, and produce the line BA to C.

=/. ABO.
AM cannot intersect Bi2, since the
exterior angle of a triangle is greater than the interior and
opposite angle. Also it cannot coincide with AI2, because then
the perpendicular to Al2 from the middle point of AB would
The angle of parallelism for
also be perpendicular to Bi2.
this common perpendicular would be a right angle, and this
is contrary to Hilbert's Axiom of Parallels.

Make Z.CAM

SOME THEOREMS ON PARALLELS

26]

Therefore ^CAi2>i.CAM, which is equal to ^lABfl.

Thus the exterior angle at A is greater than the interior
angle at B.
A similar proof applies to the angle at B.
We take now two figures of this nature each consisting
of a segment and two parallels through the ends of the segment.
'

4.

AB= the

If the segment

A =the angle

segment A'B', and the angle at

at A', then the angles at

B and

B' are equal.

D'

n'

B'

Pio. 23.

If

^ ABft

is

not equal to l

A'B'12',

one of them must be the

greater.

Let

^ABi2>^A'B'0'.

Make

^ ABC

=.1 A'B'fi'.

Then BC must cut Afi.

Let it cut it at D and on A'fl' take A'D' = AD, and join B', D'.
Then the triangles ABD and A'B'D' are congruent, so that
;

A'B'D' =:! ABD -z. A'B'i]', which is absurd.

It follows that z. AB12 is not greater than z. A'B'12', and that
the angles are equal.

z.

5.

If the segment

A and B are
four angles

AB = the

segment A'B', and the angles at

equal, as also the angles at A'

and

B',

then the

at A, B, A' a)td B' are equal to each other.

B'
Fig. 24.

If the angle at

must be the
N.-E.O.

is

greater.

not equal to the angle at A', one of them

Let it be the angle at A,

NON-EUCLIDEAN GEOMETRY

50

[CH.

Ill

At A and B draw the rays whicli make with AB an angle

equal to the angle at A'.
These rays must intersect
let them meet at C.
From A'fi' cut ofi A'C'-AC, and join B'C
The triangles ABC and A' B'C' are congruent, so that
;

^ A'B'C =z-

which

ABC =^ A'B'i]',

absurd.
Thus the angles at A and A' must be equal and it follows
that the angles at A, B, A' and B' are equal to each other.
is

6. // the angles at A and A' are equal,

B' are also equal, then the segment

and

AB= the

If

Let

the angles at

B and

segment A'B'.

AB
it

is not equal to A'B', one of them must be the greater.

be AB.
A'

From AB
Then, by

But by

cut
(4),

(3),

Therefore

off

AC = A'B', and draw

^ ACi2

L ACfi

>

CI2 parallel to AI2.

^l A'B'il' =l ABI2.
L ABfi.

AB cannot be

greater than A'B', and the two

segments are equal.

27.

From

The Angle of

26

(4),

Parallelism.
at once deduce that

we can

the angles of

parallelism coiresponding to equal distances are equal.

yy
FiQ.

Combining

this result
If

i.>j>P2,

with

then

26

2(3.

(3),

we can

assert that

n(^g)>II(^j).

SACCHERI'S QUADRILATERAL

26,27,28]

51

We

shall see later ( 41) that to any given segment we can

find the angle of parallelism, and that to any given acute angle
(

we can find the correspojiding distance.

Thus, we can say that

45)

lh=lh^

then

Il{2h)

If

Pi>i?2.

then

Jl{p^)

<U{p^).

If

Pi<P2,

then

n(^^)

> n(;?2)-

If

n(0)

= '^,

n(oo)

= o.

Also

It is

convenient to use the notation

a = n(a),
Again,
[cf.

if

41],

the segment a

and thus

7,

-a.

distance of parallelism
this

^{p^).

/3
is

And

to

[cf. 45].

complementary segment by

Thus we have

= n(&),

given,

tt

etc.

we can

find the angle oc

-oc there corresponds a

It is

convenient to denote

a'.

tt
n(a) = --n(a).
/

/\

"^r

Further, in the words of Lobatschewsky,* " we are wholly

at liberty to choose what angle we will denote by the symbol
n(^), when the line p is expressed by a negative number, so

we

shall

assume

Uln) + U(

-i->)='7r

^**

"

'''

^ *"*

ci(ml<.. >(

^ (,..u ^ ., ., ,^..
28. Saccheri's Quadrilateral.
quadrilateral in which the angles at A and B are right 7-'*'
angles, and the sides AC, BD equal, we shall call Saccheri's
Quadrilateral.
We have seen that Saccheri

The

made frequent use of it in his discussion of

the Theory of Parallels.
In Saccheri's Quadrilateral, when the right
angles are adjacent to the base, the vertical
angles are equal acute angles, and the line
which bisects the base at right angles also
bisects the opposite side at right angles.
Let AC and BD be the equal

sides,

Fio. 27.

and the angles at A and

right angles.

* Oeometrische Untersuchunyen zur TheoTie der

ParcUlellinien, 23.

NON-EUCLIDEAN GEOMETRY

52

Let E, F be the middle points of

Join EF, CE, and DE.

AB and CD

[CH.

III.

respectively.

Then the triangles ACE and ED B are congruent, and the congruence of CFE and EFD follows.
Thus the angles at C and D are
equal, and EF is perpendicular both
to AB and CD.
Further, the angles at C and D
are acute.

To prove
Cfl

and

this, at

C and D draw

D12 parallel to AB.

Then, by 26 (4),
Produce CD to E.

By 26

(3),

i.

ACfi

=l

BDI2.

^ED12>^DCU.
z.ACD=:_BDC,

Therefore, since

it

follows that

^EDB>^CDB.
Thus L ACD and l BDC are both acute

angles.

29. If in the quadrilateral A BDC, the angles at A and B are

and the side AC is greater than
BD, the angle at C is less than the angle C
E
at D.
=~1D
right angles,

we are given AC > BD, we can cut

from AC the segment AE=BD. When
this has been done, join DE.
It follows from 28 that i. AED =l BDE.
But L. AED > ^ ACD and l BDC > ^ BDE.
Therefore l BDC > l ACD.
Since

off

The converse of
namely,

as z.ACD

30.

these

theorems

that, if the angles at

= :lBDC.

// A BDC

and C a^e

so is

is

is

easily

A and B are

proved indirectly,

right atigles, according

AC=BD.

a quadrilateral in which the angles at A,

D must be acute.

B,

right angles, then the angle at

Produce BA through A to

B',

making AB' = AB.

Draw

(Fig. 30.)

B'D' perpendicular to B'A and equal to BD.

Join CD', D'A, and DA.
From the congruent triangles D'B'A and DBA, we have

D'A=DA and

z.

D'AB'

=^ DAB.

SUM OF ANGLES OF A TRIANGLE

28, 29, 30, 31]

Thus ^ D'AC

=zL

53

DAC, and the triangles D'AC and DAC are

congruent.
Therefore l D'CA

is

a right angle, and DC, CD' form one

straight line.

d'

Applying the result of 28 to the quadrilateral D'B'BD,

follows that the angles at D' and D are equal and acute.

31. The

sum

of the angles of every triangle

it

is less

than two right angles.

Case I. Let the triangle ABC be any right-angled triangle
=90.
with
At A make l BAD =^ ABC.

Fio. 31.

From

O, the middle point of AB, draw the perpendiculars OP

and OQ to CB and AD respectively.
Then the triangles POB and ACQ are congruent, and it
follows that OP and OQ are in one and the same straight line.

NON-EUCLIDEAN GEOMETRY

54

Thus the quadrilateral ACPQ, has the angles

[CH. HI.

at C, P,

and

right angles.

Therefore, by 30, the angle at A, namely _CAD, must be

acute.
It follows that the sum of the angles of any right-angled
triangle must be less than two right angles.

Case II. Consider now any triangle, not right-angled.

Every triangle can be divided into two right-angled triangles
by drawing the perpendicular from at
least

one angular point to the opposite

side (Fig. 32).

Let AD be the perpendicular referred

to in the
angles a.',

triangle
.", /3,

Then A + B + C =

ABC, and

a'

and

18

-f-

Therefore A

-f

the

figure.

+ /3)

+ y).

(a'

+ <
a" y <

But

let

be as in the

1
1

B+C

-1-

(."

right angle

right angle.

<

!'

2 right angles.

It should be noticed that no use has been

Postulate of Archimedes in proving this result.

made

of the

The diflference between two right angles and the sum of the
angles of a triangle will be called the Defect of the Triangle.

Corollary. There cannot he two triangles tvith their angles

equal each to each, which are not congruent.
It is easy to show that if two such triangles did exist, we
could obtain a quadrilateral with the sum of its angles equal to
four right angles. We have simply to cut off from one of the
triangles a part congruent with the other. But the sum of
the angles of a quadrilateral cannot be four right angles,
if the sum of the angles of every triangle is less than two

right angles.

32. Not-intersecting Lines.

from the Theorem of the External Angle (I. 16)

two straight lines have a common perpendicular, they

It follows

that

if

cannot intersect each other.

since this

26

(3)].

And they cannot

would contradict Hilbert's Axiom of

be

parallel,

Parallels

[cf.

NOT-INTERSECTING LINES

31, 32]

The converse

is

55

also true, namely, that

If two straight lines neither intersect nor are parallel, they

must have a common perpendicular.*

Fio. 33.

Let a and b be the two given

nor are parallel.

lines,

which neither intersect

From any two points A and P on the line a, draw AB and

PB' perpendicular to the line 6.
AB = PB', the existence of a common perpendicular
If
Therefore we need only discuss the case
follows from 28.
when AB is not equal to PB'.
Let PB' be the greater.
Cut off A'B' from PB' so that A'B' is equal to AB.
At A' on the line A'B', and on the same side of the line as
AB, draw the ray a' making with A'B' the same angle as a, or
PA produced, makes with AB.
We shall now prove that a' must cut the line a.
Denote the ray PA by a^, and draw from B the ray h parallel
to a^.
Since a, b are not-intersecting lines, the ray h must lie in the
region between B A and B' B produced.
Through B' draw the ray h', on the same side of B'A' as
h is of BA, and making the same angle with the ray B'B as

with B'B produced.

26 (3), it follows that the parallel from
in the region between h' and B' B.

A. does

From
lies

This proof

is

due to Hilbert

cf. loc. cit. p.

B' to

162.

h and a^

NON-EUCLIDEAN GEOMETRY

56

[CH.

III.

Therefore Ji' must cut a^.

it cut this line at T.
Since a' is parallel to B'T, it follows that the ray a' must
cut PT (Pasch's Axiom).
Let these rays a^, a' intersect at Q.
From Q draw QR perpendicular to the line h, and from the
line h cut ofE BR' equal to B'R and on the opposite side of B

Let

from

B'.

In the same way, from the line a cut oE AQ' equal to

and on the opposite side of A from P.
In this way we obtain a quadrilateral ABR'Q' congruent
A'B'RQ.
Thus QRR'Q' is a Saccheri's Quadrilateral, and the
joining the middle points of QQ', RR' is perpendicular

and

A'Q,

with
line

to a

h.

33.

Two

'parallel lines

approach each other continually, and

becomes less than any assigned

their distance apart eventually

quantity.

Let a and b be two parallel

lines.

Upon a

take any two points P and Q,

of parallelism for the lines.

PQ

being the direction

>b

the perpendiculars PM and QN to b:

and draw the perpendicular at H to the

From P and Q draw

Bisect

MN

at H,

line 6.
let it do so at K.
This must intersect the segment PQ
At K draw the ray a' parallel to b in the other direction.
This ray must intersect PM, since it enters the triangle
;

PKM

at the vertex K.
it cut PM at P'.

Let

PARALLEL LINES ARE ASYMPTOTIC

32, 33]

Since the triangles

KHM and KHN

are

congruent,

57

and

L HKP'=^ HKQ, it easily follows that P'M is equal to QN.

But P' lies on the segment PM.
Therefore PM is greater than QN, and we have shown that
as we pass along the line a, in the direction of parallelism, the
distance from h continually diminishes.
We have now to prove the second part of the theorem.
Let a and 6 be two parallel lines as before, and P any point
on the line a.

M'
Fio. 35.

Draw PM perpendicular to 6, and let be any assigned

length as small as we please.
If PM is not smaller than e, cut oE MR = 6.
Through R draw the ray a^ (RT) parallel to a and 6 in the
same

sense.

Also draw through R the ray RS perpendicular to MR.

RS must cut the ray a, since l PRT is an obtuse angle.
Let it cut a at Q and draw QN perpendicular to h.
Now the lines RQS and the line h have a common perpendicular.

Therefore they are not-intersecting lines.

It follows that Z.NQR is greater than
parallelism for the distance QN.

the

angle

of

At Q make l. NQR' = l NQR.

Then l NQR' > l NQT', T' being any point upon PQ produced.

NON-EUCLIDEAN GEOMETRY

58

From

the line b cut off NM'

from M, and from QR' cut

ofi

= NIVI, on
= QR.

[cH.m.

the other side of N

QR'

Join R'M'.
Then R'M'

is perpendicular to b, and is cut by the ray PQ

between R' and M'.
Let the point of section be H.
Then M'H < M'R', and M'R' = MR.
Therefore we have found a point on the line a whose distance
from b is less than the given length e.

The

any two not-intersecting

lines is their common perpendicular, and as we proceed along
either of the lines from the point at which it is cut by the common
perpendicular the distance from one to the other continuxilly

34.

shortest distance between

increases.

Let the common perpendicular to two not-intersecting

a and b meet them at A and B.
Let P and Q be two other points on
one of the lines on the same side of Aj
and such that AP < AQ.

Draw PM and QN
other

lines

perpendicular to the

line.

A BMP, the
M
N
B
a right angle and the angle
pi- so.
APM is acute (cf. 30).
Therefore PM > AB (cf. 29).
Also in the quadrilateral PQMN, the angle MPQ is obtuse

Then

in the quadrilateral

angle A

and PQN

is

is

acute.

Therefore QN > PM.

Thus, as we pass along the ray APQ... the distance from the
line b continually increases from its value at A.
It can be shown that two parallel lines continually diverge
towards the side opposite to the direction of parallelism, and
that two intersecting lines continually diverge from the point
of intersection. Also, the distance apart, both in the case of
intersecting lines, of parallel lines, in the direction opposite to
that of parallelism, and of not-intersecting lines will become
eventually greater than any assigned length.
The theorems of 33-4 were all proved by Lobatschewsky
cf. New Principles of Geometry with a Complete Theory of
Parallels (Engel's translation), 108

et seq.

33,34,35]

THE RIGHT-ANGLED TRIANGLE

59

35. The correspondence between a Right- Angled

Triangle and a Quadrilateral with Three Right
Angles and One Acute Angle.

The sides of a right-angled triangle

right angle, are denoted as usual by a,
and B by A and

ABC, in which C is the

the angles A
h, and c
;

and the distances corresponding to the

angles of parallelism X and ju are denoted by I and m.
Between these quantities a, h, c, I, m, A, and fx certain relations
ju

hold.
Similarly the elements of a quadrilateral, in which three
angles are right angles, the remaining angle being necessarily
acute, are connected by certain relations.

We

proceed to find the equations connecting these quanti-

and to establish a very important correspondence between

the two figures.
ties,

Fio. 38.

I.

The Right- Angled Triangle.

Let ABC be any right-angled triangle. Produce the hypothenuse through A a distance Z, and at the other end of the
segment I draw the parallel to the line CA. Also draw through
B the parallel to both these lines.
It follows from Fig. 38 that
fM

and

in the

+ U{c + l)=^U{a) = (A,

(1)

same way we have

\ + U{c + m) = n{b) = l3

(1')

NON-EUCLIDEAN GEOMETRY

60

[CH.

III.

Now draw

through A the parallel to the line BC (Fig. 39).

Also draw the line perpendicular to c, which is parallel to
BC in the same sense. This line will cut the hypothenuse, or
the hypothenuse produced, according as
is less than or

greater than
If

m<

If

m>

c.

c,

we have

c,

then we would have

\-f/3

= n(c-w)

(2)

7r-X-/3 = n(77i-c).

Fio. 39.

FiCf, 40.

With the usual notation

(cf.

27) this

reduces to

\ + /3 = Il{c-m).
In the

same way we have

IUL+(X.

CB through
CB produced which is also

Finally, produce

to

= II{C-1)
B,

(2')

and draw the perpendicular

parallel to

AB

(Fig. 40).

Also

produce AC through C, and draw the perpendicular to AC

which is parallel to AB.
From Fig. 40, if we suppose a line drawn through C parallel
to AB,

it is

clear that

U{l-b) + U{m + a)
and

similarly

2'

U(m-a) + U{l + h) = ^

(3)

(3')

THREE-RIGHT-ANGLED QUADRILATERAL

35]

II.

61

The Quadrilateral with Three Right Angles and an Acute

Angle.

Let PQRS be a quadrilateral in which the angles P, Q, R

are right angles.
We denote the sides, for reasons that will
presently appear, by Zj, a^, m^ and c^
the acute angle ,by
,

y3j

and

l^,

Produce

c^

c^

dicular at the

contain this angle

/3j.

through R a distance w^, and draw the perpen-

end of that segment.

Since II (wij)

+ 11 (w^') = -^

if the parallel through R to PQ is supposed drawn,

that this perpendicular is parallel to PQ (Fig. 41).

It follows that

and correspondingly

A, -fn(Ci-l-Wj)

yj

+ n(/i

-f-rt\)

= /3i,
= /8i

it

follows

(I.)
(I'.)

Fig. 42.

From RS

m^ then
=
n(c,-m,),
Xj+A
Yi + /3i = n(/j a^)

cut off the segment

Fig. 42 that

and correspondingly

it is

obvious from
(II.)

(IF.)

NON-EUCLIDEAN GEOMETRY

62

[CH.

III.

Finally, from QP cut off the segment m^, and from PS produced the segment &j, and raise the perpendiculars at the ends

of these lines (Fig. 43),

Q.

m,
Fig. 43.

It follows that

U{l^

+ h,) + U{m.,

.(III.)

(iir.)

and correspondingly
n(ci-K&i)
III.

We

are

tween the two

able to establish the correspondence be-

figures.

right-angled triangle is fully determined when we know

a quadrilateral of this nature, when we know Cj

and ju
and ?%'.
c

now

Let

Ci

= c and

^-yu,

m^ = m.

so that

Then

n(mi')

it

follows from

(!')

and

(2)

that

X + ^ = U(c-m),
-A+^=n(fi-t-m),
and therefore

2X =

11 (c - m) - U(c + m),
=
2^ U{c-m) + U{c + m).

35,36]

TRIANGLE AND QUADRILATERAL

But from

(I.)

and
Xi

(II.)

we have

+ |8i = n(q-7ni) = n(c-m),

/3^ = n (Cj + Wj = n (c + w).
and /3i = /3.
Aj = A

- Xi +

Therefore

From

(III')

and

we now obtain

m-

Thus

//

(III.)>

and

ttj

ftj

Therefore

63

= w - a,
= a.

we have obtained the important

a, b, c, (A, fx) are the five elements

result

of a rigid-angled

triangle,

then there exists a quadrilateral with three right angles and one
acute angle, in which the sides are c, m', a, and 1, taken in order,
and the acute angle
lies between c and 1.*

The converse of

this

theorem also holds.

36. The Closed Series of Associated Right- Angled

Triangles.

We have seen that to the right-angled triangle a, h, c, (A, fx)

there corresponds a quadrilateral with three right angles and
* This result was given by Lobatschewsky in his earliest work, On the
Principles of Oeome/ry (of. 11, 16, Engel's translation, pp. 15 and 25),
but liis demonstration requires the theorems of the Non Euclidean Solid
Geometry. TI)e proof in the text is due to Liebmann (Math. Ann,
vol. Ixi. p. 185 (1905), and NichteuHidische Oeometrie, 2nd ed. 10),
who first established the correspondence between the right-angled triangle and the quadrilateral with three right angles and an acute angle
by the aid of Plane Geometry alone.
This is an important development, as the Parallel Constructions

depend upon this correspondence, and theNon-EuclideanPlane Geometry

and Trigonometry is now self-contained.
Further, as we shall see below ( 45), the existence of a segment
corresponding to any given angle of parallelism can be established
without the use of the Principle of Continuity, on which Lobatschewsky 's
demonstration depends. Therefore, though the existence of j), when
n(jo) is given, is assumed in the above demonstration, the cori'espondence between the triangle and quadrilateral is independent of that
principle.

NON-EUCLIDEAN GEOMETRY

64

[ch.iii.

an acute angle /3, the two sides enclosing the acute angle
being c and I, and the other two a and m'.
If we interchange c and I, and m' and a, we obtain the same
quadrilateral.
a, h, c, (X, fi),

elements are
aj

It follows that, given the right-angled triangle

there exists another right-angled triangle whose

a^, h^, c^, (A^, yUj),

where

= m',

A|

b^

= b,

c-^

l,

= y,

ijl^

= -^-cl.

Thus, starting with the right-angled triangle

a,

we obtain

(X,

c,

(1)

/i),

a second right-angled triangle whose elements are

<
If

b,

we now take the

sides

in the reverse order,

i.e.

b,

we obtain another

h (y.

m',

and opposite angles of

write

I,

(^)

f-"-)

this triangle

as the triangle

it

(|-a-, yj,

right-angled triangle with the elements

m', a,

c',

(X,

|-iS)

(3)

Writing this as
m',

c',

a',

(I -A

^)'

we obtain another with the elements

V,

From

this

c',

we obtain
/',

Again, from this

b',

(|-oc, m)

(4)

in its turn
a,

m, (y,

|-^)

(5)

X),

(6)

we have
b,

a,

c,

(fi,

the last being the original triangle.

36]

THE SERIES OF ASSOCIATED TRIANGLES

65

The relation between the elements of these triangles can be

put in the form of the following rule
:

Let a, b, c, (A = IT (1), y = 11 (m)) he the sides, hypothenuse, and

the angles opposite the sides of a right-angled triangle.
Write the
letters a', 1, c, m, b'
cyclic order on the sides of a pentagon. The

six triangles which

form the closed series of associated triangles

are obtained, if ive write the letters a,^', \r, Cr, mr, br' in the same or
reverse order on the sides, starting with any one side, and take the
elements with the suffices equal to those on the same sides without
the suffices.

=r
=a

i.e.

ar

= a\

i.e.

br

mr = b\

i.e.

fir

= -2 -13;

i.e.

Xy

=y

br

= ra;
Ir = c,

IT

Cr

giving the triangle (5) above.

These results have an important bearing on certain problems

of construction. For example, the problem of constructing a
right-angled triangle when the hypothenuse c and a side a are
given, with the usual construction involves the assumption as
N.-E.a.

NON-EUCLIDEAN GEOMETRY

66

[ch.

m.

an assumption
tlie intersection of a circle and a straight line
which depends upon thel Principle of Continuity. But we
know that with the triangle a, h, c, (A, ju) there is associated a
to

triangle l',a,m, [y, -n- /S).

side a

and the adjacent angles

In this triangle we are given a

y,

and

it

can be constructed

z
associated triangle gives us

without that assumption. The

the second side h of the required triangle. This argument
depends upon the theorem proved in 41-3, that we can always
find
(p) when p is given, and that proved in 45, that
given Il{p), we can always find p.

37. Proper and Improper Points.

In the Euclidean Plane two lines either intersect or are
parallel.
If we speak of two parallels as intersecting at " a
point at infinity " and assign to every straight line " a point
at infinity," so that the plane is completed by the introduction
of these fictitious or improper points, we can assert that any
two given straight lines in the plane intersect each other.
On this understanding we have two kinds of pencils of
the ordinary pencil
straight lines in the Euclidean Plane
whose vertex is a proper point, and the set of parallels to
any given straight line, a pencil of lines whose vertex is an
improper point.
Also, in this Non-Euclidean Geometry, there are advantages
to be gained by introducing fictitious points in the plane. If
two coplanar straight lines are given they belong to one of
they
three classes. They may intersect in the ordinary sense
may be parallel or they may be not-intersecting lines with a
common perpendicular. Corresponding to the second and
third classes we introduce two kinds of fictitious or improper
points.
Two parallel lines are said to intersecif at a point at
And every straight line will have two points at
infinity.
infinity, one corresponding to each direction of parallelism.
All the lines parallel to a given line in the same sense will
thus have a common point a point at infinity on the line.
Two not-intersecting lines have a common perpendicular.
The lines are said to intersect in an ideal point corresponding
to this perpendicular. And all the lines perpendicular to one
:

and the same straight line are said to intersect in the ideal
point corresponding to this line.

36. 37, 38]

PROPER AND IMPROPER POINTS

67

We shall denote an ordinary point a proper point ^by the

usual capital letter, e.g. A. An improper point a point at
and a point
infinity -by the Greek capital letter, e.g. 12
belonging to the other class of improper points an ideal point
by a Greek capital letter with a suffix, to denote the line to
which the ideal point corresponds, e.g. Fg.
Thus any two lines in the hyperbolic plane determine a

pencil.

is

(i) If the lines intersect in an ordinary point A, the pencil

the set of lines through the point A in the plane.

(ii) If the lines are parallel and intersect in the improper

point i2, the pencil is the set of lines in the plane parallel to
the given lines in the same sense,
(iii) If the two lines are perpendicular to the line c, and
thus intersect in the ideal point which we shall denote by F,.,
the pencil is the set of lines all perpendicular to the line c.

38.

We now

enumerate

all

the cases in which two points

and the correspond-

in the Hyperbolic Plane fix a straight line

ing constructions

Two

ordinary points A and B. The construction of the

line joining any two such points is included in the assumptions
of our geometry,
(1)

An

ordinary point [A] and a point at infinity [12], The

by drawing the parallel through A to
the line which contains (2, in the direction corresponding to i2.
This construction is given below in 41-3,
(2)

line Ai2 is constructed

An ordinary point [A] and an ideal point [r,.]. This line

constructed by drawing the perpendicular from A to the
representative line c of the ideal point,
(3)

is

The line 1212'

(4) Two points at infinity [12, 12'].
common parallel to the two given lines on which 12,

is

the

12'

lie.

These lines are not parallel to each other or 12 and 12' would
be the same point. The construction of this line is given below
in 44.
(5) An ideal point [Tg] and a point at infinity [12] not lying
on the representative line c of the ideal point. The line rci2
is the line which is parallel to the direction given by 12 and

perpendicular to the representative line c of the ideal point.

The construction of this line is given below in 45.

NON-EUCLIDEAN GEOMETRY

68

Two

(6)

intersect

ideal points [F^, F^],

and are not

parallel.

when
The

the lines c

[ch.

and

line F^F^, is the

c'

m.

do not

common

perpendicular to the two not -intersecting lines c and c'. The

construction of this line was given in 32.
The pairs of points which do not determine a line are as
follows

An

ideal point and a point at infinity lying on the representative line of the ideal point.
(i)

(ii)

or

Two

meet

in

whose representative
an ordinary point.*

ideal points,

lines are parallel

39. With this notation the theorems as to the concurrence

of the lines bisecting the sides of a triangle at right angles, the
lines bisecting the angles of a triangle, the perpendiculars

from

the angular points to the opposite sides, which hold in the

Euclidean Geometry, will be found also to be true in this NonEuclidean Geometry. Lines will be said to intersect in the
Also, in speaking of triangles, it is not
sense of 37, 38.
always necessary that they should have ordinary points for
their angular points. The figure of 26 is a triangle with one
angular point at an improper point a point at infinity. It
will be seen that a number of the theorems of that section
are analogous to familiar theorems for ordinary triangles.
With regard to the concurrence of Ifnes in the triangle we
the perpendiculars through the
shall only take one case
middle points of the sides.

The perpendiculars to the sides of a triangle at their middle

points are concurrent.
Let

ABC be

the triangle and D,

sides opposite A,

B and

E,

F the middle points of the

C.

Case (i) If the perpendiculars at the middle points of two

of the sides intersect in an ordinary point, the third perpendicular must also pass through this point. The proof depends
on the congruence theorems as in the Euclidean case.

*In the foundation of Projective Geometry independent of the

Parallel Postulate, this difficulty is overcome by the introduction of
new entities, called improper lines, and ideal lines, to distinguish them
from the ordinary or proper lines, Cf Bonola, lac. cit. English transla.

tion,

App. IV.

CONCURRENCE OF BISECTORS

38, 39]

Case

(ii)

Let the perpendiculars at D and E be not-inter-

secting lines,

From

69

A, B,

and let D'E' be the line perpendicular to both.

and C draw A A', BB', and CC perpendicular to

D'E'.

Then

it is

not

difficult to

show from congruent

AA'-CC and

BB'

triangles that

= CC'.

Thus AA' = BB'.

Let

F'

be the middle point of A'B'.

From 28

it

follows that FF'

is

perpendicular to

AB and A'B'.

Therefore, in this case the three perpendicular bisectors of

the sides meet in an ideal point.

Case

(iii)

There remains the case when the

D and E perpendicular

lines

to the sides are parallel.

through

It follows

from Cases (i) and (ii) that the perpendicular to the third side
through F cannot intersect the other perpendiculars either in
an ordinary point, or in an ideal point. It must therefore be
parallel to these two lines in the same sense
or it must be
;

parallel to the first in one sense

and

to the other in the opposite

sense.

alternative we shall show to be impossible ; so

necessarily will be true.
When the angular points of a triangle are all at infinity
(ft', 0", ft'") a straight line cannot cut all three sides.
For
if it cuts two of them at P and Q, say, PQ produced must

The second

the

first

be one of the rays through

side.

(Cf. Fig. 46.)

Q which does

not intersect the other

NON-EUCLIDEAN GEOMETRY

70

[CH.

III.

But if BC
is

is the greatest side of the triangle, the angle at

the greatest angle.

Pio. 46.

then,

If,

BC.

Let

cut

at Q.
perpendicular to AC.
similar argument applies to the perpendicular through
it

Then EQ

we make .lCAP = lACB, AP produced must cut

(Fig. 47.)
it

is

F.

Fio. 47.

Therefore the perpendiculars at E and F both intersect BC.

It follows that the three perpendiculars cannot form a triangle whose angular points are all at infinity.

39, 40, 41]

THE PARALLEL CONSTRUCTIONS

71

Therefore they are parallel to one another in the same sense

a point at infinity.
intersect in an improper point
If we take these three cases together, it will be seen that the

and

theorem
40.

is

established.

The Parallel Constructions.

In Hubert's Parallel Axiom the assumption is made that

from any point outside any straight line two parallels can
always be drawn to the line. In other words, it is assumed that
to any segment 'p there corresponds an angle of parallelism
n(p).

The fundamental problems

parallels are the following

of construction with regard to

L To draw the parallel to a given straight line from a given

point towards one end.
2.

To draw a

given straight
line

shall

be parallel to one

to another given straight

which intersects the former.

In other words
1.

which
and perpendicular

straight line

line,

Given

j),

to find

11(2)).

Given 11(2)), ^^ ^^^ Vand

For both of these problems Bolyai gave solutions
one was discussed by Lobatschewsky. In both cases the
argument, in one form or other, makes use of the Principle of
2.

Continuity.
In the treatment followed in this book the Hyperbolic
Geometry is being built up independently of the Principle of
For that reason neither Bolyai's argument
Continuity.
(Appendix, 34, 35), nor Lobatschewsky's discussion * of the

second problem,

be inserted.

will

41. To draw
a Point outside

the Parallel to a given Line from

Bolyai's Classical Construction

it.

(Appendix, 34).
To draw the 'parallel

to the straight line

Bolyai proceeds as follows

AH from a given point

D,

Draw the perpendiculars DB and EA to

perpendicular DE to the line AE.

AN

(Fig. 48),

and the

*Cf. Lobatschewsky, Geametrische Unterstichunyen znr Theorie der

Also Neiv Principles

Parallelinien, 23 (Halsted's translation, p. 135).

of Oeometry, 102 (Engel's translation).

NON-EUCLIDEAN GEOMETRY

72

[CH.

III.

The angle EDB of the quadrilateral ABDE, in which, three

angles are right angles, is a right angle, or an acute angle,
according as ED is equal to or greater than AB (cf. 29).
With centre A describe a circle whose radius is equal to ED.
It will intersect DB at a point C, coincident with B, or between B and D.

The angle which the line AC makes with DB is the angle of

parallelism corresponding to the segment BD.
Therefore a parallel to AN can be drawn by making the angle
BDM equal to the angle ACB.
Bolyai's proof is omitted for the reasons named above ; but
it should be remarked that his construction holds both for the
in his language it
Euclidean and Non-Euclidean Geometries
belongs to the Absolute Science of Space.
;

42. The correspondence which we have established in 35

between the right-angled triangle and the quadrilateral with
three right angles and one acute angle, leads at once to Bolyai's
construction.

We

have seen that,

there corresponds

to the

right-angld triangle a, b,

c, (X, ju),

a quadrilateral with three right angles and an

acute angle /3, the sides containing the acute angle being c and
and the other two, a and m'.
Therefore we can place the right-angled triangle in the
quadrilateral, so that the side a of the triangle coincides with
the side a of the quadrilaterial, and the side h of the triangle
Then the hypothenuse
lies along the side I of the quadrilateral.
of the triangle will be parallel to the side c of the quadrilateral,
1,

since

it

makes an angle

^ ~

f^

with m'.

THE PARALLEL CONSTRUCTIONS

41,42,43]

73

43. Second Proof of Bolyai's Parallel Construction.

The following proof of the validity of Bolyai's construction is due to Liebmann * it will be seen that it depends
(1) on Theorem (2) of 4, regarding the locus of the middle
points of the segments A A', BB', etc., joining a set of
points, A, B, C, ...
A', B', C, ., on two straight lines,
such that AB = A'B', BC = B'C', etc.; and (2) on the con:

currence

bisectors of the sides of

Let A be the given point, and AF the perpendicular from

to the given line.
It is required to draw from A the parallel to the ray Fi2.
Let us suppose the parallel Al] drawn.

triangle

From

of the perpendicular
(cf.

Afi

39).

and

FI2

cut

off

equal segments AS and FD, and

join SD.

M and M' be the middle points of AF and SD.

From 4 we know that the line MM' is parallel
Let

to Afi

and FO.

Draw the line I2"AI2' through A perpendicular to AF, and

produce M'M through the point M.
Then it is clear that the ray M'M is parallel to the line AO".
Draw from F the parallel FJ2' to All', and let it intersect Ai2
in G.

From
The

FO' cut off FS' equal to AS. Join SS' and S'D.
GM bisects SS' at right angles, and is perpendicular

line

to the line

I2i2'.

Also the perpendicular bisector of DS' bisects the angle DFS',

and is perpendicular to 1212'.
* Ber. d. k. scichs. Gea. d.

(1910)

Wiss. Math. Phya. Klaaae, vol.

also NichttuMidische Geomeirie (2nd ed. ), p. 35.

Ixii.

p,

35

NON-EUCLIDEAN GEOMETRY

74

Lch.

m.

These two bisectors have therefore an ideal point in common,

and the perpendicular bisector of SD must pass through the
same ideal point (cf. 39) i.e. it must also be perpendicular
;

to

1212'.

parallel M'12' drawn through M' to AO'.

The bisector of the angle S2'M'i2 is perpendicular to 1212',
and therefore to SD.
It follows that M'S bisects the angle i2"M'i2'.
But M'12" and M'12' are the parallels from M' to i2"Al2'.
Therefore M'S is perpendicular to 12"Al2'.
And AS was made equal to FD in our construction.
The result to which we are brought can be put in the following words
Let the perpendicular AF be drawn from the point
A to the given line a (F12), and let the perpendicular A12' be
drawn at A to AF. From any point D on the ray F12 drop the
This line DB cuts off from the
perpendicular DB to Al2'.

Suppose the

parallel Ai2 a length equal to FD.

The parallel construction follows immediately.

We need only

describe the arc of a circle of radius FD with A as centre. The

parallel A12 is got by joining A to the point at which this arc

cuts DB.
The existence of the parallel, given by Hilbert's Axiom,
allows us to state that the arc will cut the line once between
B and D, without invoking the Principle of Continuity.*

44. Construction of a Common Parallel to two given

Intersecting Straight Lines.!
Let 012 and 012' be the two rays a and 6 meeting at O and
containing an angle less than two right angles.
From these rays cut off any two equal segments OA and OB.
From A draw the parallel Al2' to the ray 012', and from B
the parallel B12 to the ray 012.
Bisect the angles 12A12' and
By 26 (4), we know that
Z.0A12'

12B12'

by the rays

a'

and

b'.

= ^0B12.

In pjuclid's Elements the fundamental problems of construction of

Book I. can be solved without the use of Postulate 3 "To describe a
To draw the parallel from a
circle with any centre and distance."
given point to a given line can be reduced to one of the problems of
On the other hand, in the Hyperbolic Geometry, the parallel3.
*

construction requires this postulate as to the possibility of drawing

a circle.
t Cf. Hilbert,

loc. cit. p.

163^

THE COMMON PARALLEL TO TWO LINES

43, 44]

It follows that

We

= ^I2Bl2',
^i2AE = z.i2'BF = ^12BF.
^I2Ai2'

now show

shall

75

that the lines a' and

6'

neither intersect,

nor are parallel.

If possible, let

The

them

intersect at

M.

AOB is isosceles, and /-OAB=^OBA.

l BAM =z_ ABM, and AM = BM.

triangle

Therefore

M
Fio, 50.

Through M draw the

Then, since

^^^^^^^
which

The

parallel M12 to Afi

and

BI2,

AM = BM and ^MA12 = ^MBi2, by 26

^AM12 = ^BM12,

(4),

we

absurd.

is

lines

point,

and

A and

B.

AE and BF therefore do not

an ordinaryproduced through

intersect at

this proof applies also to the lines

Next, let us suppose that they are parallel.

Since the ray a' lies in the region BAI2, it must intersect BI2.
Let it cut that line at D.
Then we have z.i]AE = i.DBF, and ^ADI2 = ^BDE.
Also we are supposing DE and BF parallel, and we have
A12

and

DJ2 parallel.

NON-EUCLIDEAN GEOMETRY

7o
It follows

from 26

Therefore we have

-l

[ch.

m.

that AD = DB.
DAB = ^ DBA.

(6)

But ^ BAG =^ ABC.

Therefore we have z_DAB=^CAB, which is absurd.
Thus, the rays AE and AF cannot be parallel.
Similarly the rays EA, FB produced through A and B cannot
be parallel.

We have now shown that the lines a' and h' neither intersect
nor are parallel.
^
They must, thereforCj have a common perpendicular ( 32).
We shall now show that this common perpendicular is
parallel to

both 012 and

0S2'.

cut the lines AE and BF at U and V.

Then AU = BV, by 29.
If VU is not parallel to AS2, draw through U the ray U12
parallel to AI2, and through V the ray Vi2 parallel to Afi.
Then, by 26 (4), l AUO =/. BVfi.
Also the angles AUV and BVU are right angles, so the exterior
angle at U would be equal to the interior and opposite angle
12 VU, which is impossible (26 (3)).
Thus we have shown that the ray VU is parallel to 012.
The same argument applies to the ray UV and 012'.
Therefore we have proved that there is a common parallel
to the two given intersecting rays, and we have shown how to
construct it.

Let

it

Corollary.

A common

'parallel

can he drawn

to

any two

given coplanar lines.

If the given lines intersect when produced, the previous
proof applies.
If they do not intersect, take any point A on the line (i) and
draw a parallel from A to the line (ii).
We can now draw a common parallel to the two rays through
A, and by 25 this line will also be parallel to the two given
lines.

45. Construction of the Straight Line which is perpendicular to one of two Straight Lines containing an
Acute Angle, and parallel to the other.
Let a{OA) and
angle.

(OB) be the two rays containing an acute

44,45,46]

GIVEN ANGLE OF PARALLELISM

77

At O make /.AOB'=^AOB, and denote the ray OB'

by b'.
The common parallel to the rays b and b' will be perpendicular to

O A.

(Cf.

22.)

We

have thus solved the

second fundamental problem
of parallels. To a given angle

of parallelism to find the correIn other

sponding segment.
words, given n(^) to find p.
Incidentally we have also
shown that to any acute angle
(p), however small, or however near a right angle, there
corresponds a segment p.

Corollary.
planar

secting lines,

line

If
are

lines

we can

parallel

perpendicular

to

two

co-

not-interstill

one

draw
and

PlO. 51,

to the other.

We

need only take a point on the line (i), and draw from
a ray parallel to the line (ii). The line perpendicular
to (i) and parallel to the ray just drawn will be parallel to the
it

line

(ii).

46. Corresponding Points on two Straight Lines.

P and Q are said to be corresponding points on two straight
lines when the segment PQ makes equal angles with the two lines
on the same side.
If the lines intersect at an ordinary point O, and P is any
point upon one of them, we need only take OQ = OP, and
the point Q on the second line will correspond to P on
the first.
Obviously there is only one point on the second ray corand if R is the point
responding to the point P on the first
corresponding to Q on a third ray through O, then P and R
are corresponding points.
Also the locus of the points on the rays of a pencil, whose
vertex is an ordinary point O, which correspond to a given
point P on one of the rays, is the circle with centre O and
radius OP.
;

NON-EUCLIDEAN GEOMETRY

78

[CH. III^

47. We proceed to the case when the lines are parallel

and thus intersect at an improper point (a point at infinity).
1. // (i) and (ii) are any two parallel straight lines, there
exists one and only one point on (ii) which corresponds to a given
point on

(i).

Pio. 52.

Let P be the given point on

(i)

Bisect the internal angles at P

meet

in

an ordinary

and take any point R on (ii).

and R. The bisectors must

point.

I\

Pio. 53.

Let them meet at

pendicular to

(i)

and

S,
(ii).

and from S draw SM and SN

per-

CORRESPONDING POINTS

47]

Then SM=SN.
Through S draw
also

will

It

79

h
Si2 parallel to P12.

be parallel to

Ri2,

and

it

will bisect

^MSN,

only one angle of parallelism for a given distance.

Let S' be any point upon the parallel through S to (i) and (ii).
From S' draw S'M' and S'N' perpendicular to these lines.
By congruence theorems, it is easy to show that S'M' = S'N',
and that S'i2 bisects z.M'S'N'.
From P draw PL perpendicular to Si2, and from L draw
Lm and Lw perpendicular to (i) and (ii). (Cf. Fig. 53.)
Cut off wQ = mP on the opposite side of n from 12, and
join LQ.
Then it follows that PLQ is a straight line, and that Q
corresponds to P.
It is easy to show that there can only be one point on the
second line corresponding to P on the first.
since there

is

If P and Q are corresponding points on the lines (i) and (ii),

and R corresponding points on the lines (ii) and (iii), the
three lines being parallel t<f each other, then P, Q, and R cannot
2.

and

he in the

same

straight line.

JTl

Fio. 54.

PQR be a straight line.

the definition of corresponding points,

If possible, let

By

^fiPQ = ^l]QP,
^i2QR = ^i2Ra

we have

NON-EUCLIDEAN GEOMETRY

80

[CH.

III.

= two right angles, which is imPR would make equal alternate angles with PD,
and RO, and these two parallels would have a common perTherefore zLl2PR+^f2RP

possible, since

pendicular.

If P corresponds to Q on the parallels (i) and

R on the parallels (ii) and (iii), then P corresponds

3.

to

parallels

(i)

and

(ii),

to

and

R on the

(iii).

Fig. 55.

This follows from the concurrence of the perpendicular

bisectors of the sides of a triangle ( 39).
The perpendicular bisector of PQ is parallel to the given lines ;
the same holds of the perpendicular bisector of QR,
It follows that the line bisecting PR at right angles is parallel

to the other two bisectors, and to

Therefore P and R correspond.

48.

(i)

and

(iii).

The Limiting-Curve or Horocycle.*

We now come to oe of the most important curves in the

Hyperbolic Geometry.
The locus of the corresponding points on a pencil of parallel
lines is a curve called the Limiting-Curve or Horocycle.

It is clear that this is the circle of infinite radius,

(2) it follows that it is not a straight line.

and from

47
*

Lobatschewsky uses the terms

cycle

Bolyai speaks of the linea-L,

grenzkreis, courbe-limite,

and hwi-

THE LIMITING-CURVE

47, 48]

81

Let P and P' he any two dijferent 'points upon the same ray of a
pencil of parallel lines; the Limiting-Curve through P is congruent

with the Limiting-Curve through

P'.

Fio. 56.

We

must

first

explain what

we mean by two

Limiting-

Curves being congruent.

We suppose a set of points obtained on the Limiting-Curve
which starts at P' e.g. P', Q', R', S', etc., on any set of lines
;

1, 2, 3, 4,

...

of the pencil.

We shall show that a set of points P, q, r, s, etc., exists on

the Limiting-Curve through P, such that the segments Pq,
P'Q' are equal, the segments qr, Q'R' are equal, etc., and
these related linear segments make equal angles with the lines
of the pencil which they respectively intersect.
To prove this, take the segment P'Q'.
At P make _12P)^z.l2P'Q', and take Pj = P'Q'.
From q draw the ray parallel to P12.
Then, by 26 (4), we know that .PjI2=z.P'Q'fi.
But P' and Q' are corresponding points.
Therefore P and q are corresponding points.
Proceeding now from Q' and q respectively, we find a point
r on the Limiting-Curve through P, such that the segments qr
and Q'R' are equal, while qr makes the same angles with the
N.-E.G.

NON-EUCLIDEAN GEOMETRY

82

m.

[ch.

rays passing througli its ends, as Q'R' does with the rays
through its ends.
We have thus shown that between the two Limiting-Curves
there is a one-one correspondence of the nature stated, and in
this case we say that the two curves are congruent.
Further, it is clear that it is immaterial at which line of
the pencil we begin our Limiting-Curve.
It is convenient to speak of the point at infinity, through
which all the parallel lines of the pencil pass, as the centre of
also to call the lines of the pencil the
the Limiting-Curve
axes of the curve. Concentric Limiting-Curves will be Limiting;

Curves with the same centre.

We can now state the following properties of these curves

in the Hyperbolic Geometry corresponds to the circle with infinite radius in the EucUdean
(a)

The Limiting-Curve

Geometry.
(6)

Any two

Limiting-Curves are congruent with each other.

(c) In one and the same Limiting-Curve, or in any two

Limiting-Curves, equal chords subtend equal arcs, and equal
arcs subtend equal chords.

The Limiting-Curve cuts all its axes at right angles, and

curvature is the same at all its, points.

{d)
its

49. The Equidistant-Curve.

There remains the pencil of
the set of lines
1.

all

lines

through an ideal point

perpendicular tojthe same

line.

If two given lines have a

common perpendicular, to any

point P on the one corresponds
one and only one point Q on the

other.

Let MN be the common perpendicular to the given lines,

and P any point on one of them.
From the other line cut off
NQ= MP, Q being on the same
side of the common perpendicular as P.

Then PMNQis one of Saccheri's

and Q are equal.

at P

j\j

f,o

Quadrilaterals,

57

and the angles


-18,

THE EQUIDISTANT-CURVE

49]

Thus Q corresponds to
one point on the second
on the
2.

83

and as before there can only be

P,

line

corresponding to a given point

and

(iii)

first.

If the lines

(i),

(ii),

same

straight line, then if the point

point P on (i), and the point R on

the points

are all perpendicular to the

on (ii) corresponds to the

(iii)

to the

point

on

(ii),

P and R correspond.
(i)

(iii)

(" )

^IQ

R
-

N
Fio. 68.

Let the common perpendicular meet the lines in M, N, and

Then PIVI=QN and QN=RS.
Therefore PM = RS, and P and R correspond.
3.

The

the fact that the points

distance

from

perpendicular.

On
line.

upon a pencil of lines

an Equidistant-Curve,
locus are all at the same

locus of corresponding points

is an ideal point is called

whose vertex

from

S.

the line to

This line

upon

which

the

all the lines

of the pencil are

of the curve.

is called the base-line

the Euclidean Plane the Equidistant-Curve is a straight

On the Hyperbolic Plane the locus is concave to the

common

perpendicular.

This follows at once from the properties of Saccheri's Quadrilateral (cf. 29). Indeed Saccheri used this curve in his supposed refutation of the Hypothesis of the Acute Angle.
We have thus been led to three curves in this Non-Euclidean
Plane Geometry, which may all be regarded as " circles."
(a) The locus of corresponding points upon a pencil of lines,
whose vertex is an ordinary point, is an ordinary circle, with
the vertex as centre and the segment from the vertex to one of

the points as radius.

(6) The locus of the corresponding points upon a pencil of
lines, whose vertex is an improper point
a point at infinity

NON-EUCLIDEAN GEOMETRY

84

[CH.

iir.

a Limiting-Curve, or* Circle of Infinite Radius, with its centre

at the vertex of the pencil.
(c) The locus of corresponding points upon a pencil of lines,
whose vertex is an improper point an ideal point is an Equidistant-Curve, whose base-line is the representative line of the
ideal point.
is

ABC

at
According as the perpendiculars to the sides of a triangle
their middle points meet in an ordinary point, a point at infinity, or an
ideal point, the points
determine an ordinary circle, a limiting-curve,
or an equidistant-curve.
(Cf. 39.)

ABC

THE MEASUREMENT OF AREA.

V
\A
,

50. Equivalent Polygons.

Two polygons
up

into

are said to be equivalent when they can he broken

triangles congruent in pairs.

a finite number of

r~

'

),'

10

a;

/
/
1

1
1

\
1

,-\

'

.''

;4-

Fig. 59.

With

this definition of equivalence,

following theorem

we

shall

7/ two polygons P^ and P^ are each equivalent

Pg, then

P-y

and

now prove

the

Pg are equivalent to each other.

to

a third polygon

EQUIVALENT POLYGONS

49.50,51]

87

We are given both for P^ and P2 a partition into tria. pc'

such that to these two partitions correspond two partition^
of Pg, the triangles in the partitions of P3 being congruent in
pairs to the triangles in the partitions of P^ and Pg.
in
Consider the two partitions of P3 simultaneously
general, every triangle of the one partition will be cut into
polygons by the sides of the triangles of the second partition.
;

We now introduce (cf. Fig. 59) a sufficient number of linear

segments, so that each of these polygons shall be cut into
triangles.

By this means the two partitions of P3 are further reduced

to the same set of triangles, and this can be associated with a
set of triangles in P^ and Pg respectively.
Therefore the polygons P^ and Pg can be broken up into a
finite number of triangles congruent in pairs, and they are
equivalent to each other.
51. Equivalent Triangles.
A necessary and sufficient condition that two
equivalent is that they have the same defect.
(Cf.

The theorem stated above

will

now be

triangles are
31.)

proved.

It

has to

be taken in several steps.

1.

other,

Two
and

triangles with
the

same

a side of

the ofie equal to

a side of

the

defect, are equivalent.

Consider the triangle ABC, in which E, F are the middle

points of the sides CA and AB.
Let the perpendiculars from A, B,
and C on EF meet that line at A',

and C.
Then AA' = BB'=CC', and

B',

the
quadrilateral BCC'B' is one of Saccheri's Quadrilaterals, the angles at
B', C' being right angles, and the
being equal.
sides BB' and
Further, the acute angles at B and C in that quadrilateral
are each equal to half the sum of the angles of the triangle

CC

ABC.

Now, the quadrilateral is made up of the triangles BB'F,

CC'E, and the figure BCEF,
Also the triangles BB'F and CC'E are congruent, respectively,
with AA'F and AA'E.

NON-EUCLIDEAN GEOMETRY
84
aerefore the quadrilateral

[cH. in.

BB'C'C and the triangle ABC are

equivalent.

Next,

let

AiBjCj be another triangle with

its

side BjCi equal

and the same defect as the triangle ABC.

For this triangle we get in the same way one of Saccheri's
Quadrilaterals, the acute angles at Bj and C^ being equal to
the acute angles at B and C, while the side B^Cj = the
to BC,

side BC.

easy to see that these quadrilaterals must be congruent,

they were not, we should obtain a quadrilateral, in which
the sum of the angles would be four right angles, by a process
which amounts to placing the one quadrilateral upon the
It

is

for if

other, so that the

common

sides coincide.

ABC and A^BjCi are equivalent.

Thus we have shown that triangles with a side of the one equal
to a side of the other, and the same defect, are equivalent.
It follows that the triangles

Corollary. The locus of the vertices of triangles on the

same base, with equal defects, is an Equidistant-Curve.
2.

Any

two triangles with the same defect and a side of the one

greater than

a side of

the other are equivalent.

Let ABC be the one triangle and A^BjCi the other, and let
the side AiCj(6^) be greater than the side AC (6).
Let E, F be the middle points of AC and AB.
From C draw CC perpendicular to EF CC cannot be greater
;

than

\h.

THE AREA OF A TRIANGLE

51]

87

Construct the right-angled triangle with a side equal to

CC

and ^bi for the hypothenuse.*

Cut ofi C'Eg equal to the other side of this triangle.
Join CEg, and produce it to Ag making CE2=E2A2.
Join AgB.
Then the triangle AgBC has a side equal to h^, and the same
defect as the two given triangles.
Also the triangles ABC and AgBC are equivalent
and the
triangles AgBC and AiB^Ci, by (1).
Therefore the triangles ABC and A^BjCi are equivalent ( 50).
;

Any two triangles, with the same defect, are equivalent.

For a side of one must be greater than, equal to, or less
3.

than, a side of the other.

When it is a case of equality, the triangles are equivalent

by

(1).

In the other two cases, the same result follows from

4.

The converse

of this

theorem also holds

(2).

Any two equivalent triangles have the same defect.

From the definition of equivalence, the two triangles

can be

number of triangles congruent in pairs.

But if a triangle is broken up by transversals f into a set of
sub-triangles, it is easy to show that the defect of the triangle
broken up into a

finite

equal to the sum of the defects of the triangles in this partiFurther, following Hilbert,| it can be shown that any
given partition of a triangle into triangles can be obtained by
successive division by transversals. It follows that the sum of
the defects of the triangles is equal to the defect of the original
is

tion.

triangle.

Now
finite

the two equivalent triangles can be broken up into a

number of triangles congruent in pairs. And the defects

of congruent triangles are equal.

* The construction of the riglit-angled triangle from a side and the
hypothenuse does not involve the Principle of Continuity. The results
of 36 show that this problem can be reduced to that of constructing a

right-angled triangle out of a side and the adjacent angle.

triangle is said to be broken up by transverscds, when the partit

tion into triangles is obtained by lines from the angular points to the
opposite sides, either in the original triangle or in the additional
triangles which have been obtained from the first by division by
transversals.
:J:Cf.

Hilbert,

loc. cit.

20, or Halsted, Rationed Geometry,

p. 87.

NON-EUCLIDEAN GEOMETRY

88

[ch.

m.

Therefore the defects of any two equivalent triangles are

equal.

The theorem enunciated at the head of this section is thus

established
a necessary and sufficient condition for equivalence of triangles is equal defect.
triangle is said to be equivalent to the sum of two
5.
other triangles, when the three triangles can be broken up
into a finite number of triangles, such that the triangles in the
partition of the first are congruent in pairs with the sum of the
triangles in the partitions of the other two.
Now the defect of each triangle is equal to the sum of the
defects of the triangles into which it is divided.
It follows that if a triangle is equivalent to the sum of two
other triangles, its defect is equal to the swn of their defects.
:

52. If we regard area as a concept associated with a rectilinear figure, just as length is with a straight line, it is obvious

that equivalent figures have equal area* And if, further, we

regard the area of a rectilinear figure as a magnitude to which
we can ascribe the relations of sum, equality and inequality,
greater and less, we obtain at once from the theorems of 51
the result that the areas of triangles are proportional to their
defects. Indeed if we start with any triangle as the triangle of
imit area, a triangle which is n times this triangle will have n
times its defect.
But closer examination of the argument shows that in this
treatment of the question of area various assumptions are
made and the work of some mathematicians of the present
day has put the theory of area on a sounder logical basis.f
This more exact treatment of the theory of area in the Hyperbolic Plane is simple, and will now be given
multiplied
The measure of area of a triangle is defined as
;

'

* Hilbert distinguished between equivalent polygons, as defined above,

and polygons which are equivalent by completion. Two polygons are
said to be equivalent hy completion, when it is possible to annex to them
equivalent polygons, so that the two completed polygons are equivalent.
If the Postulate of Archimedes is adopted, polygons, which are equivaHilbert was able to establish
lent by completion, are also equivalent.
the theory of area on the doctrine of equivalence by completion without
the aid of the Postulate of Archimedes. Loc. cit. Chapter IV.

+ Cf. Art. VI. by Amaldi,*in Enriques' volume referred to above.

Also Finzel, Die Lehre vom Fldcheninhalt in der allgemeinen Geometric
(Leipzig, 1912).

51,52]

THE AREA OF A POLYGON

by

k being a constant depending on the unit triangle,

its defect,

and the unit

its

measure.

of angle

is

chosen so that a right angle has

The number

h^ is

89

for

introduced to bring the results

work in other parts of this

into agreement with the analytical

book.
It follows

from

51 that
^

two triangles have the same measure of area, they are

equivalent, and that if two triangles are equivalent, they have
the same measure of area.
1.

If

2. If a triangle is broken up into a finite number of triangles,

the measure of area of the triangle is equal to the sum of the
measures of area of the triangles in the partition.
is equivalent to the sum of two other trimeasure of area of this triangle is equal to the sum
of the measures of area of the other two triangles.

3.

If

a triangle

angles, the

The measure of area of a polygon is defined to he the sum of the

measures of area of the triangles into which it is divided in any
given partition.

This sum is independent of the partition which has been

chosen. The sum of the defects of the triangles in any partition is equal to {n - 2) times two right angles - the sum of the
angles of the polygon. This is sometimes called the Defect of
the Polygon.

With regard
theorems

to polygons

we can now

state the following

1. If two polygons have the same measure of area, they are

equivalent. For they are each equivalent to the triangle whose
defect is the sum of the defects of the given partitions.
2. If two polygons are equivalent, they have the same

For they can be broken up into a finite

of triangles congruent in pairs.
3. If a polygon is broken up into a finite number of subpolygons, the measure of area of the polygon is the same as
measure of area.

number

sum

of the measures of area of the sub-polygons.

polygon is equivalent to the sum of two other polygons, its measure of area is equal to the sum of the measures
of area of these two polygons.

the

4.

If a

Rectilinear polygons with the same measure of area will be

said to have equal area. Thus equivalent polygons have equal

90

NON-EUCLIDEAN GEOMETRY

[ch.

m.

52, 53]

The area of a polygon will be said to be greater or less

than the area of another polygon according as its measure of
area is greater or less than the measure of area of the other.
area.

53. In the Euclidean Plane we say that a rectilinear figure

contains so many square inches (or sq. ft., etc.), and by considering a curvilinear figure as the limit of a rectilinear, figure
we obtain a method of measuring curvilinear figures.
In the Hyperbolic Plane there is no such thing as a square
inch, or rectangle with equal sides, or any rectangle. To every
rectilinear figure there corresponds an equivalent Saccheri's
Quadrilateral.
To all equivalent rectilinear figures there
corresponds one and the same Saccheri's Quadrilateral with a
definite acute angle.
This quadrilateral with a given acute angle can be constructed in this geometry immediately.
The construction
follows from the correspondence established between rightangled triangles and the quadrilateral with three right angles.
If the acute angle is /3, we obtain the corresponding segment
6{/3 = n(6)}, by the construction of 45. We draw any
right-angled triangle with a side equal to h. The associated
quadrilateral has its acute angle equal to 0, and the Saccheri's
Quadrilateral is obtained by placing alongside it a congruent
quadrilateral.
All Saccheri's Quadrilaterals with the same acute angle are
equivalent.
Thus it will be seen that there is a fundamental difierence
between measurement of length and area in the Euclidean

and the Hyperbolic Plane.*

In the Euclidean, the measures

In the Hyperbolic, they are absolute. With every
linear segment there can be associated a definite angle, namely
the angle of parallelism for this segment. With every area,
a definite angle can be associated, namely the acute angle
of the equivalent Saccheri's Quadrilateral.
are relative.

* Cf. Bonola,

loc. cit.

20.

Also supra,

p. 17.

CHAPTER

IV.

THE HYPERBOLIC PLANE TRIGONOMETRY.

54. In this chapter we shall develop the Trigonometry of
the Hyperbolic Plane, as in the preceding one we have discussed
the Geometry of the Hyperbolic Plane, without introducing
the theorems of Solid Geometry into the argument.
The properties of the Limiting-Curve lead to the formulae
of Plane Trigonometry, without the use of the Limiting-Surface,
as the surface formed by the rotation of a Limiting-Curve
about one of its axes is called. The method of Lobatschewsky
and Bolyai is foimded upon the Geometry upon that Surface.
We begin with some theorems upon Concentric LimitingCurves.
1. If A, B and A', B' are the 'points in which two Concentric
Limiting -Curves cut tivo of their axes, then AB = A'B'.

n.'

Fir..

62.

Join AA' and BB' (Fig. 62).

Through the middle point M of the chord AA' draw Mi2
parallel to the rays of the pencil.

NON-EUCLIDEAN GEOMETRY

92

[CH. IV.

Then Mfl is perpendicular to the chord BB', and is symmetrical to the two parallels AB and A'B' (cf. 26 (4) and 47).
Therefore it passes through the middle point N of BB'.
Then it follows from the quadrilateral ABB' A' that AB = A'B'.
2. // A, B and A', B' are the points in which two Concentric
Limiting -Curves cut two of their axes, and P, Q, are the middle

and BB', then PQ

points of the arcs A A'

is

line of the pencil.

Fio. 63.

Since equal arcs subtend equal chords (cf. 48), the chords
are equal, and the chords BQ and B'Q are equal.
It follows that PQ is the line of symmetry for the two axes

AP and A'P

AB and

A'B', and is parallel to both (cf. 47).

Corollary. If the points P^, Pg, Pg, P^, ... divide the arc A A'
into n equal arcs, and the axes through these points are met hy the
,

Limiting -Curve BB' in Q^, Q^, Qg, Q4, ... the points
Q3, Q4, ... divide the arc BB' into n equxil arcs.
,

3.

// A,

A',

is cut

Qg,

on a Limiting-Curve, and
which a Concentric Limiting-

A", are three points

B, B', B", are the three points in

Curve

Q.y,

by the axes through A,

arc

A',

and

A A' arc AA" = arc BB'

:

A", then

arc BB".

First, let the arcs A A' and AA" be commensurable, and let
the one be tn times the arc AP and the other n times the arc
AP.
Through P draw the line of the pencil. Let it cut the second
Limiting-Curve in Q.

64,55]

CONCENTRIC LIMITING-CURVES

93

Then we know from (2) that the arc BB' = m times the arc
BQ, and that the arc BB" = n times the arc BQ,

Fio. 64.

Thus

the proportion follows.

Secondly, if the arcs are incommensurable,

same conclusion by proceeding to the limit.

we reach the

ri

Fio. 65.

55. Let us start with a Limiting-Curve whose centre is 12,

and take any two points A and B upon the curve (Fig. 65).

''

NON-EUCLIDEAN GEOMETRY

94

[CH. IV.

On the ray AQ, cut off the equal segments AA^, A^Aa, AgAg,
Let the Concentric Limiting-Curves through A^, Ag, Ag,
cut the ray Bfi in B^, B.^, Bg, ...
Then we have, by 54 (1).

....
...

A Aj = B Bj = B^ Bg = B2B3 = etc.
from 48 and 54(3),

Also,

arcAB

arc AiBj

= arc A^Bj

arc A2B2 = arc AgBg

arc A3B3 = etc.

This ratio is greater than unity, and depends only on the

length of AAi
We may choose the unit segment so that the ratio is equal to
e, when AAi = AiA2 = A2A3 = ... =the unit segment.
Let the arcs AB, A^B^, AgBg, etc., be denoted by s, s^, $2,
etc., when the segment AA^ is the unit of length.

Then we have
.

Thus

Sn

S-t

^^ oj

= se-'^, when w

is

00 ^^ 09

Ot> ^^^

^ c.

a positive integer.

deduce from this that when the segment AP

X units, x being any rational number, and the arc
denoted by Sx, then we have

It is easy to
(Fig. 66) is

PQ

is

se-

We obtain the same result for an irrational number x by

proceeding to the limit.

55, 56]

ARCS ON CONCENTRIC LIMITING-CURVES

Therefore, with this unit of length

theorem

we have the

95

following

If ABDC (Fig. 67) is a figure hounded by two Concentric

Limiting-Curves AC and BD, and ttvo straight lines AB and CD,
the straight lines being axes of the curves, the lengths a and Sx of
the arcs AC and BD are connected by the equation

Sx^se
ivhen the segments
the external curve,

AB and CD are x
BD the internal.

units of length,

and AC

is

St

Fio. 67.

another unit of length had been chosen, so that the

of the arc AB (Fig. 65) to the arc A^Bj had been
a(a>l), when AAi = BBi=the unit of length, the equation
connecting s and Sx would have been
If

ratio

Sx

= sa-^.
1

Putting

= e^,
X

we have
The number

Sx

= se

*.

the parameter of the Hyperbolic Geometrydepending upon the unit of length chosen.

56.

Since

is

Jc

we can

find

to satisfy the equation

n(;;)

there

is

a point

J,
on the Limiting-Curve through

P,

such that

NON-EUCLIDEAN GEOMETRY

9G

[CH. rv.

the tangent at Q is parallel to the axis through P, in the

opposite sense to that in which the axis is drawn (Fig. 68).

We

shall for the present denote the length of this arc by S.*
Let B be a point on the Limiting-Curve through A, such that
the arc AB is less than S (Fig. 69).

It follows that the tangent at B must intersect the axis

through A.
Let it cut 12A in D, and let the segments AD and BD be u

and t. It is easy to show that u<t.

Produce the arc BA to the point

C,

such that the arc

BC = S,

On OD produced

take the point

Then the perpendicular through

BD, and therefore to Cl2'.

A^,

such that DAi

A^ to the axis

is

= DB =

Let the Limiting-Curve through Aj meet C12' in C^.

Since the tangent at A^ is parallel to Ci2', the arc fKjC^
*Cf.

p. 119.

^.

parallel to

= S.

66, 57]

EQUATION OF THE LIMITING-CURVE

It follows

from

97

55 that
arc AC = e+'.
S-s = Se-(+')
AB through B to the point

arc AjCj

Therefore

Next, produce the arc

the arcBP = S (Fig. 70).

(1)
P,

such that

>n.

Du
t
Fio. 70.

Let the tangent at B as before cut the axis through A

D,

and

let

AD=w

at

and BD=^.

On Afl, on the opposite side of A from D, take the point Q,

such that DQi = t.
Then the perpendicular through Q to the axis is parallel to
DB, and, therefore, to P12'.
Let the Limiting-Curve through Q cut the axis PI2 in R.
Since the tangent at Q is parallel to the axis through R,
arcQR = S.
But AQ=<-w.
S-Fs = Se-

Therefore

From

(1)

and

(2),

(2)

we have
e'*

and

,s

= cosh^,
= Stanh^

(3)

(4)

57. The Equation of the Limiting-Curve.

Let Ox and Oy be two lines at right angles, and let P be the
point {x, y) on the Limiting-Curve through O, with Ox for
axis (Fig. 71).

Draw PM

perpendicular to the axis Ox, and let the ConM cut the axis through P in N.

centric Limiting-Curve through

ThenOM = PN=x, MP =
= s, and arc MN =s'.
?/.

Let arc OP

From

the construction

N.-E.a.

it

follows that

s'

< S.

NON-EUCLIDEAN GEOMETRY

98

Now

[CH. IV.

the coordinates of P, a!;(OM), and y{MP) are, respecu and t of the previous section with reference to

tively, the

the arc .9'.

Therefore

we have, from

56

(3),

e^ = cosh y

(1)

Fio. 71.

This
its

is the equation of the Limiting -Curve through

axis coinciding with the axis of x.

Also,

we have

= s'e^
= Se* tanh y,
= Ssinh?/

from

56

with

(4),

(2)

58. The Hyperbolic Functions of Complementary

Segments.
Let Oa; and Oy be two lines at right angles, and let the
Limiting-Curve through O with axis Ox have the arc OP = S
(Fig. 72).

Let A be a point upon Ox, such that OA = a;, and let the
Limiting-Curve through A be cut by the axis through P in B.
Let arc AB = s.
At A draw the perpendicular to the axis of x. Since it must
cut PB, let it intersect it at C.
Produce AC through C to the point D, such that AC = CD.
At D draw DQ perpendicular to CD.
The line DQ must be parallel to CP, since z.DCP = z.ACB,

and CB

is

parallel to Afi.

Therefore Oy, CP, and

DQ

are parallel.

COMPLEMENTARY SEGMENTS

57.58]

segments

It follows that the

OA and AD

99

are complementary,

n(0A) + n(AD) = 5.

i.e.

With the usual notation

mentary segment to x.
Therefore,

(cf.

we take

Se-* = s = Stanh-.

It follows that

x' as

the comple-

AC = S--

OA = x,

if

27)

(56(4).)

Fio. 72.

Therefore for complementary segments

e-a;

we have

= tanhl.
2

But

gx_

g-a;

sinh x

sinh

= ^ / coth ^ - tanh

x'

.'.

.'.

.'.

Also

a;

x'\_

~ cosech x'.

cosh x = \/l+

sinh'' a;

2/

sinh

x'

= coth x'.

tanh X = sech x' and coth x = cosh a;'.

sech

a;

= tanh x' and

cosech

a;

= sinh x'.

100

NON-EUCLIDEAN GEOMETRY

[CH. IV.

59. The Equations connecting the Sides and Angles

of a Right- Angled Triangle.
Let ABC be any right-angled triangle, C being the right
angle.

Produce the side AC througb C, and draw the parallel BO

through B to AC.

AB through B to L, where AL is the segment

x = U{l). [\ = ^BAC (of. Fig. 37).]

Also produce
such that

I,

Through L draw Lfi parallel to B12 and AC.

Let the Limiting-Curves through B and L, with centre at
fi,

meet the axes at

B',

D,

and D'

(Fig. 73).

Let the arcs BB', DD', LD be denoted by

the segment BD=.
Then we have

S sinh a = s = s^e"
Si + 52 = Stanh/,
2
e'*

57

s,

s^, s^,

(2).]

[56(4).]
8 tanh BL = S tanh (I -

=
= coshBL = cosh(l-c).

c),

[56(3).]

It follows that

sinh a

= cosh (I - c){tanh I - tanh (l-c)}

sinh

cosh (l-c)- cosh I sinh

cosh

= sinh c/cosh

Z.

(I

c)

and

let


THE RIGHT-ANGLED TRIANGLE

591

101

= sinh a cosh 1
1.
(Hypothenuse, side, and opposite angle.)

Thus

sinh c

From this formula, connecting the hypothenuse, a side, and

the opposite angle of any right-angled triangle, we can obtain
the relations between all the other elements, by using the
associated triangles of 36.
We know that, starting with a right-angled triangle in
which the elements are
a,

we obtain

(X,

c,

^,

/u),

(1)

successively triangles with the elements

*',
w,

c,
o',

b,

I,

(y. f-oc),

(2)

|-/8),

(3)

m',

a',

(\,

C,

b',

(|-a,

I'y

m, (y,

a,

I',

fij,

(4)

^-pj

(5)

Fromfcdid second triangle

we have

sinh

= sinh m' cosh c

= ^. cosh c, by
smhm
Therefore

-^

^ 58.

= sinh 1 sinh m

cosh c

II.

(Hypothenuse and two angles.)

Also, from the

same triangle (by

sinh

Therefore

I.),

= sinh b cosh a'

= sinh b coth a.

tanha = ^^r-^

Ill)

(Two
Now,
.

we have

since

cosh

= sinh I sinh m,

cosh

sinh a

sinh b
=

tanh a

sides

z-^.
b

tanh

and an

angle.)

NON-EUCLIDEAN GEOMETRY

102
Therefore

cosh c

[ch.iv.

= cosh a cosh b

IV.

(Hypothenuse and two

cosh a = sinh

Further,

^!"j^?

(by

= sinh(jcoshm

=-

cosh

Therefore

cosh a

,,

(bv

III.

1.).

= tanh 1 cosh m
(Side and

Applying (IV.)

and

?/i',

cosh a

this gives

V.

two

angles.)

to the triangle

c',

we have

sides.)

\-p\^

a',-U,

= cosh c' cosh m',

VI.
tanh a = tanh tanh c
(Hypothenuse, a side, and included angle.)

These six formulae are all given by a rule similar to Napier's

Rules in Spherical Trigonometry
:

(i)

Ld

the,

letters a',

1,

c,

m,

of a pentagon taken in order.

cosh of the middle part

b' he written one at each

of the sides

Then

= the product

of the hyperbolic sines of the

adjacent parts

and
cosh of the middle part

= the

product of the hyperbolic cotangents

of the opposite parts.

THE OBLIQUE-ANGLED TRIANGLE

59, 60]

60.

The Equations

for

103

an Oblique- Angled Triangle.

In the case of the Oblique-Angled Triangle ABC, the sides

opposite the angular points A, B, and C will be denoted hy a, b,
and c, as usual ; but the angles at A, B, and C will be denoted

by

X,

/i,

With
at

and

v.

this notation the distance of parallelism for the angle

be /.
proceed to prove that

will

We
I.

sinh a sinh b
:

This corresponds
nometry.

sinh c

= sech 1
Sine

the

to

sech.

Rule

of

sech n.

ordinary Trigo-

Via. 75.

ABC
From an

Let

be all acute angles.

angular point, say A, draw the perpendicular AD to
the opposite side.
We then obtain two right-angled triangles

ABD and ACD, as in Fig. 75.

Writing AD=p, we have (by
sinh c

sinh^

and

cosh

sinh p
^

cosh n

59,

I.)

from the triangle ABD,

from the triangle ACD.

Thus we have
sinh b

sinh

= sech m

sech n.

Taking another angular point say B

the same way, we would have
sinh a

sinh

= sech

and

sech

n.

proceeding in

NON-EUCLIDEAN GEOMETRY

104

[ch

iv.

Therefore
sinh a

sinh h

sinh

= sech

sech

sech

n.

If one of tlie angles is obtuse, we obtain the same result,

using the notation ri( - a;) = tt - n(x).
For the right-angled triangle, the result follows from 59, 1.
II. We shall now prove the theorem corresponding to the
Cosine Rule of ordinary Trigonometry.
We take in the fii'st place the case when B and C are acute

angles.

From A draw the perpendicular AD to BC.

AD=p, CD = and BD = a-^ (Fig. 75).
Let
Then, from the triangle ABD we have
coshc = cosh(a-g)cosh^ (59, IV.),
5',

and from the triangle ACD we have

cosh b
Also,

tanh(a -q)
mi

= cosh^ cosh q.

we have

= tanh c tanh m
T-

cosh o

Therefore

( 59, VI.),

= coshr-.c cosh ^q
cosh(a-g')

= cosh c(cosh a cosh (a~q) - sinh a sinh (a - q))

cosh {a

- q)

= cosh a cosh c - sinh a cosh c tanh {a - q)

= cosh a cosh c - sinh a sinh c tanh m.
If the angle B is obtuse, so that D falls on CB produced,
the same result follows, provided account is taken of the
notation
Ii{-x) = -n{x).
nr

If the angle
59,

is

a right angle, the result follows from

IV.

We

are thus brought to the Cosine Formula, which

be put in the form

may

cosh a

61.

Up

= cosli b cosli c -

sinh b sinh c tanh 1,

The Measurement of Angles.

till

this stage,

except in

51-2,

there has been no

our work.
The

need to introduce a unit of angle into

>

THE MEASUREMENT OF ANGLES

60, 61. 62]

105

equation a = H (a), connecting the segment and the corresponding angle of parallelism, has had only a geometrical significance.
In it oc has stood for a certain definite acute angle, which has
the property that the perpendicular to one of its bounding lines,
at a distance a from the angular point, is parallel to the other

bounding

When

line.

comes to assigning numerical values to angles, the

choice of one number is sufficient, if, in addition, the angle
zero is denoted by O.
In the Non-Euclidean Trigonometry

we

it

shall assign the

number -

to the right angle.

angles will have the numerical values proper to

All other

them on

this

scale.

In the rest of this work, when we use the equation oc = 11 (a),

both oc and a will be numbers, the one the measure of the angle
on this scale, the other the measure of the segment on one of
the scales agreed upon below ( 55), in which the unit segment
is the distance apart of two concentric Limiting-Curves, when
1

the ratio of the arcs cut ofE by two of their axes is e or e*.
It should perhaps be remarked that in dealing with the
trigonometrical formulae in the previous sections the measure
of the segment, and not the segment itself, is what we have
meant to denote by the letters in the different equations.

The Trigonometrical Functions of the Angle.

The Trigonometrical Functions

62.

sinoc, cosoc, tanoc, etc.,

are defined by the equations

sin

a=

tan a.

la

-e

- ia

ia

TT-

= sin oc

cos OL

secoc =

The fundamental equation

^^

when

cot oc

cosecoc =

-,

cos

cos oc

+e

ia,

tan

sm oc

of the Hyperbolic Trigonometry

tanha = cosa,

a = n (a).

NON-EUCLIDEAN GEOMETRY

106

We

proceed to obtain this relation

Consider the function /(oc) defined

tanh a

and

let

[CH. IV.

*
:

l>y

the equation

= cos/(a,).

us write a = A(a.).

When

0.

= -^,

When a = 0,

a
a

= 0, tanha = 0,

= 30

tanh a =

cos/(a)

cos/ (a) =

1,

Further, as a increases from

tinuously

from

= 0;

to go

i.e.fl-^j
;

i.e.

/(O)

= ~.

= 0.

/(oc) diminishes con

to 0.

Next consider a triangle ABC not right-angled and let

the perpendicular from B cut the base AC at D.
Let the
elements of the triangle ABD be denoted by AB = c, BD = a,
DA = ft, z.ABD = yu, ^BAD = X. Also let the elements of the
triangle
BDC be denoted by BC = Cj, CD = bj^, DB = a^,

lBCD = \, LDBC =
As

the side

BD

is

FH=|^

jUL^,

common, a

Tr^.)=^(*'

Fio. 76.

60, we have
- cosh (b + bj)
suih c smh Cj

Then, from the Cosine Formula,

//_

tann
*

ut

cosh

cosh
;

Cj

ir^v-

of this and the preceding sections is due to Liebmann,

Ableitimg der nichteuklidischen Trigonometrie," Ber.
Oes. d. Wiss. Math. Phys. Klasse, vol. lix. p. 187 (1907), and

The method

" Elementare
d. k. aiichs.

Another method, also indeNichteuklidische Geometrie, 2nd ed. p. 71.

pendent of the geometry of space, is to be found in Gerard's work, and
in the paper by Young referred to below, p. 136.

^*^


A FUNDAMENTAL FORMULA

62]

With the
cosf{fjL

'

we have

notation of this section,

+ fjL^) = tanh A
cosh

+ ju^)

(jut.

cosh

- cosh

Cj

sinh

sinh

COth

C,

(b

cosh

snih

ft,

sinh

Cj

sinh

^-,

;-^j

snih

'

b^)

cosh

,,

COth

107

ft

sinh

6,

sinh

Cj

^^j

;r

But we know that

tanh a = tanh

tanh m,

[ 59,

VL]

tanh a = tanh c cos/(yu).

i.e.

tanh a^

= tanh c^ cos/(^j).

coth

= coth^a cosf{/j.) cos/(/Xj).

Similarly

coth

Therefore

Further, from

sinh

We are

ftj

sinh

Cj

left

cosh

m =

sinh

-r-r
sinh

we obtain

L,

59,

sinh 6

Therefore

fj

cosh m^
^

sinh

Cj

//

sin/ (fX),
v*- /'
-^

_
~

//
\
^^^''

''

= sin/(u)
^'^^').
^'^^ sin/(ui
"^

'

with the term

cosh b cosh

ftj
*

sinh

But, from

59,

VL

and

tanh
.

Therefore

IV.,

cosh

ft

cosh

b^

sinh

sinh

Cj

Cj

we have

m_

sinh a

sinh

cosh
sinh

cosh a

_ cosh b
sinh

^ cos/(/i) cos fifx^)

sinh-^a

Thus we obtain
cos/(/i

+ ^aj) = coth^a cos/(/oi) cos/(/ij)

-

cosech-rt cosf{/u) cos f(fjL^)

- sin/(/A) sin/(jUj)

= cos/(/a) cosf{^{) - sin/(yu) sin/(/Xi)

= cos[/(ya)+/(Mi)].

NON-EUCLIDEAN GEOMETRY

108

But when

/x

[ch.iv.

= /u^ = ^t + /x^ = 0,

/(m)=/(Mi)=/(a^ +

/i)

= 0.

we have

Therefore

This is a functional equation from which the continuous

function /(/x) is to be derived.
It

may

be written

with

/(0)

Thus

sve

+ y) =f{x) +f{y),

f{x

= 0,

/(f)

f-

have
f(x

+ h)-f{x) _ f(j/ + h)-f{y)

h

Proceeding to the limit

Thus
Therefore

The
have

values of /(O)

finally

'

(^)

f(x)

= constant.

= Aa; + B.

and/(^j

determine A and

^^'^

f{x)

Thus we

= x.

are led to the desired equation

tanh a = cos a.
63. From the result proved

in last section,

tanha = cosa,
it

follows immediately that

sinha = cotcx.,

cosha = coseca,
cotha =
sech a

cosech a

seccx.,

= sin a,
= tan

oc.

B, so

that

we

62,63,64,65]
If

we

TRIGONOMETRICAL FORMULAE

insert these values in the Trigonometrical

we obtain

of 59,

= sinh c sin \ from

= tanh a cot A
cosh c cot X cot /x

cosh c = cosh a cosh 6

cosX = cosh a sin

tanh a = tanh c cos yu

sinh 6

,,

y(jt

= sinh a cosh
& = tanh a sinh
cosh c = sinh sinh m.
cosh c = cosh a cosh 6.
cosh a = tanh / cosh w.
tanh a = tanh m tanh c.
sinh

I.

sinh

Z.

the formulae of 60 for the Oblique- Angled Triangle

become

gjj^j^

gjj^j^ ^

cosh a = cosh

^Z/

Formulae

sinh a

And

109

^Aese

results

agree

^^^j^ c

cosh

= sin X

sin

/x

sin

v,

coire^onding formulae in

the

toith

- sinh h sinh c cos X.

v take the place of A, B, C, and

the Hyperbolic Functions of a, b, and c take the place of the Circular

when

Spherical Trigonometry,

Functions of

a, b,

and

yu,

c.

The Angle of

64.

X,

Parallelism.

tanh a = cos

Since

we have

1
1

- cos a. _

a.

- tanh a

+ cos a ~ 1 + tanh a'

Therefore

tan^

= e-^"^,

tan^ = e-

and

The angle

oc is acute, so

the positive sign has to be taken

in extracting the square root.

This

may

be written
tAn\'n.{p)

= e-P*

65. The formulae of 56-64 have been deduced on the

understanding that the unit of length employed is the distance
between concentric Limiting-Curves when the ratio of the arcs
cut off by two of their axes is e.
* This result is given by Bolyai, Appendix, 29, and by Lobatschewsky
in his various books, e.g. Oeometrische Untersuchungen zur Theorie der
Parallellinien, 36.

no

NON-EUCLIDEAN GEOMETRY

[ch. iv.

If a different

unit is adopted, so that the ratio of the arc

to the arc AjBj is a, any number greater than unity, we
have the equation

AB

Sa

instead of Sx = se- ^.

= 5a ~ *
I

Putting
this gives

a
Sx

= e*,

= se

*.

This parameter k will enter into

all

the equations of the

preceding sections, so that sinhy, cosh, etc., will replace

^
"'
sinh a, cosh a, etc.
And the equation for the Angle of Parallelism will be

_p

tan|n(^) = e"*.
The Euclidean Geometry now appears as a special case of
the Hyperbolic Geometry, for if we let k-^cc the formulae
of this Non-Euclidean Geometry reduce to those of the
Euclidean.
In the first place, since
,

tan|n(^) = e'<^,
the angle of parallelism becomes

when

k->oo

Further, the equations connecting the sides and angles of a

right-angled triangle, viz.
sinh

sinh

-J-

= sinh ^ sin X,

-J-

= tanh ^ cot X,

a;

cosh
.

cosh

cos

-r-

c
j^

= cot X cot
,

= cosh -r cosh -r

X = cosh

tanh -r
k

fjL,

= tanh

-r-

-y-

sin

jul,

cos u,

INFINITESIMAL GEOMETRY

65]

becomle

sin

111

A=

cotA =cot A cot

a2

B ==

>

a
1,

+ &2 =-c\

cosA == sin

B,

cosB =

when we write A, B for \ and

From the Sine and Cosine Formulae
Triangle ( 63) we get at once,
/ul.

sin

A
a2

Again,

y, y,

and

sin

= 62 +

- can be

^2

sin

C=a

for the Oblique-Angled

&

c,

-2k cos A.

made

infinitesimals

and

c tend to zero instead of h to infinity.

the Euclidean relations are obtained.
This result can be stated in other terms

by

letting a, 6,

In this case again

:

immediate neighbourhood of a -point on the Hyperbolic

Plane, the formulae of the Euclidean Geometry hold true.

In

the

Or, again

The Euclidean Formulae hold

on

,'

true in Infinitesimal Geometry

the Hyperbolic Plane.

These theorems have an important bearing upon the question

as to whether the Hyperbolic Geometry can actually represent
The
the external relations of the space in which we live.
experimental fact that, within the limits of error to which all
actual observations are subject, the sum of the angles of a
triangle is two right angles does not prove that the geometry
of our space is the Euclidean Geometry. It might be a Hyperbolic Geometry in which the parameter k was very great.
The Geometry of Bolyai and Lobatschewsky can be made
to fit in with the facts of experience by taking k large enough.
The Postulate of Euclid reaches the same end by another
means. It is a better means, for it gives a simpler geometry.

[CH. V.

CHAPTEK

V.

MEASUREMENTS OF LENGTH AND AREA, WITH THE

AID OF THE INFINITESIMAL CALCULUS.
66. In this Chapter we shall apply the Trigonometrical
Formulae found in Chapter IV. to the measurements of Length
and Areas of Curves.
The first thing to be done is to obtain the expression for
the element of arc of a plane curve.

The Element of Arc

in Cartesian Coordinates.

In the Euclidean Plane

ds^

We

shall

now prove

that in the Hyperbolic Plane

ds2

Let

P,

Q be

= cosh2 ? dx2 + dy2.

k

the points

Draw PM and QN

= dx^ + dy"^.

{x, y), (x

+ Sx,

+ Sy).

per-pendicular to the axis of

x.

Then 0M=, MP^'^, 0[^=x + Sx, and NCl = y + Sy.

From P draw PH perpendicular to QN.
Let PQ, = Ss, PH=^, HQ=^, and NH=2;.
Then, in the right-angled triangle PHQ,
^2

^p2^q2*

^Q

^\iQ

lowest order.

*This follows from 65, where we have proved that the Euclidean
Formulae hold in Infinitesimal Geometry. If we start with
cosh -r- = cosh -^ cosh

we

obtain the same result

order.

when we

neglect terms above the lowest

66]

ELEMENT OF ARC

113

Also> in the quadrilateral MNHP, the angles at M, N, and

are right angles, and the sides beginning at M are
Sx, z, q,

These correspond to

ni,

a,

in a right-angled triangle.

y.

c,

[Cf. 35.]

Fio. 77.

smh

Thus we have

Sx
j-

sinh

[Cf.59,I.]
cosh-vk

Therefore

= cosh -|- Sx,

to the lowest order.

tC

Also,

we have
cosh

^ = tanh I coth j.

Therefore y and z

Put

diflFer

and

by a small quantity when Sx is

z

=y+

Then we have tanh ^^7^ cosh

N.-E.O.

[Cf. 59, V.,

r}.

-r =

tanh

58.]

small.

NON-EUCLIDEAN GEOMETRY

114

[ch.v.

This gives, to the lowest order,

,y=-2^sinh|-cosh|-6a;2.

i.e.

Therefore y and z differ by a quantity of the second order

is of the first order.

when Sx

Now

p={y + ^y)- ^
p = Sy, to the first

Therefore
It follows

from

8s^

Ss^

order.

=p^ + q"^, that

= cosh^ 1^ Sx^ + Sy-,

to the lowest order.

Thus we have shown that the element

is given by

of arc in Cartesian

Coordinates

ds2

= cosli2|-dx2 + dy2.
k

67.

Element of Arc

in Polar Coordinates.

In the Euclidean Plane we have for the element of arc in

Polar Coordinates, the equation

We proceed to find the corresponding formula in the

Hyperbolic Plane.
It may be obtained in two ways. It could be deduced from
ds^

cosh^ ^ dx^ + dyK

k

by using the

relations connecting

cosh

x,

y and

r,

\'\7.

== cosh
cosh y\
T
r
k
k

tanh|

[Cf. 63.]

A;

tan a

sinh|
.

POLAR COORDINATES

66, 67]

simpler and more

It is

instructive

to

obtain

directly.

Let

P,

be the points

(r, 0),

{r

+ Sr, d + SO).

Fio. 78.

Draw PN perpendicular to OQ.

Let PQ = ^>, PN=^, NQ=jo, and ON = z.
Then, from the triangle PNQ, we have as

Also, from the triangle

sinh

Therefore

ONP, we have

= sinh ^ sin 66.

= k sinh j- Sd,

we have from the same

cosh
J,

triangle

= cosh -r cosh

Therefore r and z are nearly equal.

Put

[ 63.]

to the lowest order.

fC

Also,

before

='2

+ f-

|.

115
the

result

NON-EUCLIDEAN GEOMETRY

116

Then

cosh y
k

+$k

sinh

^ = cosh r

k\

[CH. V.

2ky'

to the lowest order.

Thus
i.e.r

^=91; Goth-r, to the lowest order

and

when SO

z differ
is

by a small quantity

the second order,

of

of the first order.

p = r + 8r-z.
p = dr, to the

But
Therefore

first

order.

It follows that
Ss^

= Sr^ + k^ sinh^ y SO^,

to the lowest order.

Therefore

68. The
ordinates.

ds2

= dr2 + k2sinh2~d02.
k

Element of Arc in Limiting- Curve Co-

We shall now describe a system of coordinates peculiar to

the Hyperbolic Plane. The position of the point P is given
by the Limiting-Curve and axis on which it lies, the Limiting-

Fio. 79.

Curves being
infinity

all

concentric, their

on the axis of

common

centre being at

x.

Let the Limiting-Curve through P cut off a segment of

length ^(OPfl) on the axis of x, and let the axis through P

LIMITING-CURVE COORDINATES

67,68]

117

cut off an arc of length ;; (OA) on the Limiting-Curve through

O.
(Fig. 79.)
(^, r}) are called the Limiting-Curve Coordinates of the
point P.
Now take another point Q with coordinates

Let the Limiting-Curve through

axis through O) at Qo.

cut the axis of x (the

Let the Limiting-Curve through P be cut by the axis through

S, and the Limiting-Curve through Q by the axis through

Q at
P in

R.

Also, let

through

A and B be the points where the Limiting-Curve

cut by the axes through P and Q.

is

arcOA=;;,

0P =
It

follows

Curves

[ 55],

from

arc

OB = + Stj,
r]

0Q,==^+ Si.

^,

the properties of Concentric Limiting-

that
arc

QR = Sr}e

^'
.

_i
.'.

arcQR = (5;/e

^,

to the first order.

NON-EUCLIDEAN GEOMETRY

118

Further, PR

= 8^, and we

Now we know
PQ2
88"^

.'.

as usual.

that

PR2

RQ2, to the lowest order.

= S^^ + e

* Sif, to

ds^

= d^^ ^q

Therefore

PQ = Ss,

write

[ch.

the lowest order.

k

^^2

This result could also have been deduced from that of

66 by using the equations connecting {x, y) and (^, rj). [Cf.

57 and 69

(3).]

We

apply these formulae to find the perimeter of a

of portions of the Equidistant-Curve
and the Limiting-Curve.

69.

circle,

1.

and the lengths

The Perimeter of a

In

ds^

Circle of

Radius

= dr'^ + sinh2 y

a.

dQ"^,

we put

Thus the

arc from

a and dr = 0.
to 6 = 6 is given by

= k sinh ^ x6.
k

The Perimeter of the Circle

and is given by the expression
I

2.

27rk sinh r

ds^

we put

= 2x,

= cosh^ | dx^ + dy\

= b and dij = 0.
= to x = x is given by

i/

Thus the

arc from

a;

= xcosli-.
k

The Limiting-Ciirve.
of the Limiting-Curve through the origin,
centre at infinity on the axis of x, is

The equation
with

by putting

The Equidistant-Curve y = b.

In

3.

follows

its

e^=co8h|.

[Cf.

57(1).]

ELEMENT OF

68, 69, 70]

In

ds^=^cosh'^^dx'^

dx = tanh

we put
Then

fiJs2

It follows that s

origin.

119

+ df-,

dy.

= ('l+sinh2|^rfy!

ds = cosh

Thus

j^

AI

= k sinh

j_

dy.

when we measure

from the

If we compare this result with 57 (2), we see that the

length of the arc of the Limiting-Curve, such that the tangent
at one end is parallel to the axis through the other, is unity,

when ^=1.
70. The Element of Area.
Let the arc AB be an arc of a Limiting-Curve, centre
such that the tangent at B is parallel to the axis through A.

Then we know that the length

and 69

of the arc

AB

is k.

(3).]
1

Also,

if

and

1,

the length of the arc ^fi^

= ]ce

*;

AiA2=

1,

the length of the arc k.^B^

= ke

*,

AAj

if

so on.

fS

12,

57 (2)

NON-EUCLIDEAN GEOMETRY

120

Let the area

of

[ch. v.

ABB^Aj be denoted by A^.

1

Then

(cf.

48) the area of A^BjBgAg will be A^e

AgBgBgAg will be A^e

that of

Thus the area

of

*, etc.

ABBA
1L

n-l \

i
k A.

Therefore, as 7i->

J- e
/?

area approaches a limit, namely

oo, this

A=

This is the area of the region bounded by two axes of a

Limiting-Curve and an arc such that the tangent at one end
is parallel to the axis through the other end.
The unit of area has not yet been chosen in this discussion.
We now fix it so that the area denoted above by A will be
k^ the unit of area.

With

this

measurement

Also the area of ABA,iB,i will be

Next,
arc

let

k\l - e

^).

P be a point on A B, or A B produced, such that the

AP=s.
area APPjAj area ABBjAj =s

Then

and

area

:k,

APPAn=^sVl -e ^).

first, a rational number, and then treating the

number x as the limit of a sequence of rational
numbers, we find from the above that the area bounded by the

Taking

irrational

x,

LIMITING-CURVE COORDINATES

70]

arcs of two Concentric Limiting-Curves, distant

larger one being of length s, is equal to

121

x apart, the

Fio. 82.

From this result the expression for the element of area in

Limiting-Curve Coordinates will now be deduced.
Let P, Q, R and S be the points
a,

ri),

(i+Si,

rj

+ Sr,), (i+S^,

r,),

and

(^,

+ Sf})

[cf .

Fig. 80].

_i

Then

arc

PS = 87;e

*,

[68]

PR = ^^.

and
Therefore the area

PQRS

given by

is

-1/
kSrje

When

JJ\

*(l-e V-

S^, St] are small, this

becomes, to the lowest order,

i
e

^SiSri.

Therefore the element of area in LAmUing-Cv/rve Coordinates

is

i
e

* d^drj.

equal to the product of the two perpendicular chords

infinitesimal element, and with
these units the expression for the element of area is the same
as that in the Euclidean Plane.

This

is

PR and PS which bound the

NON-EUCLIDEAN GEOMETRY

122

[CH. V.

71. The Element of Area in Cartesian Coordinates.

This result can be obtained from the expression found in
70, by using the methods of the Calculup.

We

have

j7

= A;tanh^e^
[Cf.

and
e

'^

57
69

(3).]

=coshTk

These are the equations connecting

To find the element

we need only replace

(x, y) and (A j/).

of area in Cartesian Coordinates (x, y),

lco8h'4^^Jxdy,

by

k d{x, y)

After reduction,

we

obtain

cosh \ dx dy.

^^

Fio. 83.

The

result,

however, can be found directly as follows

Let P, Q be the points (?;, y), {x + &, y + <5//).

Let the Equidistant-Curves through P and Q with Qx as
base-line meet the ordinates at R and S (Fig. 83).

POLAR COORDINATES

71, 72]

123

The figure PRQS becomes a i-ectangle in the limit, and

can use the Euclidean expression for its area (cf. 70).

we

i^

Then

arc

PR = cosh | Sx
PS =

and
Hence

69

(2)]

Sy.

the element of area in Cartesian Com'dinates is

cosh ~ dx dy.

The Element of Area

72.

As

in Polar Coordinates.

before, the result can be obtained

cosh

= cosh

J-

by using the equations

cosh ^

tanh ^
k

tan0 =
sinh

which connect

(r,

6)

and

(x, y).

Fig. 84.

But

it

is

simpler to obtain the element of area directly

from the geometrical figure

Let

P,

be the points

(r, Q), (r

Sr,

+ SO).

NON-EUCLIDEAN GEOMETRY

124

Let the circles through P and

forming the element PRQS.

figure

Therefore

cut the radii at S and

= k sinh ^ ^0, by
PR = Sr.

Then we have

The

arc PS

PRQS becomes

[CH. V.

69

R,

),

a rectangle in the limit.

Polar Coo7-dinates is

the element of area in

ksinhidrdO.
k
The area

of the circle of radius a is thus given

k sinh

-J-

Jo Jo

which becomes
or

by

dr dO,

fc

27rA;2(coshT--

1 j,

4:7rk^smh.^ -^.

2k

73. The Area of a Triangle and of a Quadrilateral

with three Right Angles.

Fig. 85.

Let

OABC

be a quadrilateral with the sides a, m', c, I, as in

C right angles A lying on

Fig. 85, and the angles at O, A,

the axis of z and C on the axis of

y.

72,73]

THE AREA OF A TRIANGLE

125

Let P be any point on CB, and PM the perpendicular from

P to OA.
Then, from the associated right-angled triangle for the
quadrilateral OMPC, we have

tanh|cosh-r

But the area

of the quadrilateral

Denote

this

Integrating,

= coshT.

by

cosh

OABC

is

(59, V.)

given by

dx dy.

S.

we have

= k\
Jo

~ dx
sinh
si
k

COsh-r
'.dx

{oV'^
= A;-sin

sinh
..

8inTr,
k^

sinh^

sinh

sinh

-r-

sinh^

-7

-um

sinh

k
But, from the associated right-angled triangle,

tanhy =

^^"^
T-

1.

(59, IIL)

sinh-j-

And

tanh| = cos^.

(62.)

we have

126

NON-EUCLIDEAN GEOMETRY

Therefore

[ch.v. 73]

sin

Therefore

Thus the area of a quadrilateral with three right angles

acute angle /3 is equal to
on

and an

this scale.

triangle ABC (Fig. 86) is equal in area to Saccheri's

Quadrilateral BCC'B', in which the angles at B and C are each
equal to half the sum of the angles of the triangle.

But a

The triangle is thus equal in area to twice the quadrilateral

with three right angles, and an acute angle equal to J(A + B + C).
Using the result just found, the area of the triangle ABC
on

this scale of

measurement

is

F(7r-2;6),

where

2/3

= A + B + C.

In other words, the area of the triangle

and its defect.

Comparing this with
area was chosen in 70.

is

the product of

k"^

52,

we

see

why

the particular unit of

CHAPTER

VI.

THE ELLIPTIC PLANE GEOMETRY.

In Hubert's Parallel Postulate, through any point A
any line b, two parallels a^ and a^ can be drawn to the
line, and these separate the lines in the plane of the parallels
which cut 6 from the lines which do not cut it.
On the Euclidean Hypothesis, the two rays a^ and 02
together form one and the same line, and there is but one
parallel to any line from a point outside it.
There is still another case to be examined, namely that in
which all the rays through A cut the line 6. In this case there

74.
outside

no parallel through a point outside a line to that line.

shall see that this corresponds to the Hypothesis of the
Obtuse Angle of Saccheri, in accordance with which the sum of
the angles of a triangle exceeds two right angles. Saccheri and
Legendre were able to rule this case out as untrue
but their
argument depended upon the assumption that a straight line
is

We

was infinite in length. Riemaim was the first to recognise that

a system of geometry compatible with the Hypothesis of the
Obtuse Angle became possible when, for the hypothesis that
the straight line is infinite, was substituted the more general
one that

it is

endless or unbounded.

The geometry

built

(Cf. 19, 20.)

up on the assumption that a

straight

unbounded, but not infinite, and that no parallel can be

drawn to a straight line from a point outside it will now be
treated in the same manner in which the Hyperbolic Geometry
line is

was discussed.
75.

We

proceed to the development of Plane Geometry

when the assumptions

(i)
(ii)

All straight lines intersect each other,

The

straight line is not infinite,

128

NON-EUCLIDEAN GEOMETRY

take tlie place of the Parallel Hypothesis of Euclid

implicit assumption that the line is infinite.

Let A and B be any two points on a given

[ch.vi.

and

his

line L.

The perpendiculars at A and B to the line must intersect,

by assumption (i).
Let them meet at the point O.
Since ^OAB = z.OBA, we have OA = OB.
At O make ^BOQ = ^AOB (Fig. 87), and produce OQ to
cut the line L at P.

Then AB = BP and Z.OPA

is a right angle.
repeating this construction, we show that if P is a point
AB, the line OP
on AB produced through B, such that AP =
The same
is perpendicular to L and equal to OA and OB.
holds for points on AB produced through A, such that
is supposed to be a positive
BP = m.AB. In each case

By

integer.

Now, let be a point on AB, such that AB =m AC, m being

to L must pass
a positive integer. The perpendicular at
through the point O, since if it met OA at O' the above argument shows that O'B must be perpendicular to L and coincide
with OB.
It follows that if P is any point on the line L, such that
.

AP = .AB,
n

and n being any two

positive integers,

OP

is

perpendicular to the line L and equal to OA and OB.

The case when the ratio AP AB is incommensurable would
be deduced from the above by proceeding to the limit.
:

THE POLE OF A LINE

75,76]

129

Now, all points on the line are included in this argument,

80 that the perpendiculars at all points of the line L pass
through the same point.
Now, let L' be another line and A', B' two points upon it,
such that the segment

The perpendiculars

AB = A'B'.

at A', B'

meet

in

a point, which we shall

call O'.

Pio. 88.

The

triangles

AOB and

A'O'B' have a side of the one equal

two angles adjacent to the sides

to a side of the other, and the

are equal, each to each.

0'A'=OA.
Thus we have shown that the perpendiculars at all points on
any line meet at a point which is at a constant distance from
It follows that

the line.

The point

will

be called the Pole of the Line.

89, produce OA to 0^, where OiA=OA.

Join OjB.
Then, from the triangles OAB and OiAB, it follows that
^OiBA = z.OBA = a right angle.
Thus OB and O^B are in a stra ight lige.
Also, AOj produced must intersect AB at a point C, such that
O^C is perpendicular to AB, and OC will be also perpendicular
to AB.
Thus OAOj produced returns to O, and the line is endless or
unbounded.
Its length is four times the distance of the pole of the line
from the given line.

76. Now, in Fig.

N.-E.G.

NON-EUCLIDEAN GEOMETRY

130

[ch.vi.

We shall denote the constant distance OA by ^, so that

with this notation the length of the line is 4^.
Thus two other assumptions of the ordinary geometry are
contradicted in this geometry

Two
Two

straight lines enclose

a space

points do not always determine a straight line.

Through the two poles of a line an infinite number of lines

can be dxawn, just as through the two ends of a diameter of a
sphere an infinite number of great circles can be drawn.
It is now clear that the argument which Euclid employs in
The exterior angle of a
I. 16 is not valid in this geometry.
triangle is greater than either of the interior and opposite
angles only when the corresponding median is inferior to W.
If this median is equal to ^, the exterior angle is equal to
if it is greater than %, the
the interior angle considered
exterior angle is less than the interior angle considered.
;

16 was essential to the proof of I. 27, it is now

geometry that theorem does not hold. Of
course, if I. 16 did hold, there would have to be at least one
In a limited
parallel to a line through any point outside it.
Also, as

evident

I.

why

in this

76, 77]

THE TWO ELLIPTIC GEOMETRIES

131

region of the plane, I. 16 does hold, and theorems dependent

upon it are true in such a region.
The plane of this geometry has properties completely analogous to those possessed by the surface of a sphere. The great
circles of the sphere correspond to the straight lines of the
plane. Like the line, they are endless. Any two points on
the surface of the sphere determine a great circle, provided the
points are not the opposite ends of a diameter.
The great
circles through any point on the sphere intersect all other
great circles.
We shall find that this analogy can be carried further. The
sum of the angles of a spherical triangle is greater than two
right angles. The sum of the angles of a triangle in this plane
The Spherical Excess
is greater than two right angles.
measures the area of spherical triangles. With suitable units
Indeed
the area of plane triangles is equal to their excess.
the formulae of this Plane Trigonometry, as we shall show
later, are identical with the formulae of ordinary Spherical

Trigonometry.*

^c^*^

77. It must be remarked, however, that in the argument

of 76 it is assumed that the point 0^ is a different point from
p. If the two points coincide, the plane of this geometry has

The length of a straight line is

instead of 4^. If two points P, Q are given on the
plane, and any arbitrary straight line, we can pass from P to. (1
Q by a path which does not leave the plane, and does not cutn /
the line. In other words, the plane is not divided by its lines]
,
*
into two parts.
The essential difference between the two planes is that in
the one the plane has the character of a two-sided surface, and
in the other it has the character of a one-sided surface .f The
that which we have been examining is usually
first plane
called the spherical plane (or double elliptic plane)
the second
plane is usually called the elliptic (or single elliptic) plane.
The geometries which can be developed on both of these
planes are referred to as Riemann's (Non-Euclidean) Geometries.
It seems probable that the Spherical Plane was the only

a wholly different character.

now 2^

Spherical Geometry can be built up independently of the Parallel

Postulate, so it is not necessary to say ordinary Spherical Trigonometry
when referring to it.

tCf. Bonola,

loc. cit.

75.

NON-EUCLIDEAN GEOMETRY

132

[CH. VI.

form in his mind. The Single Elliptic Plane and its importance
in the higher treatment of the Non-Euclidean Geometries were
first brought to light by Klein.

We

shall now show that this geometry corresponds to

78.
Saccheri's Hypothesis of the Obtuse Angle, so that the sum of
the angles of a triangle is always greater than two right angles.
The following theorem enables us to put the proof concisely
:

In any triangle ABC in which ike angle C is a

angle A is less than, equal to, or greater than a

1.

the

according as the segment

BC is less than,

right angle,
right angle,

equal to, or greater than |.

Let P be the pole of the side AC.

Then P

lies

upon BC, and PC = ^.

Join AP.

Then z.PAC = a

CB > CP,
If CB = CP,
If CB < CP,
If

right angle.

then l BAC > l PAC

then l BAC =^ PAC
then L BAC < l PAC

The converse

i.e.

i.e.

l BAC > a right angle.

l BAC = a right angle.

i.e.

L BAC < a right angle.

also holds.

Now consider any right-angled triangle ABC in which C is

the right angle.
If either of the sides AC or BC is greater than or equal to
^, the sum of the angles is greater than two right angles by
the above theorem.
If both sides are less than ^, from D, the middle point of
the hypothenuse, draw DE perpendicular to the side BC.
Let P be the pole of DE.

SUM OF THE ANGLES OF A TRIANGLE

77, 78]

Produce ED to

F,

so that

Join AF and PF.

Then the triangles
lie

ED =

133

DF.

ADF and DEB

are congruent,

and AF, F^-

in one straight line.

But we know that z.PAC>a

than

right angle, since

CP

is

greater

'g.

Therefore the sum of the angles at A and B in the rightangled triangle ACB is greater than a right angle in this case
as well as in the others.

Thus we have proved that

2.
i

In any

right-angled triangle the

sum

of the angles is greater

than two right angles.

Finally, let

ABC

be any triangle in

which none of the angles are right angles.

We need only consider the case when
two of the angles are acute.
Let Z.ABC and Z.ACB be acute.
From A draw AD perpendicular to
BC D must lie on the segment BC.
Then, from (2),
;

L ABD +

and

.'_

BAD > a

right angle

z.DAC + ^ACD>a right angle.

Fio. 92.

It follows that the sum of the angles of the triangle

greater than two right angles.

ABC

is

NON-EUCLIDEAN GEOMETRY

134

[CH. VI.

Thus we have proved that

3.

The sum of

the angles of

any

triangle is greater than two

right angles.

The amount by which the sum of the angles of a triangle

exceeds two right angles

is

called its Excess.

79. Saccheri's Quadrilateral, and the Quadrilateral

with three Right Angles and one Obtuse Angle.

Let AC and BD be equal perpendiculars to the segment AB.
The quadrilateral ABDC we have called Saccheri's Quadrilateral.

F be the middle points of AB and CD.

and
that EF is perpendicular to both AB and CD
that the angles ACD and BDC are equal.
But the sum of the angles of a quadrilateral must be greater
than four right angles, since it is made up of two triangles.
It follows that the angles at C and D are obtuse.

Let

E,

We know

Fio. 94.

Thus the Elliptic Geometry corresponds to Saccheri's Hypothesis

of the Obtuse Angle.

Now let ABDC

(Fig. 94)

be a quadrilateral in which the angles

and D are right angles.

The angle at C must be obtuse by 78.

at A, B,

Each of

To prove this, we proceed

If AC is not less than BD,
or equal to

as follows
it

must be

either greater

than

it

it.

z.ACD=/LBDC, which
if AC = BD,
obtuse and the other a right angle.
cut ofi AE=BD, and join ED.

But we know that

impossible, as one
If

a quadrithan the side opposite to it.

the two sides containing the obtuse angle in

lateral with three right angles is less

AOBD,

is

is

SACCHERI'S QUADRILATERAL

78,79]

Then we know that

But :lEDB

is

l NED

135

=^ EDB.

acute, so that both

must be

acute, which is

impossible.

Therefore AC must be less than BD.

Again, starting with AB and CD, which are both perpendicular
to BD, we find that CD is less than AB, so our theorem is proved.

We

proceed further with the formal development

geometry. There is no Theory of Parallels, for
parallel lines do not exist in it.
There is only one kind of
circle, the locus of corresponding points upon a pencil of
straight lines. The measurement of areas follows on the same
lines as in the Hyperbolic Geometry.
of

shall not

this

Two
and

triangles

which have

the

same excess have equal

conversely.

The area of a

triangle is proportional to its excess.

areas,

[CH. VII.

CHAPTER

VII.

THE ELLIPTIC PLANE TRIGONOMETRY.

80. The following treatment of the Elliptic Trigonometry
due to Gerard and Mansion. Gerard discussed the Hyperbolic Trigonometry on these lines.*
Mansion showed that
the method discovered by Gerard was applicable also to the

is

Elliptic case.f

The notation to be employed has first to be explained.

Let OA and OA' be two lines meeting at O at right angles.
Let OL be a third line making an acute angle with OA and OA'.
Let P be any point upon the line OL, such that 0P<^.

Let

We

PM and PM' be the perpendiculars to

denote OM, MP, and OP by x, y, and r

M'P by

y'

and

OA and
;

OA'.

and OM' and

x'.

* Gerard, Sur la G^om4trie won euclidienne (Paris, 1892).

Cf. also
Young, " On the Analytical Basis of Non-Euclidian Geometry," Amer.
Journ. of Math., vol. xxxiii. p. 249 (1911) and Coolidge, Non-Euclidean
;

Geometry, ch.

iv.

(Oxford, 1909).

t Mansion, Principes Fondamentarix de la 64om,6trie non euclidienne

de Riemann (Paris, 1895).

METHOD OF GERARD AND MANSION

80, 81]

81.

I.

//" P,

are

any two points on OL,

137

sicch that

OP<OQ<^,
and Pp, Qq are perpendicular

We know

that

to

OA, then LOPp<LOQ,q.

LpPQ. + LPQ.q>2 right angles.

Also

L0Pp + ^pPQ. = 2

Therefore

z.

right angles.

OPp < l PQ.q.

If S is the point on OL, such that 0S =

and Ss is perpendicular to OA, we know that AOSs = a right angle.
It follows that LQPp<LOQ/i<z. right angle.

II.

"From

Let P and

J continually increases.
be any two points upon OL, such that

to S,

OP<OQ<|C.
Pp = Qq, we must

Then we know that if

is impossible by (I.).

have

LpPQ =

i.

PQq,

which

Again,

Then
But

if

ofi pP' = qQ and join P'Q.

LpP'Q, = LP'Qq.

Pp>Q,q, cut

LPQlq<a, right angle.

(Fig. 97.)

NON-EUCLIDEAN GEOMETRY

138

[CH. VII.

Therefore l jaP'Q and l P'Qq are equal acute angles, whicli

is

impossible.

Fio. 97

Thus, as the point P [moves along

OL from O towards

S,

continually increases.
III.

From O

to S, the ratio

First, consider points

nients on OA.

- continually

increases.

upon OL corresponding

Fig. 98.

Let

P,

Q, R be three such points, so that

jyq

= qr.

to equal seg-

METHOD OF Gl^RARD AND MANSION

81]

139

Then we know that pP<qQi<rR.

From rR cut off rP' = pP, and join QP'.
Then we have PQ = QP' and -Q,Pp = i.QlP'r.

z.QRP'>^QP'R and QR<^QP'.

Therefore

Thus,

if

J>q

= i^> PQ>QR.

Therefore, for equal increments of x,

increments of
It follows

from

OL, such that

we have decreasing

r.

this that

if

P and

are any two points upon

are commensurable,

OP<OQ<^, and OM, ON

OM ON
OP

'^

OQ'

When OM and ON are incommensurable,

conclusion by proceeding to the limit.
Thus, from

IV.

From O

we reach the same

to S, the ratio - continually increases.

to S, the ratio -^ decreases.

First

that

we

consider points

upon OL

at equal distances along

line.

Let

P,

Q, and R be three such points, so that

PQ = QR.
From P and

R draw PH and RK perpendicular to Qq.

NON-EUCLIDEAN GEOMETRY

140

Then we know ttat Pj)< Hq and Rr < Kq (

But QH = QK. Therefore Q,q-Hq = Kq- Qq.

[ch.vti.

79).

from the above that Qq-Pp>Rr- Q,q.

Therefore, for equal increments of r we have diminishing

It follows

increments of
It follows

y.

from

OL, such that

this that if

P and

are

any two points upon

OP < OQ < ^, and OP, OQ

Pp
Qq
OP OQ'

are commensurable,

When OP and OQ

are incommensurable, we obtain the same

by proceeding to the limit.
Thus, as P moves along OL from O towards S, the ratio

result

^ continually
V.
limit

from

When
from

decreases.

r tends to zero, the ratio

and

above,

the ratio

r tends

towards a finite

r tends towards

finite limit

below.

From (III.) we know that x r continually decreases as r

tends to zero, so that this ratio has a limit, finite or zero.
From (IV.) we know that y r continually increases as r
tends to zero, so that this ratio either has a finite limit, not
:

zero, or

becomes

But from the

have X

> x'.

infinite.

whose sides are

Thus x:r >x':r.

quadrilateral

(Fig. 95.)

But, by (IV.),

{x, y, x', y')

we

either has a finite limit, not zero, or

tends to zero.
Therefore the limit of a; r cannot be zero, and must be some
Also x r approaches this limit from above.
finite number.
But it follows from the preceding argument that y' r has
a finite limit, not zero.
Also we know that y < y', and thus y r < y' r.
It follows that y r has a finite limit, not zero, and it
approaches this from below.

becomes

x': r

infinite, as r

These two limits Lt(^), Lt(-| are chosen as the sine

of the acute angle which OL makes with OA,* and
the other ratios follow in the usual way.

and cosine

* These limits are functions of the angle.

It can be shown that they
are continuous, and that with a proper unit of angle they are given
by the usual exponential expressions. Cf Coolidge, loc. cit. p. 53.
.

81, 82. 83]

METHOD OF Gl^RARD AND MANSION

We

82.

turn

now

and one obtuse

angles

141

to the quadrilateral with three right

angle.

Let 0AB6 be such a quadrilateral, the angles at O, B, and b

being right angles.
Produce 06, and cut off bc = Ob and cd = bc.
Draw the perpendiculars to 06 produced at the points c and
d and from A the perpendiculars to the lines just drawn.
;

Fig. 100.

We

thus obtain three quadrilaterals 0AB6, OACc, OADd, of

on the bases 06, Oc, and Od.
It is easy to show that the obtuse angles of these quadrilaterals increase as the bases increase.
Let 6B produced meet AC at H, AB produced meet Cc at I,
this nature, standing

and AC produced meet DtZ at J.

Then we have AB = Bl, AB < AH, and Al > AC.
It follows that AB > AC - AB.
Also we have HC = CJ and AD < AJ.
Therefore AC - AH = AJ - AC, and finally AC - AB
Thus AB > AC - AB > AD - AC.
83.

OMPM',
and the
and

> AD - AC.

We

return to the notation of 80 and the figure

which the angles at O, M, and M' are right angles,
sides OM, MP, PM', and'OM' are denoted by x, y, x',
in

y' respectively.

We
In

shall

now prove

the following theorem

the quadrilateral with three right angles (x, y, x', y'), in

NON-EUCLIDEAN GEOMETRY

142

which the sides

include the obtuse angle, if y' is Jcept fixed

x tends to a finite limit 0(y')

x',

and X tends to zero,

from above, and this

[CH. VII.

the ratio x':

ratio is less than <p{j).

X
Fio. 101.

As in 81 we find that x' x continually decreases as x

tends to zero. It must have a limit, which may be zero or
:

some number less than unity.

Produce MP, and draw M'Q perpendicular

From

82 we

know

to MP.

that as x decreases, the ratio

increases.

Pio. 102.

It

must therefore have a

finite limit,

infinite.

^
But
.

M'Q

- < M'P
x
X

Thus Lt(

smce

..,_

M Q <^ ..,
M P.

cannot be

zero.

not zero, or become

THE FUNCTION

83, 84]

This function

by

y',

and

Now we

is

associated with the segment M'O, denoted

be written as

will

(p{y').

have seen that

(?/')>

But

M'Q.x.

OM'PM, the side PM plays the same

the quadrilateral OM'QM.

in the quadrilateral

part as

OM'

in

Thus

<p(y)> M'P:x.

Thus we have

t/'

143

<t>{y)

0(y)<-<0(y)-.

Since x' <x, the function 0(t/') is less than unity, except for
it becomes equal to unity.

= 0, when

84.

We

Let OS

now show

that the function defined in the

continuous.
and Os be two lines meeting at O, such that
shall

previous section

is

OS = Os = |C and

i.SOs

is

acute.

"^

\,

/"
d

Xt

'

C
Fm.

103.

Then the angles at S and s are both right angles.

Let SB = aj y, SC=a;, and SD=x + y.
Let the perpendiculars at B, C, and D to OS, meet 05 at
h,

c, and d.
Through 6 and d draw bm and dn perpendicular to

Cc.

NON-EUCLIDEAN GEOMETRY

144

[ch.vii.

From 81 (III.), applied to the acute angles den and

we have cb < cd, cm :cb<cC cO, and cn:cd<cC cO.
From the second of these relations we have

bcm,

Cm

Cc

~Ss~ Ss

.BOX
Bb\

/Bb
'Bh
I.e.

Cc

cb

Ss cO

Cc

Cc

\Ss'Cm)~Ss

Then, by 83,

if

Ss,

,.

cb

Ss cO

and thus Bb and

Cm

tend to zero,

we have

^^

Ss

LtcJ = CB = y

Further,

Therefore, from (a),

and

'

LtcO = CO = ^-a;.

we have

^(^x-y)-ct>{x)<i>{y)^^^<l>{x)<t>{y).

i.e.

Again, from the inequality

(/3)

we have in the same way

^<
C\J
^,

Cu/

^(x)(f>(y)-^(x + y)^^-^cl>(x)^{y)

Adding

(^8)

and

(y),

(y)

we have

^(z-y)-(p(x + y)^-~^<p{x)<p(y)<^^,
since

(^{y) are each less than unity.

continuous function of x.

(j){x),

<p{z) is a

It follows that

A FUNCTIONAL EQUATION

84,85.86]

We

85.

shall

now show

</.(x

With the

145

that

+ y) + 0(x - y) = 20(x)0(y).

figure of 84, let the perpendicular at c to Cc

in p and q.
From cd cut off cr = cb, and

meet Dd and Bb
join

jpr.

Then we have cp = cq and pr = qb.

We

shall presently

suppose Ss to become infinitesimal.

this case the angles at

angles,

p and

and Ldpr becomes

In

q differ infinitesimally from right

infinitesimal.

compared with pd *
an infinitesimal of the first order, dr is at

It follows that dr is infinitesimal as

and that

if

least of the

Ss

is

second order.

But dp-qb = dp-pr< dr.

And dp-qb = {Dp - Dd) - {Bb Therefore we have

Bq).

/ D/?
Cc Dd Bb
\Cc' Ss~ Ss~ S^

Bq Cc\ _
Cc' Ss)~

Lt(B^) = ,(,) = LtQ).

Lt(|) = <^(x),

And

Lt(g) = 0(x +

3/),

uQ-^ = ^(x-y).

and

Thus we have
</>{x

86.

We

+ y)-i<p(x-y) =

2<f>(x)<l>(y).

proceed to the equation

<t>{x

+ y) + (p(x -y) = 2<l){x)(p{y).

We are given that (^{x) is a continuous function, which is

equal to unity when x = 0, and when x>0, <p{x) <1.
Let Xi be a value of x in the interval to which the equation
applies.

Then we can

find k, so that

*C{. Coolidge, he.

cit.

p. 49.

<f>{x^)

= coa-r-

NON-EUCLIDEAN GEOMETRY

146

Tchvii.

The function cos x has here a purely analj^ical meaning,

being defined by the equation
cosa;=l-2-j
It follows that

(l^i)

+ ^,

= cos -=-J

0(??-.rj)

= cos-^,

nx,\

/nx

Now let X be any other value of x in the

happens that this value is included in the set

interval.

If

it

nx-.

or -i,

nx.

we know that
But

if it is

0(.'r;)

= cos
(

,-),

by the above.

not included in these forms, we can still find

n by going on far enough in the scale, such

positive integers m,

that

nx.\

where

e is

But

any

(b{x)

positive

and cos

y-'

number

as small as

we

please.

are continuous functions.

k
x

It follows that

(x)

= cos r

This value of k will be related to the measure of the line OS,

in the previous sections.

denoted by

We have now to

deal with a rather complicated figure.

obtain the fundamental equation of this
Trigonometry for the Right-Angled Triangle ABC, in which C
is the right angle, viz.

87.

From

it

we

shall

cos r

Let

ABC

/ix

= cos .- cos r

\^)

be a right-angled triangle, in which C

is

the right

angle.

From
to AC.

a point b upon

AB produced draw

be perpendicular

THE RIGHT-ANGLED TRIANGLE

86, 87]

Move

the triangle 6cA along AC

up the position 6'C.
thus have the triangle h'a'C

till it

coincides with

147

C and

he takes

We

congruent with 6Ac.

In the same way move the
triangle hck along BA until h
coincides with B and the triangle
takes up the position Ba"c".
Through the middle point of
a'A draw L perpendicular to BA.
Then LI produced will be perpendicular to h'a'
We thus obtain the common
perpendicular to h'a' and BA,
the line KIL.
In the same way we obtain
the common perpendicular MJN
to AC and a"c" through the
middle point J of Ao".
Finally, we draw ?>'Q perpendicular to AB and hh" perpendicular to BC.
We have seen that as Bh tends to zero,
I

Fio. 104.

we have

Lt*^' = Lt*|
B6
bB
In the

MJ

Lt - = Lt
JA

same way

= C^,

Ave

(i)

by

(ii)

and remembering that Aa" = Bb and

bir_

may

MN

Lt^

Lt^'

be written
T ^ b'Q.

shall

(ii)

have

Lt

We

IA"

Aa

Aa

Dividing

which

IL

Lt^, = Lt^.

Thi

Aa'

(i)

now show

bb"

T ^ B^'

that this equation

0(AB) = 0(BC)0(CA).

is

.(iii)

the same as

NON-EUCLIDEAN GEOMETRY

148

From

83,

[ch.vii.

we have
0(LQ)<|^<</,(K6').

Now, when B6 tends to zero, LQ and Kfe' tend

84, 0(LQ) and (p{^b') tend to (/)(AB).

to BA,

and

from

Lt^ = 0(AB)

Thus,

In the same

(iv)

way we have
Lt(^') = 0(BC)

There remains the limit of

Let

(V)

MN
b6'

be the point at which Be" meets AC.

from 81 (I.), that s lies between C and

We know

c,

and

we have Bs>BC.
Then, since Cb'

= Bc", we have

Bh'>sc".

>0(Nc').
>
|^
MN Mn ^^

Therefore

'

Produce BC till it meets a"c" in R.

We have BR>Bc", so that BR>C6'.

From BR cut off Bc' = C6'.

Draw c'P perpendicular to MN.
Then we have

- PM - Cc').
=
v^
w < 0(CM
MN < MP S^
MP < 0(PO

I^kF

Thus

''

^(CM-PM

BJ'

-Cc')>^^TT>0(Nc")

MN

'

Proceeding to the limit,

<">

^(*^>=i^Ki;)
From (iii)-(vi), it follows that 0(AB)
the usual notation from 86,
c
COSi-

= 0(BC)<^(CA),

= COSeCOSr.

or with

TRIGONOMETRICAL FORMULAE

87,881

149

Note. At several points in this argument we have assumed

that the segments concerned are less than ^.
Once the fundamental theorem has been proved for triangles
in which this condition is satisfied, it can be extended by
analysis to all other cases.

The remaining formulae

88.

To prove
Let

tan ^

ABC be any

are easily obtained

= cos A tan ^

(2)

right-angled triangle, with

Take any point D on AC, and

C a

right angle.

join BD.

Draw DE perpendicular to AB.

Let AE=p, ED=q, AD=r, and BD=Z.
Then, from the triangle ABC, we have
cos r cos

cos

(^)

'

a
cos T

fC

fC

iC

= cos y cos J cos J + cos J sm j sin yr

= cos T cos T +
Also,

sm r sin tfC

fC

from the triangle BDE, we have in the same way

I

cos T
rri

1 here! ore

a
cos j
* k

= cos

r
T cos r

p
c
q
r Sin j: sin v.
.

+ cos

sm y sin ^ = cos .qf sm p

y- sm y
.

NON-EUCLIDEAN GEOMETRY

150
TT

cab

cos r

= cos t cos t,

= cos =7P cos Qy,

Using the equations

cos 7
k

this gives

tan

tan T
k

[ch. vii.

tan T
k

tan r

may

This result holds however small r

But we have seen that when r->0,

other than zero, and that this limit
the angle.
(81.)

is

be.

has a definite limit

~
r

taken as the cosine of

tan|
Therefore

cos A

= Lt
'-^"tan''
k
b

tan
tan J

89.

To prove that

sm-a
.

sin

]j^

A=

(3)

sin-

We

have seen that as r^O, the ratio

"

tends to a definite

limit, other

than zero, and that this limit

is

taken as the sine

of the angle.

Now

from the equation

c

COS|;

we

find that

when

a, b,

and

c^ = a'^ + b\

It follows that

= COS

sin'''

a
b
rCOST,

c are small,

to the lowest order.

A + cos^A =1.

88, 89]

TRIGONOMETRICAL FORMULAE

But, from 88,

we have

151

tatir

cos

A=
tariv
A;

tan^

Therefore

sin^ A

k
sin^ T

- tair t

- sec"^
T

CDS'*

cos^ 7

oC

COS''

8in^

,A:

Therefore

sin

A=

sin 7
^

The remaining formulae,

COS

A = cos

sm

B,

(4;

sin T

= tan -r cot A,

(5)

cos T

= cot A cot B,

(6)

can be easily deduced from those already obtained.

NON-EUCLIDEAN GEOMETRY

152

The

six equations

(l)-(6) are the equations of

Spherical Trigonometry,
a, b,

[ch. vn. 89, 90]

when

^,

c.

j,

and

-^

ordinary

are substituted for

8 90. The Trigonometry of the Oblique-Angled Triangle

follows from that of the Right-Angled Triangle, the definitions
of the sine and cosine being extended to obtuse angles. The
formulae will be identical with those of ordinary Spherical
Trigonometry, with the parameter k introduced.
The elements of arc and area can also be deduced as in
Chapter V. In this case we shall have
ds^

= cos^ J dx^ + dif-,

ds^^dr^

+ k'^sin^jdO'^,
k

d^ = cos

dA = k sin

dx dy,
T
v dr d0.

Also the Euclidean Formulae hold true in Infinitesimal

Elliptic Plane.

Geometry on the

CHAPTER

VIII.

THE CONSISTENCY OF THE NON-EUCLIDEAN GEOMETRIES

AND THE IMPOSSIBILITY OF PROVING THE PARALLEL
POSTULATE.
91. As we have already seen, the discovery of the NonEuclidean Geometries arose from the attempts to prove
Bolyai and Lobatschewsky did
'Euclid's Parallel Postulate.
a double service to Geometry. They showed why these
attempts had failed, and why they must always fail for they
succeeded in building up a geometry as logical and consistent
as the Euclidean Geometry, upon the same foundations, except
that for the Parallel Postulate of Euclid, another incompatible
with it was substituted. They differed from almost all their
predecessors in their belief that, proceeding on these lines, they
and they held that the
would not meet any contradiction
system of geometry built upon their Parallel Postulate was a
fit subject of study for its own sake.
How can one be certain
The question naturally arises
that these Non-Euclidean Geometries are logical and conHow can we be sure that continued study
sistent systems ?
would not after all reveal some contradiction, some inconsistency ?
Saccheri thought he had found such in the Hyperbut he was mistaken. Even Bolyai, many
bolic Geometry
years after the publication of the Appendix, was for a time of
the opinion that he had come upon a contradiction, and that
the sought-for proof of the Euclidean Hypothesis was in his
hands. He, too, was mistaken.
Of course, it is not sufficient simply to point to the fact that
these geometries developed into a large body of doctrine as
they have been do not offer in any of their propositions the
contradiction which the earlier workers in those fields were
convinced they must contain. We must be sure that, proceeding further on these lines, such contradiction could never be
j


NON-EUCLIDEAN GEOMETRY

154

[ch. vni.

discovered. If we can prove this to be the case, then we know

that Euclid's Parallel Postulate cannot be demonstrated.

92. There are several ways by which it is possible to

establish the fact that the Hyperbolic and Elliptic Geometries
are as logical and consistent as the Euclidean Geometry.*
Lobatschewsky, and to some extent Bolyai, relied upon
the formulae of the Hyperbolic Plane Trigonometry. These
are identical with the formulae of Spherical Trigonometry,
If the ordinary
if the radius of the sphere is imaginary.
Spherical Trigonometry offers no contradiction, their geometry could not do so. However, this proof is not complete
in itself, for it leaves aside the domain of Solid Geometry, and
does not establish the impossibility of the difficulty appearing
(Cf. Chapter II. 15, 17.)
in that field.
The most important of all the proofs of the consistency of
the Non-Euclidean Geometries is that due to Cayley and
In it one passes beyond the elementary regions within
Klein.
the confines of which this book is meant to remain. Other
proofs are analytical. The assumptions of geometry are
translated into the domain of number. Any inconsistency
would then appear in the arithmetical form of the assumptions
This form of proof also seems
or in the deductions from them.
to lie outside the province of this book.
Finally, there are a number of geometrical proofs, depending
upon concrete interpretations of the Non-Euclidean Geometries in the Euclidean. The earliest of these due to
Beltrami, and dealing with the Hyperbolic Geometry
requires a knowledge of the Geometry of Surfaces. But an
elementary representation of the Hyperbolic Plane and Space
in the Euclidean was given by Poincare.
" Let us consider," he says, " a certain plane, which I shall
call the fundamental plane, and let us construct a kind of
dictionary by making a double series of terms written in two
columns, and corresponding each to each, just as in ordinary
dictionaries the words in two languages which have the same

signification correspond to one another

Space.

of space situated above

the fundamental plane.

The portion

* For a discussion on more advanced lines, cf. Sommerville's NonEuclidean Oeometry, oh. v. and vi. (London, 1914).

REPRESENTATIONS

POINCARfi'S

91,92,93]

155

orthogonally
cutting
Sphere
fundamental plane.

Circle cutting orthogonally the funda-

Sphere.

Sphere.

Circle.

Circle.

Angle.

Angle.

Plane.
Line.

the

mental plane.

The logarithm

the anharmonic
two points and of
the intersections of the fundamental

Distance between

two

of

ratio of these

points.

circle passing through

these points and cutting it orthogonally.

plane with the

Etc.

Etc.

" Let us take Lobatschewsky's theorems and translate them

aid of this dictionary, as we would translate a German
text with the aid of a German-French dictionary.
We shall

by the

then obtain the theorems of ordinary geometry.

For instance,
Lobatschewsky's theorem
The sum of the angles of a
triangle is less than two right angles may be translated thus
'
If a curvilinear triangle has for its sides arcs of circles which
cut orthogonally the fundamental plane, the sum of the angles
of this curvilinear triangle will be less than two right angles.'
Thus, however far the consequences of Lobatschewsky's
hypotheses are carried, they will never lead to a contradiction ;
in fact, if two of Lobatschewsky's theorems were contradictory, the translation of these two theorems made by the
aid of our dictionary would be contradictory also. But
these translations are theorems of ordinary geometry, and no
one doubts that ordinary geometry is exempt from contradiction." *
'

'

93. To Poincare is also due another representation of the

Hyperbolic Geometry, which includes that given in the preceding section as a special case. We shall discuss this representation at some length, as also a corresponding one for the
Elliptic Geometry, since from these we can obtain in a simple
and elementary manner the proof of the impossibility of
* Poincard*,

La

Science

street, p. 41 et seq.

et

I'

HypolMse.

English translation by Green-


NON-EUCLIDEAN GEOMETRY

156

[ch

vm.

proving the Parallel Postulate and of tlie logical consistency

both of the Hyperbolic and Elliptic Geometries. In this
discussion the " dictionary method " of "29vwill be more fully

^^

explained.

We

shall consider three families of circles in a plane

extending the argument to spheres later. These are the

family of circles passing through a fixed point
the family
of circles cutting a fixed circle orthogonally
and the- family
of circles cutting a fixed circle diametrally {i.e. the common
chord of the fixed circle and any of the variable circles is to
be a diameter of the fixed circle). Denoting the fixed point
by O, and taking the fixed circle as a circle with centre O and
radius k, the first family of circles has power zero with regard
to O
the second, power k^
and the third, poiver P. We
shall see that the geometries of these three families of circles
agree with the Euclidean, Hyperbolic, and Elliptic Geometries,
;

respectively.

The System of Circles through a Fixed Point.

we invert from a point O the lines lying in a plane through
O we obtain a set of circles passing through that point. To
94.

If

circle there corresponds a straight line, and to every

straight line a circle.
The circles intersect at the same angles
as the corresponding lines. The properties of the family of
circles could be deduced from the properties of the set of lines,
and every proposition concerning points and lines in the one
system could be interpreted as a proposition concerning points
and circles in the other.
There is another method of dealing with the geometry of
this family of circles.
We shall describe it briefly, as it will
make the argument in the case of the other families, which

every

represent the Non-Euclidean Geometries, easier.

If two points A and B are given, these, with the point O,
fully determine a circle passing through the point O.
We
shall call these circles nominal lines.*
We shall refer to the
points in the plane of the circles as nominal points, the point O
being supposed excluded from the domain of the nominal
* In another place,

cf. Bonola, loc. cit., English translation, Appendix

and Proc. Edin. Math. Soc, Vol. 28, p. 95 (1910), I have used the
terms ideal points, ideal lines, etc. For these I now substitute nominal
points, nominal lines, etc., owing to possible confusion with the ideal

v.,

points, ideal lines, etc., of 37, 38.

93. 94]

NOMINAL POINTS,

LINES,

PARALLELS

157

We define the angle between two nominal lines as

the angle between the circles with which the nominal lines
coincide at their common point.
With these definitions, two different nominal points A, B
in this Nominal Geometry always determine a nominal line
AB, just as two different ordinary points always determine
a straight line AB.
The nominal points and lines also satisfy the " axioms of
order," * which express the idea of between-ness, when the
point O is excluded from the domain of the nominal points.
If this point were not excluded, we could not say that of any
three nominal points on a nominal line, there is always one,
and only one, which lies between the other two.
Proceeding to the question of parallels, we define parallel
nominal lines as follows
points.

The nominal line through a nominal point parallel to a given

nominal line is the circle of the system which passes through
the given point and touches at O the circle coinciding with the
given nominal line.
Referring to Fig. 106 we see that in the pencil of nominal
lines through A there is one nominal line which does not cut
BC, namely, the circle of the system which touches OBC at O.

This nominal line does not cut the nominal line BC, for the
point O is excluded from the domain of the nominal points.
It is at right angles to AM, the nominal line through A perpendicular to the nominal line BC. Every nominal line through
* Cf. Hilbert, loc. cit. 3.

NON-EUCLIDEAN GEOMETRY

158

A making with

BC on the

AM an
0AM

side of

[ch.vih.

angle less tlian a right angle will cut

which the acute angle lies.

in

Therefore in the geometry of these nominal points and lines

Euclidean Parallel Postulate holds.

the

95. Before we can deal with the metrical properties of this

geometry, we require a measure of length. We define the
nominal length of a nominal segment as the length of the rectilinear
segment to which it corresponds.
From this definition it is not difficult to show that the
nominal length of a nominal segment is unaltered hy inversion
with regard to a circle of the system ; and that inversion with
regard to such a circle is equivalent to reflection of the nominal
points and lines in the nominal line which coincides with the
circle

of inversion.
if we invert successively with regard to two circles of

Now,

the system {i.e. if we reflect in two nominal lines one after

the other), we obtain what corresponds to a displacement in
two dimensions. A nominal triangle ABC takes up the position
A'B'C after the first reflection and from A'B'C' it passes to
the position A"B"C" in the second. The sides and angles
of A"B"C" (in our nominal measurement) are the same as the
sides and angles of the nominal triangle ABC, and the point
C" lies on the same side of A"B" as the point C does of AB.
Further, we can always fix upon two inversions which will
change a given nominal segment AB into a new position such
that A comes to A', and AB lies along a given nominal line
through A'. We need only invert first with regard to the
circle which " bisects " the nominal line A A' at right angles.
This brings AB into a position A'B", say. Then, if we invert
with regard to the circle of the system which bisects the angle
between A'B" and the given nominal line through A', the
segment AB is brought into the required position.
The method of superposition is thus available in the geometry
of the nominal points and lines. Euclid's argument can be
" translated " directly into the new geometry. We have
only to use the words nominal points, nominal lines, nominal
parallels, etc., instead of the ordinary points, lines, parallels,
etc., and we obtain from the ordinary geometry the corresponding propositions in the geometry of this family of circles.
It should perhaps be pointed out that the nominal circle
with centre A is an ordinary circle. For the orthogonal
;

EXTENSION TO SOLID GEOMETRY

94, 95. 96]

159

system through A (i.e. of the

nominal lines through A) is the family of coaxal circles with
O and A as Limiting Points. The nominal lengths of the
nominal segments from A to the points where one of these
circles cuts the pencil of lines will be the same.
trajectories of the circles of the

96. The argument sketched in the preceding sections can

be extended to Solid Geometry. Instead of the system of
circles lying in one plane and all passing through the point O,
we have now to deal with the system of spheres all passing
through the point O,
The nominal point is the same as the ordinary point, but the
point O is excluded from the domain of the nominal points.

The nominal
passing through

line through two

O and

nominal points

is the circle

these tivo points.

The nominal plane through three nominal points is the sphere

O and these three points.
The nominal line through a point A parallel to a nominal

passing through
line

BC

is the circle

through

A which

lies

on the sphere through

and touches the circle OBC at the point O.

It is clear that a nominal line is determined by two different
nominal points, just as a straight line is determined by two
different ordinary points.
The nominal plane is determined
by three different nominal points, not on a nominal line, just
as an ordinary plane is determined by three different ordinary
points not on a straight line.
If two points of a nominal line
lie on a nominal plane, then all the points of that line lie on
that plane
The intersection of two nominal planes is a nominal
O, A,

B and

C,

line, etc.

The measurement of angles in the new geometry is the same

as that in the ordinary geometry
the angle between two
nominal lines is defined as the angle between the circles with
which these lines coincide at their intersection. The measurement of length is as before. Inversion in a sphere through O
is equivalent to reflection in the nominal plane coinciding with
that sphere. Displacements, being point-transformations
according to which every point of the domain is transformed
into a point of the domain, in such a way that nominal lines
;

remain nominal lines, and nominal lengths and angles are

unaltered, will be given by an even number of inversions in the
spheres of the system.

160

NON-EUCLIDEAN GEOMETRY

[ch.vhi.

Thus, the geometry of these nominal points, lines, and planes

Its
identical with the ordinary Euclidean Geometry.
every proposition valid in
elements satisfy the same laws
and from the theorems
the one is also valid in the other
of the Euclidean Geometry those of the Nominal Geometry
can be inferred, and vice versa.
The plane geometry of the nominal points and lines described
in the preceding sections is a special case of the more general
plane geometry based upon the definitions of this section.
is

97. The System of Circles orthogonal to a Fixed

Circle.

We proceed to discuss the geometry of the system of circles

orthogonal to a fixed circle, centre O and radius k. We shall
Then the system of
call this circle the fundamental circle.
circles has power
with respect to O.

Pio. 107.

Let A and B be any two points within the fundamental

and A', B' the inverse points with respect to that circle.
Then A, A', B, B' are concyclic, and the circle which passes
through them cuts the fundamental circle orthogonally.
There is one and only one circle orthogonal to the fundamental
circle which passes through two different points within that
circle

circle.

In discussing. the properties of the family of circles orthogonal to the fundamental circle, we shall call the points within
that circle nominal points. The points on the circumference
of the fundamental circle are excluded from the domain of
the nominal points.*
* In this discussion the nominal points, etc., are defined somewhat
differently from the idea^ points, etc.. in the paper referred to on p. 156.

96,97]

SYSTEM OF ORTHOGONAL CIRCLES

161

We define the nominal line through any two nominal j)oints

as the circle which passes through these two points and cuts the
fundamental

circle orthogonally.

Two different nominal points A, B always determine a nominal

AB, just as two different ordinary points A, B always
determine a straight line AB. The nominal points and lines
also obey the " axioms of order."
We define the angle between two intersecting nominal lines
line

as the angle between the tangents at the common point, within

fundamental circle, of the circles with which the nominal

the

lines coincide.

We

have now to consider in

define parallel nominal lines.

what way

it will

be proper to

Fig. 108.

Let A M (Fig. 108) be the nominal line through A perpendicular

in other words, the circle of the system
to the nominal line BC
which passes through A and cuts the circle of the system through
BC orthogonally. Imagine A M to rotate about A so that these
nominal lines through A cut the nominal line through BC at a
gradually smaller angle. The circles through A which touch
the circle through BC at the points U and V, where it meets
the fundamental circle, are nominal lines. They separate the
lines of the pencil of nominal lines through A, which cut BC
from those which do not cut it. All the lines in the angle ^
shaded in the figure do not cut the line BC
all those in the
angle ^p, unshaded, do cut this nominal line.
;

N.-E.a.

I,


162

NON-EUCLIDEAN GEOMETRY

This property is what is assumed in the

on which the Hyperbolic Geometry is based.

[oH.vm.

Parallel Postulate

We

are therefore
led to define parallel nominal lines in the plane geometry we
are investigating as follows
:

The nominal

lines through

a nominal point parallel

to

nominal

line are the two circles of the system passing through

the given point which touch the circle with which the given nominal
line coincides at the points where it cuts the fundamental circle.

Thus we have in this geometry two parallels a right-handed

and a left-handed parallel and these separate the
lines of the pencil which intersect the given line from those
which do not intersect it.

parallel

98. At this stage we can say that any of the theorems of

the Hyperbolic Geometry which involve only angle properties
will hold in the geometry of the circles, and vice versa.
Those
involving metrical properties of lines we cannot discuss until
the nominal length of a nominal segment has been defined.

Fia. 109.

For example, it is obvious that there are nominal triangles

whose angles are all zero (Fig. 109). The sides of these triangles
are parallel in pairs, and we regard parallel lines as containing

an angle

zero.

Further, we can prove that the sum of the angles in any nominal
triangle is less than two right angles, by inversion, as follows
Let Cj, Cg, Cg, be three circles of the system i.e. three
nominal lines forming a nominal triangle, say PQR. We
suppose these circles completed, and we deal with the whole
:

97, 98, 99]

NOT-INTERSECTING NOMINAL LINES

163

circumference of each. Invert the circles from the point of

intersection R' of Cj and C2, which lies outside the fundamental
circle.
Then the nominal lines C^ and Cg become two straight
lines Ci and C2', through the inverse of R.
Also the fundamental circle C inverts into a circle C', cutting Cj' and Cg' at
right angles, so that its centre is at the point of intersection
of these two lines. Again the circle C3 inverts into a circle Cg',
cutting C' orthogonally. Hence its centre lies outside C.
We thus obtain a curvilinear triangle in which the sum of
the angles is less than two right angles
and since the angles
in this triangle are equal to those in the nominal triangle, Our
result is proved.
Finally, it can be shown that there is always one -and only one
circle of the system which will cut two not-intersecting circles
of the system orthogonally. In other words, two not-intersecting
nominal lines have a common perpendicular.
All these results we have established in the Hyperbolic
Geometry. They could be accepted in the geometry of the
circles for that reason.
;

99. As to the measurement of length,

length of a nominal segment as follows

we define the nominal

The nominal

length of

any nominal segment AB

is

equal to

AV /BV^
^VAU/ BU

log

where U and V are the points where the circle which coincides
with the nominal line AB cuts the fundamental circle. (Cf.
Fig. 107.)

nominal length of AB is the same

Also the nominal length of the complete line
is infinite.
If C is any point on the nominal segment AB
between A and B, the nominal length of A B is the same as the
sum of the nominal lengths of AC and CB.
Let us consider what effect inversion with regard to a circle
of the system has upon the nominal points and lines.
Let A be a nominal point and A' the inverse of this point
in the fundamental circle.
Let the circle of inversion meet the fundamental circle in C,
and let its centre be D (Fig. 110).
Suppose A and A' invert into B and B'.
N.-B.Q.
L2

With

this definition the

as that of BA.

164

NON-EUCLIDEAN GEOMETRY

[cn.vin.

Since the circle AA'C touches the circle of inversion at C,

inverse also touches that circle at C. But the points
A, A', B, B' are concyclic, and the radical axes of the three
circles AA'C, BB'C and AA'B'B are concurrent.

its

Therefore BB' passes through O and OB 0B' = 0C2. Thus

the circle AA'B'B is orthogonal to the fundamental circle and
also to the circle of inversion.
It follows that if any nominal point A is changed by inversion
with regard to a circle of the system into the point B, the nominal
line AB is perpendicular to the nominal line with which the
.

of inversion coincides.
shall now prove that it is " bisected " by that nominal
Let the circle through A, A', B and B' meet the circle
line.
of inversion at M and the fundamental circle at U and V
It is clear that U and V are inverse points with
(Fig. 111).
regard to the circle of inversion
circle

We

BV CV
=
AU CA'

Then we have

-r

cy
cb'

AV
AU

cv^ _
BV
_ /Mvy
Bu' CA.CB~Clvr2~VMU/

AV
AU

MU

MV

Thus the nominal length

BM.

of

AV
BU

MV
MU
of

AM

BV
BU*
is

equal to the nominal length

99]

INVERSION AND REFLECTION

Therefore

we have the

165

following result

Inversion with regard to any circle of the system changes any

point A into a point B, such that the nominal line AB is perpendicular to and " bisected " by the nominal line tvith which the
circle of inversion coincides.

Fio. 111.

In other words,

Any nominal point takes up

nominal line coinciding with the

the position of its

circle of inversion.

image in the

We shall now examine what effect such an inversion has upon

a nominal line.
Since a circle orthogonal to the fundamental circle inverts
into a circle also orthogonal to the fundamental circle, any
nominal line AB inverts into a nominal line ab, and the points
U and V for AB invert into the points u and v for ab (Fig. 112).
When the circle of inversion and the nominal line AB intersect, the lines AB and ab meet on the circle of inversion.
Denoting this point by M, it is easy to show that the nominal
lengths of AM and BM are respectively equal to the nominal
lengths of aM and 6lVI. It follows that the nominal length of
the segment AB is unaltered by inversion with regard to any
circle of the system.
The same result can be obtained immediately from the
corresponding figure when the nominal line AB does not cut
the circle of inversion.

NON-EUCLIDEAN GEOMETRY

166

The preceding

results

may be summed up

as follows

[ch.vhi.

Inversion with regard to any circle' of the system has the same
effect upon the nominal points and lines as reflection in the
nominal line with which the circle of inversion coincides.

The argument of 95 can now be applied to the geometry

of this family of circles. Successive inversion with regard to
two circles of the system corresponds to a displacement in
two dimensions. We can always fix upon two circles of the
system which will change a nominal segment AB into a new
position, such that A coincides with P and AB lies along a given
nominal line through P. The method of superposition is thus
available in this geometry, and any theorems in the Hyperbolic
Geometry involving congruence of linear segments can be at
once " translated " into it.
100. We notice that the definition of the nominal length
of a segment fixes the nominal unit of length. We may take
this unit segment on one of the diameters of the fundamental
circle, since these lines are also nominal lines of the system.
it be the segment OP
Then we must have

Let

that

is,

/Pu\
log(^ J = l;

(Fig. 113).

that

is,

PU

e.

99, 100. 101]

THE UNIT NOMINAL SEGMENT

167

Thus the point P divides the diameter in the ratio e 1.

The unit segment is thus fixed for any position in the domain
"
of the nominal points, since the segment OP can be " moved
:

Fio. 113.

so that one of its ends coincides with any given nominal

point.
different expression for the nominal length, viz.,

would simply mean an alteration

and taking
would have the same

in this unit,

logarithms to the base a instead of

efiect.

101. We are now able to establish some further theorems

of Hyperbolic Geometry, using the metrical properties of this

Nominal Geometry.
place we can say that Similar Triangles are
For if there were two nominal triangles with the
same angles and not congruent, we could " move " the second
so that its vertex would coincide with the corresponding angular
point of the first, and its sides would lie along the same nominal
lines as the sides of the first.
We would thus obtain a "quadrilateral " whose angles would be together equal to four right
angles
and this is impossible, since we have seen that the
sum of the angles in these nominal triangles is always less
than two right angles.

In the

first

impossible.

We

that is, they

approach each other. This follows from the
figure for nominal parallels and the definition of nominal
also see that parallel lines are asymptotic

continually
length.

NON-EUCLIBEAN GEOMETRY

168

[cH.vm.

Further, it is obvious that as the point A moves away along

the perpendicular MA to the line BC (Fig. 108), the angle of
parallelism diminishes from - to zero in the limit.

We

shall

now prove

the segment y,

is

that the angle of parallelism,

given by

Consider a nominal line and a parallel to

Let

AM

(Fig. 114)

and AU the

it

11(2?),

^^^

through a point

A.

be the perpendicular to the given line ML)

parallel.

Let the figure be inverted from the point M', the radius of
inversion being the tangent from M' to the fundamental
circle.

Then we obtain a new figure (Fig. 115) in which the corresponding nominal lengths are the same, since the circle of
The lines AM and MU
inversion is a circle of the system.
become straight lines through the centre of the fundamental
Also, the circle AU
circle, which is the inverse of the point M.


THE ANGLE OF PARALLELISM

101]

169

becomes the circle au, touching the radius mu at u, and cutting

ma at an angle 11(59). These radii mu, mh are also nominal
lines of the

system.

Let the nominal length of

Then we have

:P

AM

be p.

= log(^/^)

= logf^V-)=logf-)^
^ \acj
\acl

Fio.

m^J

115.

But from the geometry of Fig. 115, remembering that au

cuts he at the angle n(^), we have
ac = i{l-tan(^-niH))}.

where k

is

the radius of the fundamental

Therefore

2^

= ^^g ^o*

W^

circle.

and
geometry there will be three kinds of circles.
the circle with its centre at a finite distance

Finally, in this

There

will

be

(i)

NON-EUCLIBEAN GEOMETRY

170

the limiting-curve, witli

[ch.

vm.

centre at infinity, or at a point

(iii) the equidistant-curve,
with its centre at the ideal * point of intersection of two lines
which have a common perpendicular.
All these curves are ordinary circles, but they do not
belong to the system of circles orthogonal to the fundamental
(ii)

where two

parallels

meet

its

and

circle.

As to the first, the nominal lines through a point A are all

cut orthogonally by the circles of the coaxal system with A
and its inverse point A' as Limiting "Points. Thus these
circles are the circles of this nominal geometry with A as their
centre.
They would be traced out by the end of a nominal
segment through A, when it is reflected in the nominal lines
of the pencil.
As to the second, the circles which touch the fundamental
circle at a point U cut all the circles of the system which
pass through U orthogonally. They are orthogonal to the
pencil of parallel nominal lines meeting at infinity in U.
Thus these circles are the circles of this nominal geometry
with their centre at the point at infinity common to a pencil
of parallel nominal lines. They would be obtained when the
reflection takes place in the lines of this pencil.
As to the third, all circles through U, V cut all the nominal
lines perpendicular to the line AB (cf. Fig. Ill) orthogonally.
Thus these circles are the circles of the nominal geometry with
their centre at the ideal point common to this pencil of notintersecting nominal lines.
They would be obtained when
the reflection takes place in the lines of this pencil.
These three circles correspond to the ordinary circle, the
Limiting-Curve and the Equidistant-Curve of the Hyperbolic
Geometry.
102. The Impossibility of proving Euclid's Parallel
Postulate.

We can now assert that it is impossible for any inconsistency

to exist in this Hyperbolic Geometry. If such a contradiction
entered into this plane geometry, it would also occur in the
interpretation of the result in the nominal geometry. Thus
a contradiction would also be found in the Euclidean Geometry.
We can therefore state that it is impossible that any logical
Cf. 37.

101, 102, 103]

THE PARALLEL POSTULATE

171

inconsistency could arise in the Hyperbolic Plane Geometry,

provided no logical inconsistency can arise in the Euclidean
Plane Geometry. It could still be argued that such a contradiction might be found in the Hyperbolic Solid Geometry.
An answer to such an objection is forthcoming at once. The
geometry of the system of circles, all orthogonal to a fixed
circle, can be readily extended into a three-dimensional system.
The nominal points are the points inside a fixed sphere, excluding the points on the surface of the sphere from their domain.
The nominal lines are the circles through two nominal points
cutting the fixed sphere orthogonally.
The nominal planes are
the spheres through three nominal points cutting the fixed
sphere orthogonally. The ordinary plane enters as a particular
case of these nominal planes, and so the plane geometry just
discussed is a special case of a plane geometry of this system.

With suitable definitions of nominal lengths, nominal parallels,

we have a solid geometry exactly analogous to the Hyper-

etc.,

Geometry. It follows that no logical inconsistency

could arise in the Hyperbolic Solid Geometry, since, if such did
occur, it would also be found in the interpretation of the
result in this Nominal Geometry, and therefore it would enter
into the Euclidean Geometry.
By this result our argument is complete. However far the
Hyperbolic Geometry is developed, no contradictory results
could be obtained. This system is thus logically possible,
and the axioms upon which it is founded are not contradictory.
Hence it is impossible to prove Euclid's Parallel Postulate,
since its proof would involve the denial of the Parallel Postulate
of Bolyai and Lobatschewsky.
bolic Solid

103. The System of Circles cutting a Fixed Circle

diametrally.
We shall now discuss the geometry of the system of circles
cutting a fixed circle centre, O and radius k, diametrally.
The points in which any circle of the system cuts the fixed
circle are to be at the extremities of some diameter.
We
shall call the fixed circle, as before, the fundamental circle.
The system of circles with which we are to deal has power k^
with respect to O.
Let A and B be any two points within the fundamental
circle, and A',
B' the points on OA and 08, such that

OA.OA'=

-A;2

and

OB.OB'= -P.

NON-EUCLIDEAN GEOMETRY

172

Then

and the
fundamental

circle

which passes

circle

diametrally

A, A', B, B' are concyclic,

through

them

the

cuts

[cH.vni.

one, and only one, circle cutting the

fundamental circle diametrally, which passes through two
different points within the fundamental circle.
(Fig.

116).

There

is

Fio. 116.

In discussing the properties of the family of circles cutting

the fundamental circle diametrally, two methods can be
We can restrict the nominal points of the geometry
followed.
In
to the points within and upon the fundamental circle.
this case we regard the points on the circumference at the
extremities of a diameter as one and the same nominal point.
In the other case, we extend the field of nominal points outside
the circle to infinity, and the points on the circumference do
not require special treatment.
These two alternatives, we shall see below, correspond to
the two forms of the Elliptic Geometry, in one of which every
straight line intersects every other straight line in one point,
while in the other form, straight lines have always two points
The nominal lines are the circles which cut
of intersection.
the

fundamental

circle diametrally.

When the field of nominal points is restricted to points within

or

upon the fundamental circle, any two different nominal

B determine a nominal line AB. Also any two nominal

points A,

lines

must

intersect at

a single nominal point.

THIRD SYSTEM OF CIRCLES

103]

173

When tlie domain of the nominal points is both within and

without the fundamental circle, two nominal points do not
always determine uniquely a nominal line. If the points
A and B are upon the circumference of the circle at opposite
ends of a diameter, a pencil of nominal lines passes through
A and B. Again, if the points A and B lie on a line through
O and OA 0B=

-h^, the same remark holds true.

Further, with the same choice of nominal points, every
nominal line intersects every other nominal line in two nominal
.

points.

The simplest way of discussing the properties of the system

of circles with which we are dealing, is to make use of the
fact that they can be obtained by projecting the great circles
of a sphere stereographically from a point on the surface of
the sphere on the tangent plane at the point diametrally
opposite.
If the centre of projection is a pole of the sphere,
the equator projects into the fundamental circle, and one
hemisphere projects into points outside this circle, the other
into points within it.
This projection is a conformal one,
and the angle at which two great circles intersect is the same
as the angle at which the corresponding circles in the plane
cut each other.
We define the angle between two nominal lines as the angle
between the circles with which they coincide.
We are now able to prove some of the theorems of this
Nominal Geometry.
the great circles perpendicular to a given great
the poles of that circle, it follows that all
the nominal lines perpendicular to a given nominal line intersect
at one point, in the case when the nominal points are within or
upon the circumference of the fundamental circle ; in two points,
when this field is both within and without. (Of. lb-11.)
The point of intersection is spoken of as a pole, or the pole,
of the line.
Again, in a right-angled spherical triangle ABC, in which
Since

all

circle intersect at

is

the right angle, the angle at A

=a

right angle, according

as the pole of AC lies on CB produced, or coincides with B,

or lies between C and B.
When translated into the language of the nominal geometry,
we have the theorem which corresponds to 78 (1).
Further, the sum of the angles of a spherical triangle is
greater than two right angles. It follows, since the projection

NON-EUCLIDEAN GEOMETRY

174

[ch.

vm.

conformal, that the sum of the angles of a nominal triangle

in this geometry is greater than two right angles.
(Cf. 78 (3).)
However, the metrical properties of this geometry cannot
be treated so easily as were the corresponding properties in
the geometry of the system of circles cutting the fundamental
circle orthogonally.
The same argument to a certain extent
applies, but in the definition of nominal lengths the intersections with an imaginary circle have to be taken.
It should
be added that in the extension to solid geometry the system
of spheres cutting a fixed sphere diametrally has to be

is

employed.

The fuller discussion of this nominal geometry will not be

undertaken here. If it is desired to establish the fact that
no contradiction could appear in the Elliptic Geometry,
however far that geometry were developed, there are simpler
methods available than this one. The case of the Hyperbolic
Geometry was discussed in detail, because it offered so
elementary a demonstration of the impossibility of proving
the Parallel Postulate of Euclid.

104. We have already quoted some remarks of Bolyai's

on the question of whether the Euclidean or the Non-Euclidean
Geometry is the true geometry.* We shall conclude this
presentation of our subject with two quotations from modern
geometers on the same topic
:

"

then," says Poincare, " are we to think of the

question
Is Euclidean Geometry true ?
It has no meaning.
We might as well ask if the metric system is true, and if the
old weights and measures are false
if Cartesian coordinates
are true and polar coordinates false.
One geometry cannot
be more true than another
it can only be more convenient.
Now, Euclidean Geometry is, and will remain, the most
convenient
first, because it is the simplest, and it is so not
only because of our mental habits or because of the kind of
intuition that we have of Euclidean space
it is the simplest
in itself, just as a polynomial of the first degree is simpler than
a polynomial of the second degree
secondly, because it
sufficiently agrees with the properties of natural solids, those
bodies which we compare and measure by means of our senses .f

What
:

* Cf. 16.
j-

Poincare,

La

Science

et

VHypothese.

English translation, p. 50.

103, 104]

IS

EUCLIDEAN GEOMETRY TRUE

And

175

another French geometer writes

are then entitled to say that the geometry which most
closely resembles reality is the Euclidean Geometry, or at least
one which differs very little from it
the error is too
small to be apparent in the domain of our observations and
with the aid of the instruments at our disposal.
" In a word, not only have we theoretically to adopt the
Euclidean Geometry, but in addition this geometry is physically
true."*
The matter can be put in another way. The question
whether the Euclidean Geometry is 'the true geometry has no
place in Geometry the Pure Science. It has a place in
Geometry the Applied Science. The answer to the question
if an answer can be given
lies with the experimenter.
But
his reply is inconclusive.
All that he can tell us is that the
sum of the angles of any triangle that he has observed
however great the triangle may have been is equal to two
"

We

right angles, subject to the possible errors of observation.

To say that it is exactly two right angles is beyond his power.
One interesting point must be mentioned in conclusion.
In the Theory of Relativity, it is the Non-Euclidean Geometry
of Bolyai and Lobatschewsky which, in some ways at least,
is the more convenient.
Gauss's jesting remark that he would
be rather glad if the Euclidean Geometry were not the true

geometry, because then we would have an absolute measure

of length, finds an echo in the writings of those who in these
last years have developed this new theory .f
*

Hadamard,

Let^ons de Odomitrie dementaire, vol.

i.

p.

286 (Paris, 1898).

Also the letter to

t Cf. the letter to Taurinus, quoted on p. 24.
Gerling given in Gauss, Werke, vol. viii. p. 169.
A similar remark is to be found in Lambert's Theorie der ParallelLinien, 80
see Engel u. Stackel, loc. cit. p. 200.
;

\
:

INDEX OF NAMES OF AUTHORS.

Numbers
Amaldi, U., 88.
Archimedes (287-212

Klein, F., 32, 39, 132, 154.

B.C.), 5, 15,

18, 88.

Bartels,

M.

J.

C.

(1769-1836),

Beltrami, E. (1835-1900), 12, 154.

Beman, W. W.,

4.

W.

(1775-1856),

20,

19,

27, 28, 36.

R.

(1875-1911),

Legendre, A. M.

Ch.

I.

12,

17,

Cayley, A. (1821-1895), 32, 154.

W. K. (1845-1879), 39.

P., 136-141.

Napier, J., Baron of Merchiston

(1550-1617), 102.

Clifford,

Pasch, M.,

CooUdge, J. L., 136, 145.

Dedekind, J. W. R., 4.
Dehn, M., 18, 19.

Poincar^

33, 46, 68, 63, 71, 175.

Enriques, F., 4, 88.
Euclid (circa 330-275 B.C.), Ch.

H. (1854-1913), 154.

Proclus (410-485), 10.

I.

Riemann, G.

F. B. (1826-1866).
27, 32, 38, 39, and Ch. VI.

Saccheri, G. (1667-1733), Ch. I.

Schumacher, H. K. (1780-1850),
25, 37.

Finzel, A., 88.

W.

B., 17, 18.

Friedlein, G. (1828-1875), 10.

Gauss, C. F. (1777-1855), Ch. I.

Gerard, M. L. J., 106, 136-141.
Gerling, G. L. (1788-1864), 19,
21-23, 175.
Greenstreet, W. J., 155.
J. S., 175.

Halsted, G. B., 5, 27, 30, 36, 40,

87.

Heath, T. L.,

3, 42, 44, 56.

J.

155, 174.

Engel, F., 13, 17, 18, 20, 24, 27

Hadamard,

17.

(1752-1833), 12,

15-18, 39, 127.

63, 73, 106.
Lindenau, B. A. (1780-1854), 21.
Lobatschewsky, N. J. (1793-1856),

Mansion,

26, 36, 39, 47, 68, 90, 131, 156.

Frankland,

H. (1728-1777),

J.

23, 175.

Liebmann, H.,

Bertrand, L. (1731-1812), 17, 18.

Bolyai, J. (1802-1860), Ch. I.

Bonola,

Kiirschilk, J., 36.

Lambert,

32.

Bolyai,

refer to pages.

10, 12, 15.

Hilbert, D., 5, 18, 42, 48, 54, 71,
74, 87, 88, 127, 157.
1, 4,

Schweikart,

F.
21-23, 26, 37.

K.

(1780-1859),

Sommerville, D. M. Y., 33, 154.

Stackel. P., 13, 17, 18, 20, 24, 27,
28, 30, 31, 33, 36, 175.

Taurinus, F. A. (1794-1874), 23,

24, 175.

Thibaut, B. F. (1775-1832), 25.

Townsend, E. J., 5.
Vasiliev, A. V., 33.
Vitali, G., 4, 5.

Wachter, F. L. (1792-1817),

Young,^W. H.,

106, 136.

21.

SUBJECT INDEX.
Numbers

rejer to pages.

Absolute Science of Space, 27-29.

Absolute unit of length, 17, 25.
Absolute units as compared with

Element of

arc, in Elliptic Geometry, 152.

in Hyperbolic Geometry, Car-

tesian Coordinates, 112-114.

relative, 17, 90.

Angle of parallelism, 41, 50, 109.

Associated Right-angled Triangl es
63-66.

Arc

CSrcle,

Area

in

Limiting-Curve Coordinates,
117-118.

Polar Coordinates, 114-116.

Element of area, in Elliptic Geometry, 152.
in Hyperbolic Geometry, in Cartesian Coordinates, 122-123.
in
Limiting-Curve Coordiin

of, 118.

of, 124.

of infinite radius (see LimitingCurve), 80.

Three kinds of, in Hyperbolic

Geometry, 83, 170.
One kind of, in Elliptic Geometry, 135.

Complementary segments,

51.

Hyperbolic functions of, 98-99.

Congruence, Axioms of, 2, 5.
of infinite areas, 17.

Consistency of the Non-Euclidean

Geometries, Ch. VIII.
Correspondence
between rightangled triangle and quadrilateral with three right angles,
59-63.
Courbe-limits (see Limiting-Curve),
80.

nates, 119-121.
in Polar Coordinates, 123-124.
Equidistant-Curve, 82.
Base-line of, 83.
Concave to base-line, 83.
Arc of, 118.
Equivalent polygons, 84.
Equivalent triangles, 85.
Theorems on, 85-88.

Euclid's imexpressed axioms, 3, 4.

Parallel Postulate or Parallel

Hypothesis, 2.
Postulates I. -II., 3.
Postulate III., 5, 74.
Excess of a triangle, 134.
Exterior angle. Theorem of

(I. 16).

3, 130, 131.

Defect of triangle, 54.

of polygon, 89.
Direction of parallelism, 45.
Direction-theory of parallels, 11.
Displacement equivalent to two
reflections or inversions in the
nominal geometry, 158, 159,
166.

in

triangle

with

one

point at infinity, 48.

Geometry, Absolute,
Astral, 22.
Elfiptic, 39, 131.

Euclidean, 2.
Hyperbolic, 39.

29.

angular

SUBJECT INDEX

178

Measure of area of triangles and

Greometry, Absolute

polygons, 88-90, 120, 124-126,

Imaginary, 34.
Nominal, 157.
Nod -Archimedean,
Non-Euclidean, 1.

135.

Measurement of

18.

Parabolic, 39.

Semi-Euclidean, 18.
in the infinitesimal is
Euclidean, 111.
on the Limiting-Surface is
Euclidean, 29.
on the Sphere is independent of
the Parallel Postulate, 29.
Grenzkreis (see Limiting-Curve),

Geometry

80.

Hubert's Axiom of Parallels, 42.

Horocycle or horicycle (see Limit-

One-sided surface, 131.

Order,

Axioms

of, 157.

Pang^omdtrie, 36.
Parallel constructions, 71-77.
Parallel lines, Euclid's treatment
of, 1, 2.

Bolyai's treatment of, 28, 29.

ing-Curve), 80.
Infinite,

angles, 104, 105.

Napier's Rules, 102.

Not-intersecting lines, Two, in a
plane have a common perpendicular, 54.
diverge on either side of common
perpendicular, 58.

contrasted

with

im-

bounded, 38, 39, 127.

Hubert's treatment of, 42, 43.

Lobatschewsky's* treatment of,
33-35, 40-42.

the Nominal Geometry corresponds to a reflec-

Right-handed and left-handed,

tion, 158, 166.

Theorems on, 43-50,

Inversion

in

Hypotheses of the
Acute Angle, Right Angle, and
Obtuse Angle, 15-17.
Length of a line in Elliptic Geo-

Legendre's

metry, 129-131.
Limiting-Curve, or horocycle, 80.

Arc of, 119.

Area bounded by arc

two of its
Axes of, 52.

of,

and

axes, 120.

56.

Parallel Postulate, Euclid's, 2.

Impossibility of proving, 10,
29-32, 35-36, 170-171.

Two

theorems independent

of,

8-10.

Pasch's Axiom, 3.
Perpendicular bisectors o sides
of a triangle, 68-71, 84.
Points, at infinity, 47, 66.
Ideal, 67.
Nominal, 156, 160, 172.

Proper and improper, 66.

Coordinates, 116.

Equation of, 97.

Theorems on, 81,

43.

82, 95-97.

Limiting-Curves, Concentric, 82.

Area bounded by arcs of two

concentric, and two of their
axes, 120.

Theorems on Concentric, 91-95.

Limiting-Surface, 35.
Linea-L, 80.
Lines, Asymptotic, 15, 29, 56.
Divergent, 58.
Ideal, 68.

Points, Corresponding, 77.

Theorems on, on parallel lines,
78-80.

Pole of a line, 129.

Postulate of Archimedes,

5, 18.

of Dedekind, 4.
Principle of Continuity, 4.
Problems of construction, independent of Parallel Postulate
and Principle of Continuity,
5-8.
in Hyperbolic

Geometry,

65.

Nominal, 156, 161, 172.

Not-intersecting, 15, 34, 40, 54.
Parallel, 1, 28, 34, 40, 42.

Quadrilateral, Saccheri's, 51, 134.

with three right angles, 52, 134.

SUBJECT INDEX

179

Ray, Definition of,

Relativity, Theory

of, 175.

Representation of

Non -Euclidean

Geometry

Euclidean, 154-

pkssing through a fixed point

of proof for
of angles of a triangle, 25.

Extension to Solid Geometry,

in

System of circles

5.

Extension to Solid Geometry,

156-159.

156.

Rotation-method

sum

170.

HsTpotheses of Acute
Right Angle,
and
Obtuse Angle, 13-15.
Similar triangles impossible in

Saccheri'a

Angle,

Non-Euclidean Geometry,

169-160.

Trigonometrical

(see Limit-

ing-Surface), 29.
Squaring of the circle, 29.
Sum of the angles of a triangle,
and
the
hypotheses
of
Saccheri and Tjcgendre, 12-

angle, in
136-152.

Elliptic

and the Postulate of Archi-

in Hyperbolic Plane, 100-102,

108-110.

The Cosine Rule, 104, 109.

The Sine Rule, 103, 109.
Truth of the Euclidean Geometry,
174, 175.
of Bolvai, 27-32.
Gauss, 19-26.

Lambert,

18.

in Elliptic Geometry, 132-134.

in Hjrperbolic Geometry, 53.
System of circles, cutting a fixed
circle diametrally, 171-174.

Extension to Solid

Geometry

174.

orthogonal
160-170.

to

fixed

circle,

tri-

Plane,

Work

18.

medes,

105,

Trigonometry, of right-angled

54,

135.

Sphere of infinite radius

Functions,

140.

17.

Legendre, 15-19.
Lobatschewsky, 32-38.

Riemann, 38-41.
Saccheri, 12, 15.

Schumacher,

26.

Schweikart, 21.
Taurinus, 23.

Wachter, 21.

PRINTED BY ROBERT MACLEHOSE AND C. LTD. AT THE UNIVERSITY PRESS,

GLASGOW, OREAT BRITAIN.