SCANNED BY PRESERVATION SERVICES

29/06/94
gfrinlnmjb Mathematical gracts
No. 6

AN INTRODUCTION
TO THE THEORY OF
AUTOMORPHIC FUNCTIONS
 AN INTRODUCTION
TO THE THEORY OF
AUTOMORPHIC FUNCTIONS

BY

LESTER R. FORD, M.A. (Harv.)

LECTUKER IN MATHEMATICS IN THE UNIVERSITY OF EDINBURGH

DEPARTMENT OF MATHEMATICS
UNIVERSITY OF TORONTO

LONDON
G. BELL b SONS, LIMITED
5

YORK HOUSE, PORTUGAL STREET

1915
3*7
 PREFACE
OWING largely to the researches of Poincare and Klein the

domain of Automorphic Functions has expanded enormously

during the last thirty-five years and the ramifications of the
:

subject have extended into many and diverse h'elds. This has
caused embarrassment in the selection of materials for a book
of modest dimensions, and has necessitated a brief treatment,
or in some cases the exclusion, of many important and attractive

subjects. The aim throughout has been to present in as thorough

a manner as possible the concepts and theorems on which the.
theory is founded, and to describe in less detail certain of its

important developments.
The present trart had its origin in a series oi' lectures on

Automorphic Functions given to the Mathematical Research Class

of the University of Edinburgh during the Spring Term of 1 !)!").

I wish to express a grateful acknowledgment of my indebted-

ness to Professor Whittaker. who has
read the manuscript during
the course of its
preparation, and lias made many valuable

ions: and to Mr Herbert Bell, who has assisted in the

preparation of the Bibliography.

L. R. F.

THK UNIVERSITY OF EDINBURGH,

1915.
 CONTENTS
CHAPTER I

LINEAR TRANSFORMATIONS
PAGE
1. The Linear Transformation 1

2. The Fixed Points of the Transformation . 3

3.

4.

5.
Conformal Transformations
Inversion in a Circle
The Transformed Elements
.........
of ....
Length and of Area
4
6
9
6. Types of Linear Transformations . . . . . . . 10

CHAPTER II

GROUPS OF LINEAR TRANSFORMATIONS

Notation 16
7.

8. Definition of a
Group ...
... ..... 17
'..

10.
11.
Examples of Groups
Transforming a Group
.

....
Continuous and Discontinuous Groups
17

1 -2.

13.
The Modular Group
The Fundamental Region ........
14.
15.
1C.
Fuchsian Groups ..........
Fundamental Region for the Modular Group

.....
Fundamental Region for a Fuchsian Group 26
17. The Cycles 29
18.

H).
The Transforms
.......
of the Fundamental Region

Groups without Principal Circles

32
35

CHAPTER III

AUTOMORPHIC FUNCTIONS

21.
tomorphic Functions .........
Properties of Simple Automorphic Functions
37
38
22. The Thetafuchsian Series 44
_':;. <
'onvergence of the Series ... 46
 ACTOMOKI'HIC FUNCTIONS

CHAPTER IV

TFIK KiKMAN.x-ScinvAHZ TRIANGLE FUNCTIONS

mal
tttial lunation
of Half-Plane
.... PAGE
50
51
Map .

: ii.-tiull .53
: ; reo Kinds ..f
(Jn.ups
54
\+ u +v=} 54
!.
\+n+v<l ...
...... 55
56
iMphii- Projection
.

ise III. A + M + '^-l 57

jular Solids 59

CHAPTER V
NON-EUCLIDEAN GEOMETRY

63
34.
solute, or Cay ley an, Geometry
The Three Types of Geometry ... 64
35. lu-piv.-entution of Non-Euclidean Straight Lines by Circles in the
Plane. The Elliptic Case 65
3*:. Tli.-
Hyperbolic Case 67
37. A i 'plication to Groups of Linear Transformations . . .69
CHAPTER VI
UNIFORMISATION

38. i: Surfaces 72
73
onectivity of Surfaces
of the Fundamental Region
roblem of Uniformisation
of a....
.......
Group .
74
76

;:v

41.
(Jniformisation
Whin.-.
by means of Automorphic Functions ....81 77
78
Group
\'>. \\ .

Group
ik of S.-lmttUy and liuniside .... .87
85

88
 CHAPTER I

L1XEAR TRANSFORMATIONS

1. The Linear Transformation. If z is a complex quantity

whose real part is x and whose imaginary part is iy, it is
customary to represent z by a point in a plane whose abscissa
is x and whose ordinate is y, the coordinates being referred to

perpendicular axes. If x, or y, or both become infinite, the point

recedes to an infinite distance from the origin of coordinates.
This is expressed by z = oo and infinity is considered to be a
,

single point in the complex plane. In this respect the conception

of infinity differs from that employed in projective geometry,
where the infinite region is considered to be a line.*
Consider z' =f(z), where f(z) is a function of z, and let the
variable z be represented on a second plane. To each point z
for which the function f(z) is defined there correspond one
or more values of z' given by the above equation. To points,
curves,and areas in the 5-plane there correspond by virtue of
the equation points, curves, and areas in the z'-plane.
The configurations in the one plane are said to be trans-
formed into configurations in the other plane. The whole theory
of automorphic functions depends upon a particular type of
transformation, defined as follows :

DEFINITION. The transformation

(1)

where d are constants (real or complex) and ad

a, b, c, be =t=

is called a linear transformation.^

* The reason for this is that the kind of transformations most
frequently con-
sidered in the theory of functions of a complex variable transform the infinite region
into a point in the finite part of the plane whereas ordinary projection in geometry
:

transforms the infinite region into a line.

t Also called a linear substitution, a homographic transformation, or a homo-
graphic substitution. If ad-bc = Q the equation reduces to z' = constant
) but this
;

case is without interest.

1
 AUTo.MukrilK' FUNCTION- [CH. I

/ he is called the determinant of the trans-

11 1)0 convenient to have always

ad-bc=I . . .
(2)

in the general case taking this value if the

nui illuminator of the fraction in the second member

y J(od he).
:i
(1) be solved for z.

cz -a
called the inverse of the transformation (1). It is a
i -a information with the same determinant as (1). We
state then
THKUJIK.M I. The inverse of a linear transformation is a
iDi'iiml'ion.

To each value of z there corresponds one, and only one, value

(
1) and conversely to each value of z there corresponds
;

and only one, value of z by (3). In other words, the z-plane

is transformed into the
z'-plane in a one-to-one manner.
It will be most serviceable to represent the values of z' not
.n a different
plane, but on the same plane and with the same
of coordinates as are used for representing z. In all
that follows but one plane will be used unless the contrary is

licitly stated. The results just found can be stated in the

,ving
THKOKKM 2. The z-plane is transformed into itself in a
ne manner by a linear transformation.
be subjected to a linear transformation,

a function of 0,

,
= =
. ,

^formation (1) and then making (4) is equivalent

ansl'nniiMtiun (;")). Now (5) is also a linear trans-
rmijiaiit in tin- form in which the fraction
'm ft-y).
It is wortli noting that if the
! ili and (1) are I-JK-II
unity, the determinant of
 1, 2] LINEAR TRANSFORMATIONS 3

(5) is unity also without further change. If z" be subjected to

a linear transformation, we get again a linear transformation of
0, and so on. Thus we have, in general
THEOREM 3. The successive performance of any number of
linear transformations is equivalent to a single linear trails-

formation.
The linear, or homographic, transformation has the property
that it leaves invariant the cross-ratio of four points. For if

zv % 3> ^4
are ^ our points, and z/, z z ', z3 ', z are the points into
which they are transformed, we have, on substituting for z {
'.

:/ their values from (1),

-?/)
-z -
(z/- z,' )(,?.;
"
__ (Z T 2 )(z 3
6

2. The Fixed Points of the Transformation. The points

which remain unchanged under the transformation (1) are found
by solving the equation
z = (az + b)i(rr. + d), or cz* + (d -
a)z
- b = . .
(7)

If c =4=
0, the fixed points are

t _- x
~2^~
---. _a-d- N /[(g-^)g

If c -
0, the fixed points are

There is but one fixed point if (a c?)

2
+46c = 0; or, when
ad - be = 1, if (a + d) = 4.
z
It is the point

=(a-rf)/2c. . .
(10)

Equation (7) will have more than two roots only in case it is
identically satisfied ;
that is, if c = d a = b = 0. In this case
every point unchanged in position.
is

THEOREM 4. The only linear transformation with more than

two fixed points is the identical transformation z' = z.
By means of the last theorem we are able to establish the
following proposition
THEOREM 5. There is one, and only one, linear transforma-
tion which transforms three distinct points z v z.2 z into three
3 ,

distinct points z/, z 2 ', z 3 '.

Weshall prove first that there is not more than one such trans-
 ro.MOltl'HK 1 I/NCTIUXS [CH. I

i. :'.= (,/; + /,)

=
(cz+d), and z" (aZ + P)l(yz+3) be
arrying %, z z z 3 into z^, z 2 z 8 Eliniinat-
, ,
'.

function of z', z" = (Az' + T$)/(Cz' + 'D).

a lin.-ar
11 Theorems and 3, without actually performing
I

"
depends linearly upon z for z is derived
:
;

from ;, by a linear transformation. When

z :,. by hypothesis z" = z' 0/, z 2 or Z B '. The
\ve liave ,

ation from :' to z" has then three fixed points, 5^',%',
ml by Theorem 4 z" = z. The two transformations are
which proves that there is not more than one
bhe same,
with the desired properties.
>

ni i.-it ion
That tln-iv is always one such transformation we prove by
it up. If none of the six values is infinite,
lly setting
l-r the transformation defined by
1 3 2 1 1 \

^ 3n q i^-j

;uati<.nwhich expresses the equality of the cross-ratios

and (zz v z 2 z 3 ). This is of the form (1) when solved
i

in terms of 0. It obviously transforms z v z 2 z 3 into z-^, z 2 z s , ', '.

(nation (11) depends upon both variables if each set of

points is distinct; the determinant of the transformation
then does not vanish [see footnote, Section 1].
:j
= oo 3 2 =oo or 3 = oo, it is necessary to replace the
1 1
1

, ,

.-I member of (11) by

, or 1!
2-2.,

and a similar change is necessary in the first

1

y ;

ber for an infinite value of /, z 2 or Z B In any case there ,

'.

transformat ion with the desired

properties, and the theorem
kblished.
n is a convenient form for use in
( 1 1
)
actually setting
a information carrying three given points into three

3. Conformal Transformations. -Let p be a

point in a plane,
l!il1 '
two curves issuing from p. Let
l)(1
be the
tithe curves (ie. the angle between their
tangents
tuenee \v.- shall consider the
angle positive if
'" M point on d toapoint on C, through the region
2, 3] LINEAR TRANSFORMATIONS 5

of the included angle the motion is counter-clockwise around p.

Now let a transformation of the plane be made such that p, C v C 2
'
are transformed into a point p' and two curves C/ and C 2 issuing
therefrom The transformed angle 0' will be that angle between
'

C/ and C 2 which includes the transforms of the points within

/
the original angle 0. It is positive if in passing from C 1 to
C./ through the included angle the motion is counter-clockwise
around p. The cases with which we shall be chiefly concerned
are those in which 6' = 0.

DEFINITION. A
transformation preserving the magnitude
of angles a conformed transformation. It is directly
is called
or inversely conformal according as the sign of the angle is
preserved or changed.
It is a theorem in the elements of the Theory of Functions
that a transformation of the complex plane z' =f(z), where f(z)
is any analytic function of z,
preserves both the magnitudes and
signs of angles, except at the points where f(z) has singularities.*
Laying aside the two points z
= 00 and z= die, the latter of
which gives :' = QO since we have not defined the angle between
,

linesmeeting at oo we can state for reference the following

,

THEOREM 6. The linear transformation is directly con~

formal.
It a theorem of the Theory of Functions, whose proof is
is

beyond the scope of this tract, that the most general directly con-
formal transformation of the 0-plane is of the form z' =f(z), where
f(z) is an analytic function of z."\ In general the correspond-
ence of the points of the 0-plane, and these of the 0'-plane is not
one-to-one. For example z Jz gives two values of z for each
value of z\ Z=\T\Z gives infinitely many values of z for each
value of z' . The question naturally arises whether the linear
trans Eon nation the most general one possessing the quality of
is

uniqueness mentioned in Theorem 2. This is in fact the case.

For let z f(z) be the most general function yielding a one-
to-one transformation. Let 3 = 9 be the point for which z'=cc ;

f(z) then has a pole at q and nowhere else. The function can
be written in the form

A,,, A,.,_ A,
m~ l
(z -</)'" (z-q) z-q
*
See Forsyth, Theory of Functions, sec. 9. f Ibid., sec. 253.
 ..MOIM'HIC FUNCTIONS [CH. I

wher. 'rder of the pole. Clearing of fractions we

f

nation of the m-th order in z. To each value of z

values of :. Hence m must equal 1. Therefore

s' = Jh_ + An .
"'7
,
which is of the form (1). We
-<f q:

ha\. ilished the following

TiiF.ouKM The most general one-to-one directly con-
7.

is a linear
iformation of the plane into itself
on.

4. Inversion in a Circle. There is an intimate relation, as

hall see presently, between the linear transformation of
variable and the geometrical transformation of the
"
known
inversion in a circle."
as
Consider a circle with centre at and of radius r. Inversion
in th- ircle is the transformation resulting from replacing

b P of the plane by a point P' lying on the line OP and

,r-e from O satisfies the equation OP'.OP = r2 . P

y
ami \ with respect to the circle. If 0,
are said to be 'inverse
f f
1'. 1' have the coordinates (/, g), (x, y), (x y ) respectively, the ,

tions of the transformation are

,
-.
'

If tli- 'lit re moves to infinity, the circle becomes a straight

Tin- point P' approaches a position such that the line PP'
pendicular to and is bisected by the given line. That is,
iiiv. rsion in a straight line is a mere reflection in the line.
\\". state li'-re for reference some well-known
properties of
inversions:
s
!
are preserved in magnitude but changed in sign;
an-fnrniatinn inversely conformal.
is

are transformed into circles, the straight line

pecial case of the circle.

.nts inverse with respect to a circle C are trans-
>

.ints invrrs." with respect to the transformed

'II inversions and linear transformations

Tin- inversion is a one-to-one transfor-
plam- into its. -11' which preserves the magnitude
3,4] LINEAR TRANSFORMATIONS

of the angle but changes its sign. If now two inversions, or

any even number, be made in succession, the result is a one-to-

one transformation of the plane which preserves both the
magnitudes and the signs of the original angles. According to
Theorem 7, such a transformation is a linear transformation.
Hence
THEOREM 8. The successive performance of an even number
of inversions is equivalent to a linear transformation of the

plane into itself.

The converse of this theorem will now be proved. The
transformation

can be broken up into the succession of the following transfor-

mations :

(a) z = z + a,
l
where a = d/c ,

(6) z2 = e*% ,
where c 2 /(ad - be)
= Ae* ,
A> ,

(c)
z = Az
s 2 ,

(d) -1/z,,
r^= = ale
(e) z =z 4 + fi ,
where (3 .

That these are equivalent to the given transformation is

verified by eliminating z v z^ z3 z. , We shall show that each
of the five transformations is equivalent to an even number of
inversions, which proves that
the given general linear trans-
formation is equivalent to an
even number of inversions. (X
(c) Putting z = x + iy,
zl = x + iy v
1
a =a 4- ia", the
transformation, on breaking
into real and imaginary
parts, is
xl = x + a' , yl = y + a".
This is a translation. Let D
Oa (fig. 1) be the line joining B
the origin to the point a.

Each point of the plane is FIG. i.

translated parallel to the line

Oa a distance equal to the length Oa. Let AB, CD be two
lines perpendicular to Oa and at a distance iOa apart. A re-
 A r'I< ).M< mi'HIC FUNCTIONS [CH. I

is in figure) followed by a reflection in CD is

C'|iii\
tin- translation under consideration.* Thus (a) is
'.jui
tw<> inversions.
' in the form cos# + i sin 0, we get in this
ring
ting n-ul and imaginary parts of the two members of
(nation.
.'.,
= jr r

l
cos & -y
l
sin 6 , y c)
= x^ sin +y 1
cos ,

.' ion about the origin through an angle 6.

A A and B'B be two lines through the origin such that
the angle AOB is J0 (fig. 2).
We verify at once that a re-
flection in A'A followed by a
reflection in B'B is equivalent
to the rotation in question.
This case is thus reducible to
a pair of inversions.
(c) This is the case of ex-

pansion from the origin

It is equivalent to the two

FIG. 2..
inversions: (1) an inversion
in a circle centre at the origin
and radius r; followed by (2) an inversion in a circle centre at
iirin and radius rJA. For if P', P" are the successive
-
of

This is
.Mjuivalent to an inversion in the unit circle with
re at tin* origin
x

lint, in this and succeeding cases it is sufficient to verify that three

by the even number of inversions and by the linear
follow! from Theorem ;'..

" nut.- that the points <>n AT. are unchanged by a reflection
Oa a distance equal to the length Oo by a reflection
4, 5] LINEAR TRANSFORMATIONS

followed by a reflection in the imaginary axis

(e) This, like (a), is a translation.

Combining these results, we get the converse of the preceding
theorem
THEOREM 9. Any linear transformation is equivalent to
the successive performance of an even number of inversions.
It is evident that the representation of a linear transformation
as a succession of inversions is not unique ;
that the transforma-
tion can, in fact, be broken up into an even number of inversions
in infinitely many ways.
The properties of inversions furnish us with valuable in-
formation concerning linear transformations. Thus (/3) and (y)
preceding yield the following important theorem
THEOREM 10. When the complex plane is subjected to a
linear transformation,
(a) Circles are transformed into circles ;

(6) // two points are inverse with respect to a circle, the trans-
formed points are inverse with respect to the transformed circle.
Since three points determine a circle, a given circle can be
transformed into a given circle by setting up a transformation

carrying three distinct points of the first circle into three

distinct points of the second circle (Theorem 5). These points
can be chosen in infinitely many ways, Hence
THEOREM 11. There exist infinitely many linear trans-
formations which transform a given circle into a second given
circle.
COROLLARY. There exist infinitely many linear transforma-
tions which transform a given circle into itself.

5. The Transformed Elements of Length and of Area. In

the transformation
z = (az + b)/(cz + d)

let z be given an increment dz\ the transformed point z is

given an increment dz' = (dz'/dz)dz. The length of dz is \dz\,
the absolute value or modulus of dz and the length of dz is
;

|dz'/dz \\dz\. A small rectangle at the point z of sides dz l and

dz z has the area \dz$zz Since angles are preserved, the
\.
 10 AUTOMORPHIC FUNCTIONS [CH. I

r
transformed area is a rectangle at z . Its area is \dz^dz^ ,
or
. Now
dz ad - !><

and if the determinant of the transformation, ad be, is unity,

we can state
THEOREM 12. If the determinant of the transformation is 1,
elements of length in the neighbourhood of a point z are multi-
~2
plied on transformation by cz-fd and elements of area are ;

i, i H ',
t
>1 ',<(!
by |cz + d|~ 4 .

6. Types of Linear Transformations. We have already

noted a separation of linear transformations into two classes.
Laying aside the identical transformation z = z, a transformation
r

may have either two fixed points or one fixed point. The
number of fixed points and the behaviour of the transformation
with reference to its fixed points furnish a useful basis of
classification. We turn first to the larger class with two
fixed points.
If in Equation (11) we put ^ and f2 the fixed ,
points, for
:,
and zy the transformation becomes

?l^|i
= K^|i .
(12)

where K = (s - 3 2 )(s 3
'
- ^)/(z3 - ^)(%' - f 2 ).
K is called the multi-
plier of the transformation. depends apparently upon the
It
particular point 3 chosen, but such is not the case; if for
.' we put any other
pair of corresponding points, we see by
Iv nation (6) that
|
is K
unchanged.* The character of is used K
to classify the transformations with two fixed and we
points ;

shall so(i
presently that it determines the behaviour of the trans-
formation with reference to the fixed
points.
will
It
simplify the problem to transform the plane in which
z and -/ are n-pivscnted by a linear transformation. Let

Z = r4> z '=^4l 03)

f=l
I'onn.vJun in the form (1), Section 1, we note that when
K = (a - -
il
ties, ci)j(a c 2 ), & and 2 having the values
Ql (9).
5, 6] LINEAR TRANSFORMATIONS 11

The transformation becomes

Z' = KZ (14)

in which the fixed points ^2 are now shifted to

,
and oo .

Having studied this latter transformation, the results can be

carried back by the inverse of (13) to the plane of z, z'.

Writing K in terms of its modulus A (>0) and its amplitude

ft so
K = Ae',
we distinguish the following- cases :

(a)The hyperbolic* transformation, = A. K

=
The transformation 71 AZ is the expansion from the origin
studied in Section (c). We observe at once the following facts
4>

concerning a straight line through the origin (that is, a

it: (1)

circle through the fixed points 0, oo) is transformed into itself;

(2) any circle with centre at the origin (and hence orthogonal to
the family of fixed lines) is transformed into some other circle
with its centre at the origin.
Let the ZZ'-plane be now transformed back into the zz'-plane
by means of Equations (13). The points
oo become and _

and 2 respectively. The straight lines through the origin

become circles through the points ^ and 2 and the circles with ;

centre at the origin become, since angles are preserved, circles

orthogonal to the family through and ^ .

In the hyperbolic transformation then (1) any circle through

the fixed points is transformed into itself (2) any circle ortho- ;

gonal to the family of fixed circles is transformed into some other

circle orthogonal to the fixed circles. We note also that, since
and oo are inverse points with respect to any circle with
centre at the origin, the points and f2 are, applying Theorem
10 (6), inverse points with respect to any circle orthogonal to the
fixed circles.

Fig. 3 shows the two families of circles just mentioned. The

way in which a region is transformed is indicated, each shaded
region being transformed into the next in the direction of the
arrow.
*
The names elliptic, hyperbolic, parabolic, were first used in this connection by
F. Klein in 1878. They were suggested by the connection of substitutions of these

types with displacements in absolute geometry of the elliptic, hyperbolic, and

parabolic type.
 AUTOMORPHIC FUNCTIONS [CH. I

(//) The transformation,

elliptic
=Q K i6
.

=
The transformation Z' e Z is the rotation about the origin
ie

discussed in Section 4 (6). The circles of the preceding case play

a part here, but their roles are changed. The circle with centre
at the origin is transformed into itself a line through the origin ;

i^ transformed into another line

through the origin, making an
an^le with the given line; the points and co are inverse
points with respect to the fixed circles.
Transferring these results to the sz'-plane :
(1) a circle through

FIG. 3.

th- fixed points transformed into a circle through the fixed

is

points intersecting the given circle at an angle 0', (2) the circles
orthogonal to the circles through the fixed points are the fixed
<-i
ivies; (3) the fixed points are inverse points with respect to
arli of the fixed circles.
Tin: behaviour of the transformation is shown in fig. 4.
'J=.Ivr, and only one of the fixed points is shown. Each
^h.-idi-cl
region is transformed into the next in the direction of
tin- arrow.

'/'I u> loxodromic

transformation,
* = Ae X i6
.

This r.msi'onnatioii. /;' =

t
.V%
is a combination of the
preced-
'
Tin- ti-riu l<).i'n,li'omic was introduced by F. Klein in 1882.
 LINEAR TRANSFORMATIONS 13

ing ones. It involves both a stretching from the origin and a

rotation about the origin. In general there are no fixed circles ;

but in the particular case 6 = 7r, a stretching from the origin

followed by a rotation through two right angles, each line
through the origin is transformed into itself, the half of the line
on the one side of the origin being transformed into the half on
the other side.
In the 3^-plane, then, the loxodromic transformation has
in =
general no fixed circles. In the particular case O TT each
circle through the fixed points is transformed into itself, the two

Fir,. 4.

parts into which the circumference is divided by the fixed points

being interchanged.
We note an essential difference between this particular case
and the two preceding types with reference to the fixed circles.
In the elliptic and hyperbolic transformations the interior of the
fixed circle is transformed into itself in the loxodromic trans-
;

formations in which there are fixed circles, the interior of the

fixed circle is transformed into its exterior.

(d) The parabolic transformation, one fixed point.

When determining the fixed points we found that if
(a = the transformation has but one fixed point,
elf 4- 4}bc
namely =(a d)/2c. In order to express the transformation in
terms of the fixed point we note that the points z= oo d/c are ,

transformed into z' = a/c, GO.

Setting up the unique transf orma-
14 AUTOMORPHIC FUNCTIONS [CH. I

- we can write the trans-

tion carrying oo , d/c into a/c, oo ,

formation in the form

- - -g -$ ,
z'
a/c d/c a/c d/c _

tituting for (
its value, this becomes

1 1
T + p, where p=a . .
^
1
J
z z- +a

Transforming the plane in which 0, z are represented as follows

the transformation becomes

Z' = Z + fi ..... (17)

the fixed point now being at infinity. This is a translation of

the plane parallel to the line 0/3 joining the origin to the point /3.

Fm. 5.

Any line parallel to O/3 is transformed into itself; any other

hi liui- La transformed into a parallel straight line. Trans-
forming hack to tin-
:;'-phiiK;, the point oo becomes the point f.
I'aralN-l become circles
straight lines, intersecting only at oo ,

ecting only at and hence tangent at f.

Hence, in the
].a!,-il)olic transformation a one parameter family of
there is

nt circles tlm.ugh UK- fixed point each of which^is trans-

6] LINEAR TRANSFORMATIONS 15

formed into *
itself ; through the fixed point is
any other circle
transformed into a second tangent to it.
circle

Fig. 5 shows a parabolic transformation in which the fixed

circles are those with a horizontal tangent. The shaded regions
connected by the arrows show how the parts of the plane are
transformed. The circles with a vertical tangent at furnish
an example of a family each circle of which is transformed into
some other circle of the family.
*
One
of the fixed circles is a straight line through |. The fixed line through
0, 0, in the ZZ'-plane is transformed into a circle through oo', |+l/)3, { in the
oo

cz'-plane that is, a straight line through | and f + l/j8. Thus the fixed circles are
;

those circles through whose tangent is parallel to the line joining and 1/0.
 CHAPTER II

GROUPS OF LINEAR TRANSFORMATIONS

f
7. Notation. The linear transformation z =(az+b)/(cz+d) is

frequently denoted by

^}
cz + dl
or (
z
ll,
\z-fc,
K2 -|
z - ft.

the transformation being the process of substituting the second

term in the parentheses for the first. Also it will often be con-
venient, when not likely to lead to confusion, to represent the
transformation merely by a capital letter, as S, T, U.
If a transformation S be made, and then to the transformed

plane a second transformation T be applied, the single linear

transformation equivalent to the succession of the two will be
represented by T(S), or simply TS. ST will be different in
general from TS. Thus if S is the transformation above and T
is the transformation sf = (aZ+/3)/(yz+8), we have, making use
of Equation (5), Section 1,

(aa + b y )z + ap_b8\

which are different in general. Linear transformations are thus

not commutative.
T2 T3, , etc., are the transformations equivalent to
performing
T twice, thrice, etc. T" 1 will represent the inverse of T, and
T~ n = (T~ 1 ) w . The identical transformation z'=-z is represented
by 1. The fact that a transformation followed by its inverse
l.-avcs the points of the plane
unchanged gives the equation
T rr = l.
Combinations of T and its inverse T" 1 obey the
ordinary laws of the addition of exponents in multiplication.
In combining transformations it is
easily seen that the associative
law holds l.e. that S(TU) = ST(U). No confusion then can arise
i

writing STU.
i.e
7-9] GROUPS OF LINEAR TRANSFORMATIONS 17

To find the inverse of ST, we make on

the plane transformed
by ST first the transformation S' 1 then T' 1 T~ 1 S- 1 (ST) =
,
:

T- (S- 1 S)T = T~ T = 1.
1 1
Thus T^S' is the transformation which,
1

applied after ST, leaves the points unchanged ;

that is,

T^S- 1
. In general the inverse of ST .' . . UV is

8. Definition of a Group. A set of transformations, finite

or infinite in number, is said to form a group if
(1) the succession of any two transformations whatever of
the set is a transformation of the set ;

(2) the inverse of every transformation of the set is a trans-

formation of the set
The definition applies to all kinds of transformations, but we
shall be concerned only with groups of linear transformations.
The group properties, using the notation of the last section, are
these (1) If S is a transformation of the group (not necessarily
:

different from T), ST belongs to the group and (2) if T is. a ;

transformation of the group, T" 1 also belongs to the group.

Since T and T' 1 belong to the group, T~ 1 T(-1) does also;
hence, every group contains the identical transformation z' = z.

9. Examples Of Groups. We give here several examples of groups.

Group of Real Tr<>/is?"/ionx. All transformations of the form z' = z + k, where
(1)
k has any real value, constitute a group. For if T is z' = z7t\, T" is z' = z-Jc 1
l ,

which belongs to the set. If S is z' = z + Jc 2i ST is z' = z + k} +k.2t which belongs to

the set. If k be restricted to positive values, we do not have a group for T~ x does ;

not then belong to the set.

(2) Group of Rotations about the Oriyin. The transformations z' = e'9;., where
6 is any real quantity, are easily shown, after the manner of the preceding case, to
satisfy both group properties.
(3) Group of the Siinplii Periodic Functions. The set z' = z + mw, where w is

constant and m
a positive or negative integer or zero, forms a group.
is

(4) Group of the Doubly Periodic Functions. The set z' z + mot+m'o' } where =
ca, ca' are constants and m
and m' are positive or negative integers or zero, forms a
group. It is assumed that the ratio o>/o>' is not real ; although this is not necessary
to establish the group properties.

(5) Cyclic Group. The m transformations z'sseZjrtn/mgj n = Q, 1, m-l, . . .

form a group. They are the successive rotations about the origin through an

angle

*
The order of procedure in a combination should be carefully noted. Thus
STU is U, followed by T, followed by S. If S is z'=/3 (), T is s'=/2 (z), and U
is s' = /i(), the combination STU is z' =/3 {f2[fl (z)]}.
2
18 AUTOMORPHIC FUNCTIONS [CH. II

Ratios. The six transformations

(ti) Group of the AnJiarmonic
-'--
- ~~ " '
!
~
c
'
i- -L '
i
i ..
' -izl,
~
.
JL.
-
..
i
i

We that both group pro-

form a group. verify by combining the transformations
perties are satisfied.
This group is so named for the reason'that if z is any one of
ratios are given by
tin- anliarmonic ratios of four points on a line, the six anharmonic

the transformations of the group.

10. Transforming a Group. From any given group of

linear transformations infinitely many other groups can be
derived by applying a linear transformation to the plane in
which s and its transformed values are represented. Let T be
a transformation of the original group, and let T carry s into z'
Let a transformation G be applied to the plane, and z' becoming
z l and 0/ respectively. The transformation T x carrying z l to z^
is a linear transformation equivalent to (1) that carrying
zl :

to z followed that z to z' followed by (3) that

(2) by
carrying
=
carrying z to ^. Hence T\ GTG'
1
.

Now, let all the transformations of the given group be trans-

formed in this manner. We shall show that the new set of
transformations form a group.
Thus, ^-^(GTG-^-^GT-KJ- which belongs to the set
1
,

since T" 1 belongs to the original group, and the second group

property is satisfied. And if S 1 = GSG~ 1 where S belongs . to

the given group, S^ = GSG^GTG' = GSTG' 1 1
,
and this belongs
to the set since ST
a transformation of the given group, and
is

the first group property is satisfied. Therefore, the transformed

set forms a group.
\\V si all
i find that it will frequently facilitate the study of a
>iip
to transform All configurations in the
it in this manner.
lane can then be carried back to the s-plane by the trans-
formation G" 1 .

11. Continuous and Discontinuous Groups. Let z be an

arbitrary point in the plane, and let us apply to this point all
tin-
bransformatipns of some given group. Considering the
i i<.ns ni' lie
i
mints derived from z by these transformations,
]

.ill arise: fidicr all the transformed

points will lie
on' hlicicntly small circle with as its centre (excluding,
z
bhe ioVntical transformation which leaves the position
MiH-liii' i
there will be infinitely many of the trans-
9-11] GROUPS OF LINEAR TRANSFORMATIONS 19

formed points clustering about z. In the former case the group

i-said to be discontinuous] in the latter, continuous. Thus, in
the group of the simply periodic functions (No. (3) of Section 9),

given any finite point z, there is no transformed point within

a circle of radius eo drawn about z. The group is then discon-
tinuous. Similarly the group of the doubly periodic functions
(No. (4) of Section 9) is discontinuous. In the group of real
translations (No. (1) of Section 9), on the other hand, we can,
by taking k sufficiently small, obtain a transformed point as near
as we like to the given point. This group is continuous there-
fore. A group with only a finite number of transformations, as
Nos. (5) and (6) of the same section, is necessarily discontinuous.
We lay down the following
DEFINITION. If there exist transformations of the group,
transformation excluded, which change the posi-
the identical
tion of the z-plane by an arbitrarily
an arbitrary point of
small amount, the group is continuous. Otherwise it is dis-
ronfinuous.
If the transformed point be represented by z + Sz, the trans-
formation can be written

z + 8z = (az + b)l(cz + d) ;

2
whence, cz +(da)z b+(cz+d)8v=Q. If the group be con-
tinuous, there will exist transformations, whatever be the point
z, such that o: is arbitrarily small. Hence, there must exist
transformations in which c, d a, b are arbitrarily small quantities
(but not all zero, for that is but the identical transformation).
In order that the group be continuous it is not necessary
that the coefficients of the transformation should vary con-
tinuously asis the case in Nos. (1) and (2). Thus consider the
group z = e where n is a positive or negative integer or zero,
ine
z,

and where incommensurable with TT. This is the group of

is

rotations through positive and negative multiples of the angle 0.

By rotating through the angle a sufficient number of times
any point can be brought back as near as desired to its original
position hence, the group is continuous.
;

Continuous groups have been made the subject of many

investigations, particularly by Lie, and have yielded important
results in the theory of differential equations. However, for a
reason that will appear later (Section 20, footnote), such groups
 AUTOMORPHIC FUNCTIONS [CH. II

no part in the theory to which this tract is devoted; and

not have occasion to consider them further.
iiall

12. The Modular Group. In this section we make a detailed

>tudy well-known group. The properties of this group

of a

uplify many theorems to be subsequently established; and

the study of the group at this point furnishes a concrete back-
ground for future developments of the theory.
Consider the transformations 2f = (az + b)/(cz + d) in which
a, ^. c, d are positive or negative integers or zero subject to the
c< mdition that ad be = 1. The inverse of such a transformation
the coefficients d, b, c, a (Eqn. (3), Section 1), which are
also positive or negative integers or zero and with determinant
</>/ bc = l. If z
/
=
(az+/3)/(yz 8) is a second transformation +
whose coefficients a, /3, y, 8 are of the desired form ;
then the
transformation followed by the second is a transformation
first

with the coefficients aa + /3c, ab + /3d, ya + Sc, yb + Sd (Eqn. (5),

Section 1). These coefficients are integers or zero, and the
determinant is (ad bc)(a3 /3y)
=1 . Both group properties are
therefore satisfied.
All transformations (az
(z, +
b)/(cz -f d)) in which a, b, c, d are
integers a/nd ad bc=l constitute a group called the modular

group.
Let us consider the two transformations S (z, 1/0) and :

T (z,z + I), which belong to the modular group. It will be shown

:

in a moment that any transformation of the modular group

formed by combinations of S and T with their inverses.

',<

\\V v.-rifythatS-^S or S 2 = l; T is (z, z + n) T- is (z, z-n). ;

\V<- shall now indicate a method of

decomposing a trans-
formation of the modular group into combinations of S and T,
taking first a particular example:

(4x25)-(9xll) = l.

The method to be followed is to reduce the coefficients in the denominator by

mbiiiutions of S and T. Since replace z by z + n, n being chosen
25>11 we
that 25 shall be ivplaee<l by as small a number as possible. Put c-2 for z.

UT, = (5, (45 + l)/(llc + 3)), where T, = (5, 5-2) or T^T" 2

.

of fraeth/ns, the first coefficient in the denominator

.Her.

*(, (-4/5 + 1 )/(-!]

11-13] GROUPS OF LINEAR TRANSFORMATIONS 21

Continuing,
UT 1 ST, = (.:, ::/(3:4-l)), where T, = T 4 = (~, :. + 4).
UTjSTaS =(,-!/(* -3))
UT ST 2 ST = (c,-l^),
1 3 where T 8 = T 3 = (;, ~
+ 3)
UT1 ST2ST8S=1.
From this last equation U is the inverse of TiSTsSTgS. Hence

U = S-iTg^S^To^S- 1 1
! !- 1 = ST- 3 ST- 4 ST-.
U is thus expressed in terms of S and T.

The method followed in this particular case furnishes material

for a complete proof for the most general case. When \d\ >\c\
a transformation of the form (':, :d=) reduces the transformation
to r which \d\<\c\. Putting
in I/: for we interchange 2;,

c and can reduce again. The determinant at each step is

j

.ity, since the transformations which are combined have unit

determinants. So long as c 4= 1 the substitution (z, zn) will
| |

not reduce d to zero, for the condition ad be = I requires that

c and d be prime to one another. We eventually reduce the
denominator to unity, and ad be = a = 1 and the transformation
is of the form (z, zn), which by means of (:, C=FTI) becomes

equal to the identity. We then express the given transforma-

tion in terms of the transformations used in reducing it as in
the example. Hence, any transformation of the modular group
can be expressed in combinations of S and T.
It is clear that S and T are not the only transformations that
can be regarded as generating transformations of the group.
Any pair in terms of which S and T can be expressed will serve.
All transformations, being expressible as combinations of S and
T. will then be expressible in terms of the new pair. Thus we
mi o-ht have used U = 0, - I/z) and V = (c, - 1/0 + 1)); since S = U,
T=UV.
13. The Fundamental Region. Before proceeding further
with the study of the modular group it is desirable to introduce
the definition of the fundamental region, a conception of the
foremost importance in the theory of discontinuous groups.
DEFINITION. 7 \vo points, two curves, or two areas are said
to be. roiKjriu'id n'ith respect to a given group if one can be
f/'
iixfurmed into the other by some transformation of the group.
DEFINITION. A region such that no two of its points are
congruent with respect to a given group and such tJtat every
2-2 AUTOMORPHIC FUNCTIONS [CH. II

ii
adjacent to it contains points congruent to points in the
n region is Billed a fundamental region for the group.
Fig. 6 shows a fundamental region for the
Zniir
cyclic group of six rotations (z, e ^z}. No
two points in the shaded region can be carried
into one another by rotations through multiples
of the angle 2ir/6 but points in any adjacent
;

region (and in this case any other point in the

plane) can be carried into some point in the
region by such a rotation.
The shaded area in fig. 7 is a fundamental
region, or "parallelogram of periods, "for the group
of the doubly periodic functions (, z + mu
+ m'u').
The vertices of the parallelogram are the points
0, a;, 00' ,
co + (a'.
FIG. 6. We note that the rays bounding the region
in fig. 6 are composed of congruent points.

Strictly speaking, only one of these rays should be considered as belonging to the
region. A similar remark applies to fig. 7.
No point can have two congruent points in the fundamental
region ;
for these two points, being congruent to the same point,
would be congruent, which is

contrary to hypothesis.
It is clear that the funda-
mental region is not uniquely
determined. We can, for ex-
ample, add to the given region
a small adjacent region which
has no two of its points con-
gruent, provided we remove
from the original region the
part congruent to the added
n-^inn. The new region is then
fundamental. We note also
that any region congruent to
tin? fundamental region will
also serve as a fundamental
FIG. 7.
ion. For example, the un-
shaded ivo-ions n fig s> Q ^nd 7 are j
congruent to the shaded
lona and any one could be
employed as a fundamental region.

14. Fundamental Region for the Modular Group. Tin*

^formation T=*(z, :+l) is represented geometrically by a
13, 14] GROUPS OF LIXEAR TRANSFORMATIONS 23

translation in the positive direction parallel to the real axis

through a unit distance. The transformation S = (z, 1/z) breaks
up into the real and imaginary parts

This is equivalent to

V+ tfo r-o+Vo

that is, a reflection in the imaginary axis followed by an

FIG

inversion in the circle whose centre is at the origin

O and whose
radius is unity.
Consider the region lying outside the unit circle # 2 + 2/ 2 = l
and between the lines # = -0. This is the shaded region in
fig. 8. The translation T carries this region into a congruent
one abutting along the boundary x=\. The translation T" 1
carries it into a congruent region abutting along the boundary
x i. The transformation S carries it into a congruent region
abutting along the circular portion of the boundary. Hence
the given region satisfies the second part of the definition of a
fundamental region .
 AUTOMORPHIC FUNCTIONS [CH. II

It remains to show that there are no two points of the

fundamental region congruent. To do this it will be proved that
the transform of a point of the region lies outside the region by
i-y
transformation of the group except the identical transfor-
mation. It is necessary to consider three cases, according to the
value of c in the transformation (z, (az + b)/(cz + d)) :

(1) c = 0. The determinant ad= I gives a

= d= 1.

The transformation then a translation (z, z-\-n) where n is a

is

positive or negative integer. This translation carries a point of

the region into a position outside the region.
If c 4= 0, we can write the transformation

-a/c =
z - + d).
l/c(cz

(2) c= 1. z' + a|=l/ zd\.

Now 1
zd |
is the distance from z to the point ^d, and z'^a is
|
\

the distance from z' to a, a and d being integers. The point

z in the given region lies outside the unit circles drawn about
the integral points on the real axis. Then, \zd\>l. Hence
|
z'^fa |
< 1, and z' lies inside the unit circle about a. It is thus
outside the given region.

K-a/c| =
2
(3) |c|^2. l/c |

The points and d/c lie on the real axis. The points of
a/c
the given region lie above the line
y = %, then z + d/c\>^.
\z'
z
a/c\<2/c =%, since c
2
^ 4. The point z' is below y = }$,
and is therefore outside the given region.
Hence when z lies in the given region no transformation of
the group carries it into some other point of the region. The
given region then contains no two congruent points and is
tlx-ret'ore a fundamental
region. Since we have proved that
tin- modular
i^roup has a fundamental region, we have established
tip fact that it is a discontinuous
group.
\V<- found that the transformations, T, T~\ S, carry the
fundaim-ntal H- ion
into congruent regions adjacent to it. The
isi'onnalion carrying the fundamental into an
region adjacent
"ii I! carries tin- three regions surrounding it into three

surrounding I!. There are transformations of the group

Yin- tin- iiindaim-ntal region into these regions adjoining R.
14. 15] GROUPS OF LINEAR TRANSFORMATIONS 25

There are transformations of the group carrying the funda-

mental region into regions adjacent to the new regions, and so
on. Continuing this process, we cover the whole upper half-
plane with regions congruent to the fundamental region. This
partition of the plane is shown in fig. 8.
We get a similar division of the lower half-plane by a
reflection in the real axis. The fundamental region and its
reflected region are inverses in the real axis and by Theorem ;

10(6) the various transforms of the region in the lower half-

plane will be the reflection of the corresponding regions in the
upper half -plane.
Infinitely many regions cluster along the real axis. The
points on this axis are for this reason called the points of
nitinuity of the group.

15. Fuchsian Groups, or Groups with Principal Circle.

The coefficients of the modular group being real, any real point
is transformed into a real point by a transformation of the
group. The real axis is then & fixed circle for each transforma-
tion of the group. This group is a member of an important
class of groups having the property designated and called
Fuchsian groups.*
DEFINITIONS. // all the transformations of a discontinuous
nip leave one and the same circle invariant, and if points in
<'/)'<

tli* interior of the fixed circle are transformed into points in

1

tJte interior by each of the transformations, the group is called

a Fuchsian group.
The fixed circle is called the principal circle of the group.
A reference to the article on Types of Linear Transformations
and to tigs. 3, 4, and 5, where the fixed circles are drawn, shows
that the transformations of a Fuchsian group are :

(1) hyperbolic transformations with the fixed points on the

principal circle ;

(2) elliptic transformations with the fixed points inverse to

one another in the principal circle ;

(3) parabolic transformations with the fixed point on the

principal circle.
*
This term was introduced by Poincare, Comptes rcndus, xcii. (1881), p. 333.
It is not very appropriate, but will lie retained here.
 AUTOMORPHIC FUNCTIONS [CH. II

xodromic transformation can belong to the group for the ;

only loxodromic transformations possessing fixed circles trans-

form the interior of each of the circles into its exterior.
It will simplify matters to transform a given Fuchsian

group, in the manner described in Section 10, by a transformation

which carries the fixed circle into the real axis. Any Fuchsian
^roiip can be thus transformed into one with the real axis as
principal circle, the interior and exterior of the fixed circle be-
coming the two halves of the plane separated by the real axis.
The real axis being the principal circle, any real point is trans-
formed into a real point. If z
f
=
(az -\- b)/(cz d) is a transforma- +
tion of the group, a, 6, c, d are real quantities, or at most contain
a common complex factor which can be removed by division.
The requirement that the interior of the fixed circle should
be transformed into itself (and, of course, the exterior into itself
also) gives a further condition on the constants. Breaking the
transformation into its real and imaginary parts :

Id -
,
= ,
=
~ (ad bc)y
(CX + d)1 + <?l?
y
(CX + df + 6'
V '

we see that the necessary and sufficient condition that y' have
the same sign as y, and hence z' and z be on the same side
of the real axis, is that ad bc>0. Dividing numerator and
denominator by the real quantity J(ad bc), the determinant
becomes +1, the coefficients remaining real.*

16. Fundamental Region for a Fuchsian Group. Consider

a Fuchsian group with the real axis as principal circle. Let a
be a point of the plane at which there are no infinitesimal
transformations. Since the group is assumed to be discontinuous,
Mich a point exists. We shall suppose a to be in the upper halt'-
plam-. None of the points congruent to a lies in its neighbour-
region R about a all of whose
hood. Tln-n- will exist a small

points an- transformed into points outside R by the transforma-

inted "lit in the early pages of the book that the deti-rminant can
!*' -di-ivd -l-l. If, however, ad-bc<0, the <h-trrminant becomes +1
<MI!V division - an imaginary quantity, and the
..ii
by J(ad ir), coefficients no longer

d transformations with ii<"_Mtive determinants are the

TT mentioned in Section 6 (c), which carry the
rior.
15, 16] GROUPS OF LINEAR TRANSFORMATIONS

tions of the group, the identical transformation excluded. That

is, if R is sufficiently small, no two of its points are congruent.
Now let R be enlarged by the addition of adjacent regions so

long as this can be done without adding points congruent to each

other or to points already in the region. When R is enlarged

as muchas possible in this way, we have a fundamental region

for the group for no two points of the region are congruent, and
;

no adjacent region can be added without introducing points con-

gruent to points already in the region.
Let b be a point on the boundary of this region (see fig. 9).
There is a transformation of the group carrying b and a portion

FIG. 9.

of the boundary on one or both sides of b into some boundary

point b' and a portion of the boundary near b'. If this were
not the case, we could add to the given region a small adjacent
region at b without introducing points congruent to any in
the fundamental region which is contrary to hypothesis. Call
this transformation S. Let AB be the part of the boundary
containing 6 which is transformed by S into CD
containing b'.
The fundamental region is carried by S into an adjacent
region, CD being the common boundary.
1
Similarly S" carries
the fundamental region into an -adjacent region, AB being the
common boundary.
Considering a point c not on AB or CD, we get a portion EF
of the boundary containing c as interior or end point congruent
28 A FT< >.M ORPHIC FUNCTIONS [CH. II

to another portion HG by a transformation T of the group.

Furt li.-r, AB, CD, EF, HG contain no parts in common; for a
point lying just outside the region near a common portion of
tin- boundary would have two congruent points inside the
.Inch is impossible. Continuing the above process, we
conclude that the fundamental region is bounded by a fir
or infinite number of pairs of congruent curves.
Almost without exception the Fuchsian groups that have
been investigated are those in which the fundamental region is
bounded by a finite number of pairs of congruent sides and it ;

is with this case only that we shall be concerned.

The fundamental region may lie entirely above the real axis,
as in the case of the modular group or it may lie on both sides
;

of the real axis and contain a portion of the axis. In the latter
case we need consider only that portion lying above the axis,
for the remaining portion can be derived from this by a re-
flection. But it should be borne in mind that both portions
belong to the fundamental region.
Concerning the nature of the boundary of the above funda-
mental region we know nothing. We propose now to simplify
the region by replacing the boundary by another whose nature
we know. Let a circle with its centre on the &-axis be drawn
through A and B (fig. 9). This circle is perpendicular to the
axis. It is transformed by S into the circle through C and D
with its centre on the ^-axis. [The circle through A, B is ortho-
gonal to the -axis. Since angles are preserved and the cc-axis
is transformed into itself, this circle is transformed into a circle

orthogonal to the #-axis which we know passes through C, D.]

\V.- can
replace the former boundaries A6B and C6'D by these two
jnient circles; for any portion of the original region removed
near one boundary is replaced by a congruent region near the
other boundary. The modified region is a fundamental region.
Proceeding in a similar manner for the other parts of the
boundary, we gut the entire boundary composed of arcs of
circles, llmcc. I he fundamental region can be chosen UK a
circular polygon, the bounding circles having their centres on
I/;/ arranged in congruent pair*.
f l" i"l f " i

\V- shall sprak of the transformations of: the

group by which
tin- of sides are connected as the g, 'inrutnnj
congruent pairs
tramsformatione of tl,r
growj*. It will be proved later that any
16, 17] GROUPS OF LINEAR TRANSFORMATIONS 29

transformation of the group can be expressed as a combination

of these generating transformations.

17. The Cycles. The sides of the fundamental region being

arranged in congruent pairs, to each point on the boundary there
corresponds in general one other point congruent to it and lying
on the boundary. This is not generally true, however, of the

FIG. 10.

vertices of the polygon, each of which lies on two sides. Corre-

sponding to a vertex there may be one congruent vertex, there
may be several, or there may be none.
DEFINITION. Each set of congruent vertices of the funda-
mental polygon is called a cycle.
A. cycle may consist of one vertex

or of several.
In fig.
10 is shown a fundamental
region in which the congruent sides
are connected by arrows, S, T, being U
the generating transformations. Per-
O the transformation S, AB is
forming
A
transformed into ED, E being the
C
FIG. 11.
transform of A. Performing T' 1 EF ,

goes into CB, C being the transform of E. Performing U,

CD goes into AF, A being the transform of C. The vertices
A, E, C are congruent vertices, and constitute a cycle. Starting
with B and making in succession T, U" S" we get the vertices
1
,
1
,

F, D, B. Hence
B, F, D
form a second cycle.
In fig. 11 there are three cycles: AC, B, and D. A vertex
lying on the axis as A
is called a parabolic
point. The two circles
there are tangent. Since the points congruent to a point on the
30 A! IOMORPHIC FUNCTIONS [CH. II

Iso lie on the axis the vertices of a cycle containing a

parabolic point are all parabolic points. One of the vertices of
the fundamental region for the
modular group is a parabolic

point the vertex at infinity in

the shaded region of fig. 8. It
constitutes a cycle.
Fig. 12 is an example of a
fundamental region lying on
both sides of the principal circle.
The region considered is that

lying within the large circle and

without the three small circles.
FIG. 12.
It is the fundamental region for
r's
group mentioned later. This particular region has no
vertices.*
Two useful theorems connected with the cycle will now be
established
THEOREM 13. The sum of the angles in each cycle not con-
t i n Ing parabolic points is a submultiple of ZTT.

Let
1? 2
A A ,
.
.,
An
be the vertices of the cycle. Let T l be the
.

transformation by which A x is transformed into A 2 a side ,

adjacent to A x becoming a side adjacent to A 2 T 2 the trans- ;

formation carrying A 2 and an adjacent side into A 3 and an

adjacent side and so on to T n which carries A n and an adjacent
; ,

sid- into A
and an adjacent side. Each of the transformations
l

>n<* of the
generating transformations or its inverse. Gon-
MI!T the n regions congruent to the fundamental region by
applying to the fundamental region the n transformations of
,roup: T n> Tn T ?i -i, . .
.,
T n ...T 3T 3 ,
TM . . . T2Tr Call
the fundamental region R
and the congruent regions just
iiH-ntioiHMl l!j,
R 2,
. .
.,
R respectively. T n carries A,, and an
yi

adjacent side into A x

and an adjacent side. Hence R x has a
nid adjoins R along one of the sides terminating
angles an- preserved, K x has a vertex of angle Aw f
* T:
at here departs from
in thai of
the papers iu the Poineaiv
early voluiii" I
// ^fu//te//)tl/i><l i
wlu-re the idcii
the cycle first appeared. ..f

.iiTt-d ^ould be Hilly tli;il part above the axis, the four

ling part ..I'tlm boundary. The region tlien has ei^hl vertices
kind are "
"
if tliis
cycles iii 1'oiiiciuv's
..]..-ii
terminology.
:in U
; by A, I lie
alible at the verti \
17] GROUPS OF LINEAR TRANSFORMATIONS 31

at Ar carries A M _! to A 1
The transformation TnT,^ Further, .

R 2 adjoins x along
a side terminating R
at A for T H _j carries R 3 ;

into a region R' adjacent to R at A n-1 and the transformation ,

T,, carries R into R x and R' into R 2 adjacent to it. Continuing,

we get the regions R 3 R,^, R M as shown in the figure, the , . .
.,

vertices at A x having the angles A n _ A 2 A 1 If the angle 2, . .

., ,
.

at the point A x is not yet filled, we continue with the regions

R, i+l ,
R w+2 ,
. . . derived from R by the transformations
TM . . .
T.Tn, Tn . . .
TxTnT,,.!, etc. Now, none of the angles
A 1?
. .
.,
A n is zero, for the bounding sides are circles with
their centres on the jj-axis and no two such circles can be
tangent at a point not lying
on the axis. Hence, by the
above process the angle about
the point is eventually filled.

Now, none of the angles A 2,

. .A can overlap the original
.
, ;4

angle A for the points in these 1 ;

jingles are congruent to points

in the angles A, A,, of
the original region R and any ,

points common to the angle A x

and one of these angles would
have another congruent point in
the fundamental region, which
is
impossible. Hence the angle x
A
of R n R, n or Ryw coincides
, ,
. .
.,

with the angle Aj_ of R . If R, each of the angles appears once,

t,

and the sum is 2-rr ;

if R2n ,
each appears twice, and the sum is
2-2; in general each appears p times, where
p is some integer,
and the sum of the angles is 2?r/p, which was to be proved.
THEOREM 14. To each cycle not containing parabolic points
there corresponds a relation between the
generating transforma-
tions of the group.
We
have just found that when the transformation (T n
T2 T 1 applied to R the transformed region Epn has the angle
;)
) is ,

Aj_ in its original position. The points in this angle must coincide
with their original positions, since otherwise the angle A would
1
contain pairs of congruent points which is impossible since A
:

belongs to the fundamental region. But if three or more points

are unchanged, all points are unchanged, and the transformation
 AUTOMORPHIC FUNCTIONS [CH. II

isthe identity. Hence (T n T n _! T 2 T 1 ) P = 1, a relation between . . .

the gem-rating transformations of the group.

These two theorems can be illustrated by the modular group.
The point C (fig. 8) constitutes a cycle of angle 2vr/2. The sides
at C are connected by the transformation S whence the second ;

theorem gives S 2 =l. A and B make a cycle of angle 2?r/3. A

returns to itself by the transformation T followed by S whence ;

=
(ST) = STSTST 1. This can be verified directly from the'
3

(
i
nations of the transformations.

18. The Transforms of the Fundamental Region. Let

T\, T2 ,
.
.,
Tn
be the generating transformations connecting
.

the n
pairs of congruent sides of the fundamental region,
R .

As has been pointed out, R can be transformed into a region

adjacent to it along any one of its sides by means of one of the
transformations Tp . .
.,
T B T^ 1
, T,r
l
Let be some
,
. . . . T
transformation of the group, let R' beand
the region into which
it transforms R . Let A'B' be any side of the polygon R'.
There is a transformation of the group transforming into a R
,<'</
ion R" adjoining along the arc A'E' For let AB be the R .

side of the polygon which is carried by the transformation R

T' into A'B'. Let T be the generating transformation carrying ft

II, into R* adjacent to along the side AB. Then T' carries R
R fc
into a region R" adjacent to R' along the side A'B'. There-
fore T /r
J.\, which is a transformation of the group, carries R
into R".

Starting with the region R we transform it by means of ,

tin- generating transformations and their inverses into regions

adjoining it along its %n sides. We can then transform R into
ions adjoining these new regions along each of their sides,
ami so on. Unless the group consists of only a finite number of
formations we get in this way infinitely many transforms
.

of the fundamental region. The question now arises How much :

of th<- plane is covered by these repetitions of the fundamental

n ''.

THI:<>KI;M 15. The transforms of the fundamental region

'! tihfme manner (or, if the region lies on both sides of the
/ / lull
part lying above the axis) cover the entire upper
half-plane.
To prove this, it will be shown that any given point Q in the
18] GROUPS OF LINEAR TRANSFORMATIONS 33

upper half-plane can be reached by repetitions of the fundamental

region according to the method indicated in a finite number of
steps. Let P be a point in the fundamental region, and join P, Q
by a straight line. PQ lies entirely above the real axis let ;
K
be its minimum distance from the axis. We shall suppose for
the moment that the fundamental region does not extend to
infinity. Let M
be its maximum distance from the 3>axis. Let
AB be the side of the fundamental region from which the line
PQ issues (fig. 14). Construct the region R : adjacent to R along
the side AB. Let PQ leave Rj by the side A 1 B 1 Construct R 2 . ,

the transform of R adjoining R x along the side A 1 B 1 etc.* ,

Consider C D the portion of the line PQ lying in the region

t t ,

FIG. 14.

R;. Let z=(a z + b )l(c z-\-d )

i i i i
be the transformation by which
R; is derived from R . Then CJ)* is congruent to an arc CD
rot

joining two sides of R . The length of CiD^ is /

\dz'\. By
'

Theorem 12, if 0,^ 6iC f = l, this length is /

-
\ Now,
'C 1 C;0 i \Ju4
|

c {z + d. |
2 = (^x + di) + cfy 2 2
;
and from the equations of Section 15

c&<>
- Therefore dz
=H X len th of

arc CD.
Suppose there are infinitely many segments
Their is greater than K/M times the sum of the lengths
sum
of infinitely many arcs CD crossing R from side to side. Now
this latter sum is infinite. For the only case in which it would
not be infinite is that in which all the arcs CD, save for a finite
number, cross the region near the vertices of the polygon but ;

this would mean that PQ winds round a point at which vertices

meet, which is impossible. The assumption that the number of
segments Cf>i is infinite leads then to the conclusion that the
* In
case PQ crosses a vertex of one of the polygons a slight deformation of the
line will make it cross a side.
3
34 AUTOMORPHIC FUNCTIONS [CH. II

sum of ilu'ir lengths is infinite. But this is impossible since the

sum of their lengths is not greater than PQ, a finite quantity.
1 1. Mice Q can be reached by repetitions of the fundamental region
in :i finite number of steps.
This proof does not hold if the fundamental region extends
to infinity, for in that case we cannot conclude that y, the

imaginary coordinate of a point on CD, is bounded. In that

case consider the congruent regions adjacent to R One or .

more of these adjacent regions will not extend to infinity save

in the one case in which the fundamental region consists of only
two sides with infinity as a parabolic point. This is the case in
which the fundamental region is that between two vertical lines,
the "period strip" of the simply periodic functions, and the
transformations are real translations ; in this case the half -plane
is obviously covered. We lay this case aside. Now, take P in
this new region ;
is bounded, and the preceding proof applies.
y
Since any point Q in the upper half -plane can be reached in a
finite number of steps, it follows that the whole upper half -plane
is covered by these repetitions of the fundamental region.
We note that all the transformations involved in these re-
petitions of the fundamental region have been combinations of
T I}
. . .
., Tn with their inverses. The group has no other trans-
formations. For let T be any transformation of the group. It
carries into a regionR in the upper half-plane. R R must
coincide with one of the regions R* already found for ;
if there
were overlapping, the region common to R and R* would contain
points having two congruent points in R which is impossible. ,

Whence T = T k where T k is the transformation carrying R into

,

RA and is a combination of the generating transformations.

Hence
THEOREM. // TI} Tn are the transformations connect-
. . .

ing the n pairs of congruent sides of the fundamental region

of a Fuchsian group, any transformation of the group can be
expressed as a combination of Tv Tn with their inverses. . .
.,

group has only a finite number of transformations, the

If tin-

half-plane will be covered by a finite number of repetitions of

lip- fundamental
region. If the group has an infinite number of

transformations, the half-plane will be divided into an infinite

numU-r of regions. Since any finite point Q above the axis can
I' n-.-i'-ln-d in a 1'mito number of steps, any finite region lying
18, 19] GROUPS OF LINEAR TRANSFORMATIONS 35

entirely above the axis will contain in its interior only a finite
number of regions or parts of regions. Hence the points about
which infinitely many regions cluster lie on the real axis. These
cluster points may comprise all the points of the axis, as in the
case of the modular group or only certain points of the axis, as
;

in the case of groups in which the fundamental region crosses

the axis.
The division of the lower half -plane into regions can be made
by reflecting the upper half -plane in the real axis. If R lies

entirely above the axis, its reflection will be a fundamental region

R ' lying entirely below the axis. The repetitions of will, byR '

virtue of Theorem 10 (6), be the reflections of the repetitions of

R . fundamental region crosses the axis, the lower half
If the
of the region can be chosen as the reflection of the upper half ;

and the reflection of the whole half-plane supplies the lower

parts of all the transformed regions.

19. Groups without Principal Circles. A discontinuous

group which possesses no principal circle (called Kleinian by
Poincare) exhibits several of the properties already found for
Fuchsian groups. For example, the fundamental region can be
simplified by the use of circular arc boundaries, although now
the bounding circles will not be orthogonal to a fixed circle.
The sum of the angles of a cycle not composed of parabolic points
(i.e. vertices where the adjacent sides are tangent) will be a sub-

multiple of 2?r and each such cycle yields a relation between

;

the generating transformations of the group. And we can

partition the area surrounding the fundamental region by means
of transforms of the region, as in Section 18. It is here that we
come to an essential difference, however the copies of the funda-
mental region do not fill the interior of a circle. In the simpler
cases the whole plane may finite number of
be covered by a
regions (as shown infig. 23), or by an number
infinite clustering
about a point (as in fig. 7), but in general the division is much
more complicated. A common phenomenon is that in which
only a portion of the plane, bounded by an irregular curve, is
covered and several different fundamental regions are necessary
;

to obtain the division of the whole plane (just as in the Fuchsian

case two regions, one on each side of the principal circle, are
usually required).
36 AUTOMORPHIC FUNCTIONS [CH. II, 19

Poincare has introduced a method of treating these groups by means of three

dimensional transformations. Each transformation is equivalent to a certain even
number of inversions in circles in the plane. The transformations in the plane will
V the same these inversions be made in spheres with the same centres and radii
if

as the circles and furthermore by the use of these spheres the transformations are
;

denned for points above and below the complex plane. It can then be shown that
there exists a fundamental region for these generalised transformations bounded by
spherical surfaces orthogonal to the complex plane. The parts of the plane contained
in this three-dimensional fundamental region furnish the several fundamental regions
in the plane whose repetitions cover the whole plane.*

*
Poincare, Mtmoire sur les groups kleineens, Acta Math., vol. iii. (1883), pp.
49-92 ; Forsyth, Theory of Functions, pp. 610-618.
 CHAPTER III

AUTOMORPHIC FUNCTIONS

20. Automorphic Functions. Automorphic functions are the

natural generalisation of the circular, hyperbolic, elliptic, and
certain other functions of elementary analysis. A circular
function, as sin z, has the property that it is unchanged in
value if z is replaced by s + 2m7r, where in is
any integer that ;

is, the function is unchanged in value if z be subjected to any

transformation of the group (s, z +
Zm-n-). A
hyperbolic function,
as sinh z y
is unchanged in value if z be subjected to a trans-
formation of the group (z, z + Zmiri)', an elliptic function, as
the Weierstrassian function @(z), retains its value under trans-
formations of a group of the form (z, + mw-f-mV).
DEFINITION. A function F(z) is said to be automorphic
Tk\
7 I

-
where
(o j]=F(z),
. - ,. /. ,
7

CZ -j~ Q/
is any transformation of the group*
It will be proved later that there are automorphic functions
connected with any discontinuous group. For the present we
assume this fact and proceed to derive some of the properties
of the functions. We shall confine our attention to groups
(Fuchsian or Kleinian) for which the fundamental region has
a finitenumber of sides, and to those automorphic functions
having no other singularities than poles in the fundamental
region. To avoid long circumlocutions in the statement of the
*
There are no functions, other than constants, which are unchanged in value
under all the transformations of a continuous group. For, let be a point at which
such a function has no singularity. Since the group has infinitesimal substitutions,
there are infinitely many points in the neighbourhood of % at which F()= F(z ). It
is a well-known theorem that a function which has the same value at infinitely many

points in the neighbourhood of a point at which it has no singularity is identically

equal to that value. Hence, F(s)=F(z ), a constant. It is for this reason that we
confined our attention in the study of groups solely to those which are discontinuous.
37
 AUTOMORPHIC FUNCTIONS [CH. Ill

rema we shall call such functions simple automorphic

t'luiis* Owing to its automorphic character we shall have
mplete knowledge of the function when we know its

li.-liaviour in the fundamental region.

21. Properties of Simple Automorphic Functions.

THEOREM A. A simple automorphic function takes on its

vain* p ti /nes, or some multiple thereof, at a vertex belonging to

icle the sum of whose angles is 27r/p.
Ki : ) is said to take on its value s times at a point Z Q if it

can be written F(z) = F(z )-{-(z-z Q ) (/)(z), where In

s

r/>(^)^0.f
considering the number of points at which the function has the
valu- F(:,,) the point z is counted s times. Let the function be
simply automorphic, and let A be a vertex belonging to a cycle
of angle 2ir/p. We can write F(z) = F(A) + (z- A) s (p(z), where
0(A) 0, and we are to prove that s is a multiple of p.
=4=

It was found in Section 17 that in the Fuchsian case there

i> a transformation T of the
group with A as fixed point which
rotates points near A through an angle 2-jr/p. The other fixed
point is A', the reflection of A in the real axis. T then is
z' A 2ni/p
z A
A similar fact can be
the transformation e
zA! ,
z
T-,-
A.

shown for the Kleinian group, A' being a point not in the
neighbourhood of A. Applying this transformation to the
function, we get

Viz)
= F(A) + (z - 4)W) = F(A) + (z - A)'6^
The being automorphic, is unchanged in value. By
function,
taking near A, z' is also near A; and, since A" is a distant
z

point, the function in square brackets can be made to differ

arbitrarily little from </>(z). Hence e 2ir */p=l, or s is a multiple
of //
: which establishes the theorem.
If there are n in the cycle, np regions have a
vertices
\.
Dividing the s-fold value there equally among
tin -in. 8 nj> will !) said to belong to the fundamental region.
A similar remark applies to each of the other congruent

II th- ^rouji Fuehsian, the simple autoraorphic function

is is usually called a
lisiiiu Inaction.
'" l1 "' "f absent and
pole at negative in the formula.
'
1" '
l
, F(c ) is s is
20, 21] AUTOMORPHIC FUNCTIONS 39

vertices, yielding s/p (an integer) points belonging to the

fundamental region at which the value F(A) is taken on.
In the neighbourhood of a parabolic point P the behaviour
is less
simple. Since infinitely many regions cluster about P
the function has there an essential singularity (i.e. a singularity
which is not a mere pole, for near a pole F(z) can take on no
value more than a finite number of times). Let F(P) be the
value approached by Y(z) when z approaches P in the funda-
mental region. We can show that there is a parabolic trans-

formation of the form p = z- p + /3 by which the

z'-P
function

is unchanged. It is unchanged by any repetition of this

transformation: -^ p
= ^ ~~ p + n/3. By transforming the point
% ~~" JL XT

P to infinity by the transformation Z = l/(z P) the transforma-

tions become Z' = Z 4-^/3, a group of simply periodic transforma-
tions. From this we can show that F is a function of t having
at most poJes where t = e 2irizff Hence
*.

THEOREM B. At a parabolic point a simple automorphic

function is a function of t having at most poles, where

This kind of singularity is called logarithmic. The number

on its value at the vertices of the cycle
of times that ~F(z) takes
to which P belongs is determined by the index in the m
expression for F form F(z) = F(P) + f n 0(f).
in the
THEOREM C. A simple automorphic function, not identically
zero, fix dfi many zeros as it has poles in the fundamental region.
It

If f(z) has no other singularities than poles in a region en-

closed by a contour and if the function and its derivative are
continuous on the contour and the function does not vanish

there, then the integral =. I d log f(z) taken around the contour
in a counter-clockwise direction is equal to N M, where N is

the number of zeros and M the number of poles in the

region.* We
shall apply this theorem to the automorphic
function F(z), integrating around the boundary of the funda-
mental region. We
suppose first that the function has no
zeros nor poles on the boundary. Consider the parts of the
*
Whittaker and Watson's jfodem Analysis, p. 119 ;
or Osgood, Lehrbuch der
Funktioncnthcorie, 2nd ed., p. 333.
40 AUTOMORPHIC FUNCTIONS' [CH. Ill

integral arising from two congruent sides AB, CD (fig. 15),

/" 1
'
/*
^
I d log F(z)+ d log I
F(0). Since F(z) has the same values
JD JA.
A B
f
along the congruent sides, this sum is I d log F(z) +
[
I cZ
log F(z)
= 0.

Similarly the parts arising from other pairs of congruent sides

cancel and = or
;
N M N = M,
which was to be proved.
The cases in which F(z) has zeros or poles on the boundary
If there is a zero or a pole on
require further consideration.
the side AB, there is one also on the side CD. Change the
path of integration by means of two congruent arcs as shown
in fig. 16, and the integrals cancel as before. Only one of
the points is included in the new region and it is clear that

A /
C
/A :

FIG. 15. FIG. 16. FIG. 17.

since the points are congruent only one should be considered as

belonging to the region.

If F(z) has a zero or pole at a vertex A, we integrate as in

fig. 17. We shall suppose it is a zero of order s. Let 2-rr/p be

the sum of the angles in the cycle to which A belongs. The

integral ^~.ld log F(z) taken in a counter-clockwise direction

entirely around is s; hence, the A

integral taken in the clockwise
direction along the small arc shown in the figure is s. l_ A/2-Tr.

There will be a similar integral at each of the congruent vertices,

,-u id these will not cancel. Hence N (s/2-7r)S L A
-= sip M= t ;

or N-f-/p = M. But s/p is precisely the number of zeros belong-

in-- to the
region which have not been included in the contour;
and wo see that the theorem holds here also. The case of a
.it A is treat CM! similarly.
\V- can show likewise that the case of a pole or a zero at a
olio point produces no exception.
Tin- methods above dispose of every case except those in
winch tln-iv are infinitely nuiny zeros or
poles on the boundary
1)1 '!" J'.ut th,- function cannot have
region, infinitely many
21] AUTOMORPHIC FUNCTIONS 41

poles without having an essential singularity which we assume

itdoes not have and it cannot have infinitely many zeros with-
;

out being identically zero. Hence the theorem is established.

THEOREM A simple automorphic function which has no
D.
singularities in the fundamental region is a constant.
Let F(s) be such a simple automorphic function and let its ;

value at Z Q a point in the fundamental region, be C. F(z) C

,

is a simple automorphic function with a zero at z and having

no poles. Such a function, according to Theorem C, is identi-
cally zero. -
Therefore, F(z) C=0, or F(z)=C.
THEOREM E. A simple automorphic function which is not
a constant takes on every value in the fundamental region the
same number of times.
Consider the function F(z) C, where C is a constant. This
is a simple automorphic function with the same poles as F(z).
It has as many zeros as it has
poles. The zeros of F(z) C are
the points at which F(z) takes on the value C. Since C may be
any constant, we conclude that the number of times F(z) takes
on any given value is equal to the number of its poles, which
establishes the theorem.
THEOREM F. Between two simple automorphic functions
belonging to the same group there exists an algebraic relation.
Let F^z) and F 2 (z) be the two functions with /^ and k 2 poles
respectively. We are to show that there exists a relation
-1
F 2 ) = A 1 F 1 F2 + A 2 F 1 F2 + . . . + A (M+1MB+1)S = 0,
where A A 1? 2 , etc., are constants not all zero, the relation being
independent of z. The function <1>, whatever values be given
the constants, is a simple automorphic function which has no
poles other than at the points where occur the poles of F 1 (0) and
F 2 (z). The number of its poles is not greater than mk^-\-nk.2 .

We can choose the constants A p A (m+1)(n+1) so that will have . .

., <|>

zeros at (m + l)(n+ 1) c
1, points v c etc., not coinciding with the 2 ,

poles of F^z) or F (z) by satisfying the (ra + l)(n + 1)

2
1
equations
A 1
F 1 m (c )F 2 M (t- )+
i <
. . . +A (
,,, +1)( , m) = 0, t = l,2, . .
., (f+l)(n +!)-!.

Constants, A x A 2 etc., not all zero, can always be chosen to

, ,

satisfy these equations since there is one more constant than

Now if m and n be large enough
equations to be satisfied.
(m+l)(n+'L)--l>mkl +nk2t and the Fuchsian function $ lias
 \rro.MORPHIC FUNCTIONS [CH. Ill

than possible poles. According to Theorem C,

<!>( Fv F2 )=0, the poles having mutually cancelled. Hence the
theorem.
It will generally happen, if the algebraic relation be
found in this manner, that $(F 1? F 2 ) is reducible: $(F 19 F2)
-'lyF^FjJ^F^Fj,) . . .
<
n (F l5 F2 ). Some one at least of the
irreducible factors must vanish, say ^(F^ F 2 )=0. This irreduc-
il>lerelation will contain both functions unless one of them is
a constant, in which case the irreducible relation is of the form
i-c^O.
easy to see in the general case what the degree of this
It is
irreducible equation in F 1 and in F 2 will be. The degree in F l is
the number of values of F x satisfying the equation when F 2 is
given a fixed value. For a fixed value of F 2 there are k 2 values
of z at which F9 (z)
takes on that value, 7c 2 as before being the
number of poles of F 2 (0). For each of these values F-^z) has a
value satisfying the equation. Hence, i 1 (F 1 F 2 ) = is of degree < >
,

/,-._,
in Fr
Similarly of degree in F 2 For particular
it is \ .

functions some of the values of F 1 arising from the k z values of z

may always be equal (as in the functions F and F 2
),
in which
case the degree in F a is less than k z .

THEOREM G. Any simple automorphic function can be ex-

prt **ed rationally in terms of two simple automorphic functions
connected with the group.
Let us consider two functions F^z) and F 2 (0) with /^ and Jcz
poles respectively whose irreducible algebraic relation $(F 1? F 2 )
=
is of k
degree z in F l
and /^ in F 2
as mentioned in the preceding ,

paragraph. Let F8(#)

be a third Fuchsian function connected
with the group. Each pair of values of F 1? F 2 satisfying the
algebraic equation is given by a single point z in the funda-
mental region (save for a few exceptional points). This value of
one value only of F 3 (^). F 3 is then a single- valued
function of F, and F2 so long as these satisfy the algebraic rela-
tion 'I'CFj,
F 2 = 0.
) Such a function, when it has only polar singu-
lariti. s as in tli.' case of F 3 we know from the theory of algebraic
,

fund ions to !>< a I- ;L | ional function of the variables in the

algebraic
-<|uation. ll<-;ico

,A F 1 F,+.
1
.
._
2) '
i;,F/'i7+ -" '
]

which was to be cst al.lisli.-d.

21] AUTOMORPHIC FUNCTIONS 43

The derivative of an automorphic function is not, in general,

automorphic. = (az-\-b)/(cz + d) be a transformation
Let z of the

group, where ad-bc = l. Then F(Y) - F(z) but ;

and F'(X)*F(X), except in the case (cz + d) = I 2

for all trans-
formations of the group. This condition holds only in case the
transformations are all translations, z' = z + k* We note that
the ratio of the derivatives of two automorphic functions is
automorphic; i.e. F 1 '(z')/~F 2 \z')
= 'F l (z)/'F 2 (z).

THEOREM H.
Two simple automorphic functions connected
with a group satisfy, in general, an algebraic differential
equation of the first order.

For r is a Fuchsian function. By the pre-

az iI~T^
7-^ > >

az "TTTT
ar 2

ceding theorem this is expressible rationally in general in terms

of F]_ and F 2 Hence, dF1/dF2 = R(F1 ,F2 ) the variable z not
.
J

explicitly appearing.
For convenience
let us represent an automorphic function by

x(z). Concerning the inverse function z(x) we shall now prove

the following remarkable theorem
THEOREM I. // x(z) is an automorphic function of z, z can
be expressed as a function of x by the quotient of two solutions
of a linear differential equation of the second order of the form

Z = jR(w, x)y, ivhere <(w, x) = 0,

R being rational in w and x and <I> a polynominal.

Consider the function y-^z) = ^(dxjdz). We shall show first

that - 1
-r^ is an automorphic function. Let us subject z to a,

2/i
ax
transformation of the group, and represent by y^ and x'( = x)
the transformed values of the functions. Making use of the
equations dx'/dz' ^z+d^dx/dz and l/y l
2 = dz/dx, we obtain
the following:
*
The singly and doubly periodic functions constitute this exceptional class in
which the derivative is also automorphic.
44 AUTOMORPHIC FUNCTIONS [CH. Ill

(cz + (i) i
+
+Cc
da; dx
i__^
'<fo>2 y^
Hence,

This function being automorphic, is expressible rationally in

terras of x and some other automorphic function w, and w and
are joined by an algebraic relation <&(w t a;)
= 0. Whence o/j

satisfies an equation

4,
= R(0, a;) v,
where $(?y, a;)
= 0.
ax 2

Now let 2/ 2
be a second solution of this equation independent of
the first ; y^ and y t satisfy an equation of the form

where c is a constant different from 0. [In general the ex-

fFdx where P(x) is the coefficient of dyjdx.
pression equals ce~ ,

=
In this case P(se) 0.*] Dividing by y^,

l_^/2_J/2. ^l = _ll = c ^.
y l dx y^ dx y-f dx

Whence, integrating,

or

where //.,
is ;i solution of the equation. The theorem is thus
Wished
\\V note iluit two independent solutions of the equation are
/:) and =

22. The Thetafuchsian Series. We

have heretofore assumed
functions automorphic with
respect to a given
7th, ./ '/'/ ;! K'jinittunn, -ith cd., p. 114.
21, 22] AUTOMORPHIC FUNCTIONS 45

group. We shall now demonstrate their existence, and in fact

construct such functions, by means of a series discovered by
Poincare.* We treat here the case of Fuchsiari groups.
Let the unit circlewith centre at the origin be the principal
circle of the group ;
and let

CjZ -f- &i

0, 1, 2, 3, . .
., M<-^ = 1

be the transformations belonging to the group. We shall suppose

the identical transformation to be z = z. Let K(z) be a rational
function of z none of whose poles lie on the principal circle. We
consider now the following series

We shall prove presently that this series

is
convergent (save
for certain exceptional points) both within and without the
principal circle, provided is an m
integer greater than 1. This
series is called the Thetafuchsian series. Let us assume the
convergence of the series and investigate the properties of the
function 9(0) which it defines.

If z be subjected to a transformation of the group

the series becomes

If we represent by %
the transform of the point z when the
j-transformation followed by the ^-transformation is made, the

second factor of each term of the series is H( -^

CijZ
- ^ )
. The

fi
first
,.
factor
.

is (
^ j ' '

r ^ 7
^
2

The first of these factors, being independent of i, is the same

for all terms. The series then is

But the expression under the sign of summation is 6(0) for the
;

i- transformations of the original series included all the trans-

*
Compies rendus, xcii. (1881), p. 333; M6moire sur les fonctions fuchsiennes,
Acta Mathematics, vol. i. (1883), p. 193.
46 AUTOMORPHIC FUNCTIONS [CH. Ill

formations of the group, and the ^'-transformations of the last

series include also all the transformations of the group, the
order of the terms being changed. Hence,

By means of Thetafuchsian series we can set up infinitely

many automorphic functions. Let G^z) and 6 2 (V) be two series
formed from two rational functions H (^)
1
and H (0), the integer
2

??i
being the same. Consider the function Y(z) = G1(0)/G2().

F(z) then an automorphic function.

is

form other combinations of 0-series which are

It is easy to
Thus if the numerator be a sum of terms of the
automorphic.
form G^ ... 6* where the integers used in forming the series
are 15
mm.2 ,
mk respectively, and the denominator be also
. . .

a sum of such terms, the function will be automorphic if

mi _[_ m2 4. t . .
_|- m k is the same for all the terms. If F(z) is
an automorphic function, we found that its derivative is multi-
plied by (CjZ + dj) on performing a transformation of the group.
z

m
Hence 0(z)/[F\z)] is an automorphic function, and we can
readily deduce other combinations which are automorphic.

23. Convergence of the Series. There are certain points of

the plane at which individual terms of the series become in-
finite. If a is a pole of H(0), there is an infinite term when

(a0+b )/(c&+dj)=a.
t That is, a term becomes infinite when-
ever z approaches any point congruent to a pole of H(0). A
t'Tiu becomes infinite also when 2 = djc t that is, at points con- ',

gruent to the point infinity. Since each transformation carries

the exterior of the circle into itself, these latter points are all
rior to the circle. Let a small circle be drawn about each
pole, of II(:) and let the regions within these circles and all the
ul.'ir
regions congruent to them be excluded from considera-
tion. Similarly let a large circle K
be drawn with no point
n -nt to infinity on its exterior, and let the region outside
it

this circlr and the congruent circular regions about the points
u< !<<!.
I The portion of the plane that remains
that inMli- K and outside the circles about the other excepted
22, 23] AUTOMORPHIC FUNCTIONS 47

points is finite in area, Let C be any contour lying within

this region and containing none of the excepted regions on its

interior. We
suppose also that no two points within C are
congruent. C may lie entirely within the principal circle, or
entirely outside it, or it may contain a portion of the circle on
its interior in case the fundamental region of the group contains

a portion of the circle. We shall now prove that within C the

Thetaf uchsian series converges uniformly.
To do this we whatever value z has within C
shall prove that
the absolute value of the term (c^ + c/ 2
)~ '"H(0 ) is less than a i j

constant U^, and the series 1 +

U U
2 + converges. Let M, . . .

and in be the maximum and minimum values of eg + d \~ l in

i {

C . Now \CiZ + di\- l = \d - l \z+d jci \-\ and \>s + djc is the i i \

distance from the point z to the excepted point djct. This

excepted point lies within the circle K and without the region
C . If D be the greatest distance from a point of C to a point
on K, and d be the least distance from a point of C to any of
the circles surrounding the excepted points, we have M*< l/c -D and t

tiii > 1/6$ whence M < dnii/D.

; f

Let C lf C 2 be the regions congruent to C

,
. . . and let the ;

areas of the regions be A A 15 A 2 Applying the formula ,

. . . .

found for areas in Section 5, we get the following

/ /
|
c,z +d t \~'dxdy > | j

Hence m *<A /A
< < ,
and M f
4
<cZ 4 A,/D 4 A .
Now, M +M +4
1
4
. . .

< (cZyD A 4
)(A + A + 1
. . .
)==S, a finite quantity, since the sum of
the areas A A ,
1?
. . . is less than the area of K. Hence M <S*, f

whatever the value of i\ and M am = M

< 4
2TO - 4
M <S<
4 (ro - 8)
^A
4
< _/D
4
A .

This inequality is true on the assumption that m^2.

Consider now H(s,). If z is in C each transformed point z lies i

outside the excepted regions around the poles of H(0). Hence

H(f) is finite and we can choose a constant greater than the H
value of the function for every point outside the excepted regions.
Wehave now the material for establishing the convergence
of the series. Each term (c z 4- d )~ 2m 'H.(z ) is less in absolute t i i

value, whatever the value of z within C than the constant ,

Mr""H. But
 AUTOMORPHIC FUNCTIONS [CH. Ill

The s.-ris ^M H (
thus converges; hence the original series con-
tbsolutely and uniformly in C and the sum 0(0) is an
;

analytic function of
:.*

one of the excepted points a term of the series becomes

infinite, but if this term be suppressed the series converges. 9(0)
then has a pole at the excepted point. Let us consider the
function in a fundamental region within the principal circle.
There is a point in the fundamental region congruent to each
pole of H(0) within
the principal circle. Hence, the number of

poles of 9(z) within the fundamental region is in general equal

to the numberof poles of H(z) within the principal circle. In
particular cases two poles of H(0) may be congruent points, and
the order of the pole at their common congruent point in the
fundamental region is not greater than the highest order of the
two poles.
In a fundamental region outside the principal circle the
number of poles of 9(0) is equal to the number of poles of H(0)
outside the region increased by 2m, the latter arising from a
term (c^ + c^)" 2" at a point congruent to infinity. In a group
1

having a fundamental region crossing the circle the number of

poles is the sum of the two preceding numbers. The function
has essential singularities at the points of the principal circle
where the excepted points cluster.
If H(z) has no poles within the principal circle, or if the poles
are at congruent points and are of such a nature that the poles
of the infinite terms arising from them cancel, 0(0) has no singu-
larities within the circle. Poincare has shown, in the memoir
cited above, that of all such functions, formed with a given value
>f m,
only a finite number are linearly independent [i.e. for a group
whose fundamental region does not cross the principal circle].
The poles of the automorphic function F(0) = 9 1 (0)/9 9 (0) arise
from the poles of the numerator and the zeros of the denominator ;

zeros arise from the zeros of the numerator and the poles of
tin; l.'iiominator. Now, the function F(0) has the same number
zeros in the fundamental region.
oi' conclude We
from this that for all functions 9(0) formed with a
given m
of a similarly formed series for a group without a principal
Ml.lisln-d by slight, and obvious changes of the proof just given,
idi-d infinity is not a point of discontinuity of the The series is called
group.
23] AUTOMORPHIC FUNCTIONS 49

the difference between the number of zeros and of poles is a

constant.*
It should be noted that in case all points of the principal
circle are points of discontinuity of the given group that is,
points at which infinitely many fundamental regions cluster

every point of the circle is an essential singularity for the

functions defined by the series outside the circle and the function
inside. Each of the functions then has the circle as a natural

boundary. Since the function inside cannot be continued

analytically into the one outside, the two functions, although
defined by the same series, are distinct functions.
* Poincare
proved that in the case of a fundamental region lying entirely
within the principal circle, the number of zeros exceeds the number of poles by
m(n-l -S/27J-), where 2n is the number of sides of the fundamental region and S
is the sum of the angles of its vertices.
 CHAPTER IV

THE UIE.MAXN-SCHWARZ TRIANGLE FUNCTIONS*

24. The Differential Equation. Given a linear differential

equation of the second order

and two independent particular solutions y = y 1 (x) and y = y.2 (x):

then the general solution is of the form y = Ay 1 + ~By z The con- .

nection with linear transformations, which we have already noted

in Section 21, Theorem I, arises in the following manner. Let
= 2/1/2/2 be the quotient of the two solutions and let 0' = y^y* be ;

the quotient of two other solutions, y^' = ay jrby l z , y2

'
'

= cy\ + dy 2 .

Then
g
,
= ay + by z = az +
l
b
'

cy 1 i- dy z rz +d
In particular ^
and y% may be other branches of y^ and 2/ 2 in
these solutions are not single-valued. The determinant of
'
this linear relation will vanish only if y^ and y z differ by a
constant factor.
\Vc shall consider the hypergeometric equation with three
^insularities and with the exponent differences X, /u, v. The
real

<j(juati(jii can in all cases be reduced to the following

an iMjuation whose singularities are at the points 0, 1, ao .

tin;
ordinary method of the substitution of a power series
in tin- differential
equation and determining the coefficients, we
'ind that it' a is a
point not coinciding with 0, 1, or oo there are ,

!i|.-].eii(l..]i1 solutions of the form

typerg, /,'<///' in his Jr<>rfa, Nachtrage-,
-I/"'//,. 75 (1873), ji. 2

50
'24.
25] THE RIEMAXX-SCHWARZ TRIANGLE FUNCTIONS 51

//,,
= (a- a) {1 + l\(.f -a)}

where P 1 and P 9 are ascending power series in x a having no

constant terms. Similarly at the singular points we find the
solutions
y lf0 -a*{l + P8 (*)}
2/2,0 =1+P 4 (*)

l\j(l-X-^-,
ay

Further, the coefficients in the series P 3 ,

. .
., un> real, and
the coefficients in P 1 and P 2 are real if a is real.

25. Conformal Map of Half-Plane. Let ; = //,(<)//,(>) be

the ratio of two linearly independent solutions. ; is in
general
a many-valued function of x, but in a region not surrounding
the singular points any branch of the function will be single-
valued. Such a region is the upper half-plane, and we now
propose to find into what region in the s-plaiie the upper half-
plane is mapped by the function : = y l '//.,.

Consider a point a on the real axis, between and 1, say.

The function : = y Jy. a = (x a ) { 1 + P(,r a)} maps the neigl -
lt 2t
i

bourhood of the point x = a on the neighbourhood of c = 0. rt

The coefficients of the series P(x-~ a) are real; hence, the real
axis near a is transformed into the real axis in the s -plane near tt

the origin. By virtue of a remark in the preceding section,

:=(f,,, + (r :, + r/ 1 ), a linear transformation by which the
/;
l
)
]
!

real axis in the 2,,-plane becomes a circle in the z-plane. Hence,

the real axis near a is mapped on an arc of a circle in the 0-plane.
If the series for z a does not converge throughout all the region

1, we consider c,,, the ratio of the principal solutions at a point

6,such that the region of convergence of the series for z b overlaps

that of the preceding series. We get the axis at b mapped on
an arc in the :-plane. It is a continuation of the arc previously
found, since it coincides with it in themap of the common region
of convergence. Continuing, we get the line 1
becoming, by
the transformation z =y l /y z ,
a circular arc in the 0-plane.
:,L> .u TOMORPHIC FUNCTIONS [CH. iv

At the origin the ratio of the two solutions of the preceding

ion is : = //,, //-.-. o
= ^{l + PO(^)}- Taking the real branch
M!' the function, when x>0, z >0; the positive part of the real
axis near = () becoming the positive part at z = 0. But when
./

./
passes above the origin and then traces the negative portion
of the axis, z is in Although P (a?)
general no longer real.
K
remains real, containing no fractional powers, X does not.
Writing x = pe
ie
the circuit of the origin gives x = pe iu on the
,

itive part of the axis. Whence x^ = p x eik7T and z has the ,

factor (>'**. The axis x<0 becomes a line in the 2-plane making
an angle XTT with the positive real axis, the region in the upper
=
half-plane at x Q being transformed into the region inside this
angle. By means of = (a z + 6 )/(c 2 + c ) this line becomes an
arc in the 0-plane making with the arc previously found an
angle XTT.

Similarly we can show that the part of the real axis for x>l
is transformed into an arc in the z-plane meeting the transform
of the axis 0<x< 1 at an angle ^TT. If we continue the mapping
of the parts x>l and x<0 to infinity, a consideration of the
solutions y^ ,, y 2iao shows that the two arcs meet at an angle
J/TT. Thus we have the result that
The ratio of any two solutions z = y 1 (x)/y 2 (x) maps tlte upper
half-plane of the variable x on a circular arc triangle in the
plane of z, the sides of the triangle being the transforms of the
lines 0], loc ,
and oo 0. Its angles are ATT, JULTT, VTT.

Let us denote the vertices of this triangle by z z v S M If, ,

.

starting at a, we had traversed the real axis passing below the

singular points we should have had a map of the lower half of
the plane on a triangle z^z' n with angles XTT, /ULTT, J/TT, the side
:,,*-! being common to the two triangles. It is easy to find the
n-lation between these triangles. Since each of the auxiliary
functions introduced, z at z etc., are series with real coefficients,
,

conjugate imaginary points in the #-plane become conjugate

-
imaginary points in the -plane, etc. These go by the linear
transformation into pairs of inverse points with respect to the
'ii-dr into which the
real axis is transformed in the s-planf.
ll'-ncc. UK- nut j) of the lower
half-plane jn*t found Vx derived
il,''
tricmgle /,/-,/, by an inversion in the side z z r
'

Now, is a
:
many-valued function of x. If we start with a
on the negative part of the axis,
j>>int using the branch of the
25, 26] THE RIEMAXX-SCHWARZ TRIANGLE FUXCTIOXS 53

function which mapped the line Ooc on Z O Z M and map the lower ,

half-plane with this branch, passing below the singularities, we

get a triangle which is an inversion of the original triangle in

the side z z x In general, there is a branch of the function

.

mapping either half-plane 011 a triangle derived from the map

of the other half -plane by inversion in a side. These triangles
are in general infinite in number and overlap.

26. The Triangle Function. The relation z =y l (x}/ y. ,(,>)

defines x as a function of say x=f(z). Under what circum-

z,

stances is f(z) a single-valued function ? The triangles furnish

a ready answer. It is that there must be no overlapping; for
a point z belonging to two maps of the .i'-plane gives two different
values of the function. At z the two triangles mapping the
upper and lower half-plane form an angle 2\7r, and in order
that there shall be no overlapping by continued reflections
in the sides meeting at Z Q it is necessary and sufficient that
,

2\7T be a submultiple of 2?r. A similar remark applies to

the other vertices. Hence, the angles of the triangle \TT, JULTT,
VTT are submultiples of TT, or A, /m, v are the reciprocals of

integers*
For convenience let each triangle which is a map of the

upper half -plane be shaded. The shaded and unshaded triangles

alternate. Each shaded (unshaded) region is derived from an
adjacent unshaded (shaded) region by inversion. Hence, any
shaded (unshaded) region is derived from any other shaded (un-
shaded) region by an even number of inversions that is, by a ;

linear transformation. easy to see that these transforma-

It is
tions form a group (both group properties of Section 8 being
satisfied) and that a shaded and an <lj<'<>nt unshaded triangle
;

together fown a fundamental region for fJ/r group.

The fundamental region is a map of the whole ft-plane. By
a transformation of the group this region is carried into some
other map of the plane congruent points being the transforms
;

of the same point in the ^-plane. Hence, if z and z' are any two
congruent points f(z') =f(z) ;
or f (z) is automorphic with respect
*
It is possible also to have zero values. If \ = Q, in finding and y-2 o the
T/I, ,

method of substituting a series of undetermined coefficients yields but one solution.

A second solution with a logarithmic term can be found, and we can show that in the
triangle ^ ^z x the sides -^ and ^^.^ are tangent at c The angle of the triangle is
.

thus ATT as before.

 AUTOMORPHIC FUNCTIONS CH. IV

This uutomorphic function is called a triangle

27. The Three Kinds of Groups. The groups arising in

connection with the triangle functions exhibit strikingly different
* characteristics according as
the sum of the angles of
the triangle is equal to, less
than, or greater than TT. We
divide the groups, for pur-
poses of study, then into the
three cases :

Case I.

Case II.

Case III.
It will be a simplification
to transform the 0-plane as
follows. Let the sides of
FIG. 18.
the triangle meeting at, say,
~
meet again at Let a linear transformation be made carrying
infinity. Two of the sides
of triangle then become
the
straight lines. In fig. 18 is
shown the character of the third
side in each of the three cases.

28. Case I.

In this case the third side is

a It is obvi-
straight line.
ous that by continued reflec-
tions in the sides the
triangles
"/' i' iff i a number u<]
I he tt'hu/r.
/>lne. There
four possible cases: in which
FlG 19<
i/havethe values (i.) \, .', 0;
-

<>) I I }
: and (iv.) |, :
'

;
,
:
'

;
. The division of the
plum- I'or (ii.) is shown in th.-
accompanying figure.* On trans-

it will l.e oljs'-rvr.l Unit several of the triangles in this case we

l>y eonil.iniiig

opposite sides are congruent by transformations of

:
Thus in lig. 1!) the twelve either a side or a \
triangles having
20-29] THE RIEMANN-SCHWARZ TRIANGLE FUNCTIONS 55

forming back to the original s-plane we get infinitely many

triangles with the point z', which in this case is the common
intersection of the three sides, as a cluster point.

29. Case II. A + ,u + i'<l. The third side BC (fig. 18) is

convex toward A, hence A lies outside the circle of which BC is

FIG. 20.

an arc. From A we
can draw a real tangent AT to this circle.
With A as AT as radius (T being the point of
centre and
tangency) let a circle K be constructed. is K
orthogonal to
each of the sides of the triangle. By an inversion in any side

on the vertical line through the centre of the figure is such a parallelogram, the
translations which carry a side into an opposite side transforming the system of
triangles into itself. This is then the period parallelogram for a doubly periodic
group, which is a subgroup of the original group.
The reduction of the triangle functions to elliptic integrals in the four cases in
which \ + + v = l has been effected by A. Polakov, 1'ecueil Math, de Moscou, 27
/j.

(1911), p. 424.
 ."; AUTOMORPHIC FUNCTIONS [CH. IV

of the triangle, K is transformed into itself. The new triangles

arising will also have their sides orthogonal to K, and for all
Mirn-.-ding inversions K will be a fixed circle. The linear trans-
formations arising from an even number of these inversions have
K as a principal circle. The group is Fuchsian. It follows
from Theorem 15, Section 18, that the triangles will be infinite
in number and entirely fill the circle K.
There are infinitely many possible values for X, ^ v such that
their sum is less than unity. Fig. 20 shows the case in which
the values are J, J, \.
The modular group is an example of the
case of the values J, J, 0. The fundamental region shaded in
fig.
8 is composed of two triangles. The single triangle is the
half of the shaded region formed by a vertical line through the

point C. The same figure can be used to illustrate the case of the
values \, J, 0. In this case the shaded region is a single triangle
and the fundamental region consists of the shaded region and
an adjacent region. The group is no longer the modular group,
being composed of only a part of the transformations of that
group that is, it is a subgroup of the modular group.
;

30. Stereographic Projection. Before discussing the third

case we shall introduce a widely used method of representing
the z-plane on the surface
of a sphere. Let T be the
0-plane, and let P be a point
not lying in the plane. Now
let an inversion be made in
a sphere K with its centre
at the point P. T is trans-
formed into a sphere S pass-
ing through P (fig. 21). The
F[r 21
same inversion transforms
the sphere S into T. This
inversionis called the
stereoyraphic projection of the sphere
on plane or of the plane on the sphere. Obviously the
tli.;

'onvspmdence of the points of S and T is one-to-one.

two outstanding properties of inversion in a sphere are
Tlii
that spheres
(including planes) are transformed into spheres;
( 1 >

mil <-!>tli;it the

angles l>et\voen lines are preserved in mao-nituoV.
From these \v can state the following facts conrernino- corre-
29-31] THE RIEMANN-SCHWARZ TRIANGLE FUNCTIONS 57

spending figures on S and T (1) angles are preserved in

:

magnitude (2) a circle in T is transformed into a circle on S,

;

and conversely (3) points of T inverse with respect to some

;

sphere R will correspond to points on S inverse with respect to

the transformed sphere R', and conversely.
By virtue of Thereom 7, Section 3, we can state at once that
every one-to-one and directly conformal transformation of the
sphere into itself corresponds to a linear transformation in the
plane. Thus the rotation of the sphere about an axis corre-
sponds to a transformation of the plane whose fixed points are
the stereographic projections of the extremities of the axis.*
It is easy to find the transformation of the sphere into itself

corresponding to an inversion of the plane in a circle C. The

effect on the plane is the same if the inversion be made in a
sphere R with the same centre and radius as C. This sphere is
orthogonal to the plane. By means of (3) above we see that
the corresponding transformation of the sphere is an inversion
in R', the transform of R by means of the inversion of the

stereographic projection. R' is orthogonal to S and intersects it

in C', the circle corresponding to C. Hence inversions of 8 in
orthogonal */>lieres correspond to ordinary inversions in the

/>f "tie T. Such an inversion changes the sign of the angles on S,

but an even number of the inversions preserves the signs of the
angles and corresponds to a linear transformation of the plane.
A particularly useful inversion in an orthogonal sphere is a
reflection in a diametral plane. The succession of two such
reflections (or in general of an even number) is equivalent to a
rotation of the sphere, the line of intersection of the two
diametral planes being the axis of rotation. Further, reflection
in a diametral plane does not change the magnitudes of figures
on the sphere, and transforms great circles into great circles.

31. Case III. \ + /A-hi/>l. There are four possible cases,

A, JUL, v having the values (i.) J, J, 1/n, where n is any integer;
*
This is the most general transformation of the sphere into itself by rigid motion.
Fur, given any rigid motion, there is one fixed point A on the sphere, since every
linear transformation has a fixed point. The centre is also fixed in rigid motion ;

hence the line OAa fixed axis, and the motion is a rotation.
is The circles on the
sphere in planes perpendicular to OA are the fixed circles. Projecting these on the
plane, we note that the transformation corresponding to a rotation of the sphere is
elliptic in type.
 AUTOMOKl'IIIC FUNCTIONS [CH. IV

.1.
.[
: and (iv.) I, !. 1. In these cases the
A "f tlif triangle (see tig. '2'2) lies within the circle of
'

\vhich 15< is an arc : t lit? re is no real tangent to the circle from

oimnon orthogonal circle. It will now be shown
'

that A IK --an be projected stereographically on a sphere in such

a manner that the three sides become arcs of great circles.
Through A draw the chord TT' of the circle of the third side
which is bisected at A (TT is
perpendicular to the radius
through A). The circle K, with A as centre and AT as radius,
is intersected by each side of the
triangle at two points which are
at opposite ends of a diameter.
Now plane be projected
let the
stereographically on a sphere S
with A as centre and AT as radius.
The points on the circle K are
unchanged in position. The three
sides of the triangle become three
circular arcs each of which passes
through opposite ends of a dia-
meter. That is, the sides of the corresponding triangle A'B'C'
on the sphere are arcs of great circles.
From the remarks of the last paragraph of the preceding
<>n we see that inversions in the sides of the triangle ABC
:

''rn-spond to reflections in the diametral planes of the sides of

the triangle A'B'C'. The sides of the new triangles arising by
th'-sr; rellt'Otions are also arcs of great circles; and the continued
inversions in the sides of succeeding triangles in the plane are
all reflections in the diametral
planes of the corresponding sides
nangli's on the sphere.The group of linear transformations
ing from combinations of even numbers of the inversions in
is a.
(/I'dtijj of rotation* of fit c xnlin'c.

Tin- an-a of the triangle A'B'C' is unchanged by reflection

in a diametral plane. Hence by repeated reflections the whole
sphere will be covered by a finite number of
Projecting these triangles on the plane, we can state
thnt /'// f/ lf
>

case X 4- yu -f > 1 flic triangles are finite in number

i/

'I in/ /ill Hit- ifliolr in-.

fi/it
\\V c.-in determine the number of the triangles in any case.
the Bpherical triangle A'B'C' is -( AT)-(X + /x + i/- 1),
31, 32] THE RIEMAXN-SCHWARZ TRIANGLE FUNCTIONS 59

2
and the area of the whole spherical surface is 4?r(AT) Hence.

the numbertriangles of is N = 4/(X + + /* i/ 1). In the four

possible cases mentioned above we have (i.)
N = 4?i (ii.) = 24
;
N :

(iii.)
= 48 and (iv.) = 120.
N ;
N In each case half will be shaded
and half unshaded. The number of copies of the fundamental
region, which consists of some two adjacent triangles, is X 2 ;

from which we conclude that the group of linear transforma-

tions has N/2 distinct transformations, including the identical
transformation.

32. The Regular Solids. Case III., in which the triangles

in the plane can be projected stereographically into ordinary
spherical triangles all equal in size, is connected with the groups
of rotations about axes and reflections in planes of symmetry
by which a regular solid is transformed into itself.*
We treat here only case (iii.), in which X, /UL, v have the values
\, -J-, \. The division of the plane into the forty-eight triangles
is shown in fig. 23. If these be projected stereographically on
a sphere 8 of which K is a great circle, we get a system of forty-

eight spherical triangles. These are shown in fig. 24. The

triangles are not equilateral.
Now consider the division of the plane into eight triangles
indicated by heavy lines in fig. 23. Each of these larger
triangles is composed of six of the smaller ones. The angles
of each are Tr/2, ?r/2, 7r/2 and each of these triangles becomes
;

an equilateral triangle an octant on the sphere. Now let the

bounding arcs of these triangles be replaced by the chords
joining their extremities, and we have the edges of a regular
octahedron inscribed in the sphere. This octahedron is shown
in fig. 24.f
Now let the rotations of the sphere which
us consider
correspond to the transformations in the plane by which a
shaded triangle, say, is carried into another shaded triangle.
The transformations by which one of the small shaded triangles
at the centre of fig. 23 is carried into the other three shaded
triangles at the centre correspond on the sphere to rotations
through angles vr/2, TT, STT 2 about an axis connecting two
*
Sclnvarz, loc. cif. ; Klein, Ma'.h. Ann., 9 (1S75), p. 183.
t I am indebted to Professor Crum Brown for the construction of the model of
which this figure is a photograph. L. R. F.
60 At.'TOMOlJI'HIC FUNCTIONS [CH. IV

of the octahedron. The transformations by

which one of these triangles is carried into the other two
shaded triangles lying in the same large triangle correspond to
rotations through angles 2-7T/3, 4?r/3 about an axis through the
centres of opposite faces of the octahedron. In each of these
rotations the octahedron is transformed into itself. Similarly
in show that the linear transformation by which a shaded

is transformed into
triangle any of the other twenty-three
shaded triangles corresponds to a rotation of the sphere by which
tahedron is transformed into itself. Hence, including the
no rotation (the identical transformation), their m n
I'

group of twenty -four rofofions by which the octahedron is

/'///"
transformed if.^-lf.
Hut we can say more. The inversions by which a shaded
>rn<-d into an adjacent unshaded triangle, and in
ato any shad-d triangle carry each of the eight large
32] THE RIEMAXN-SCHWARZ TRIANGLE FUNCTIONS 61

triangles either into itself or into some other large triangle.

On the sphere these inversions are reflections in diametral
planes by which the octahedron is transformed into itself. The
planes in question are the planes of symmetry of the solid, each
of which passes
through either four vertices or the mid-points
of four faces. Since a shaded triangle may be carried into any
one of the unshaded regions, there are twenty-four of these
reflections. They do not form a group, since the succession of

FIG. 24.

two reflections is not a reflection but a rotation.

But the twenty-
four reflections ttxjctlicr -with the
twenty -four rotations mentioned
<ij>ove constitute a group of transformations
by ivhich the
octahedron is transformed into itself.
Suppose that, instead of grouping the forty-eight triangles
into eight large equiangular triangles each
containing six, we
now group them into six quadrilaterals each containing eight
triangles. Take the eight triangles at the centre of fig. 23 as
one quadrilateral. The four quadrilaterals arising by inversions
in its sides together with that composed of the eight triangles
meeting at infinity constitute the other five. By projection on
the sphere the central figure becomes a spherical quadrilateral
(]-2 AUTOMOHPHIC FUNCTIONS [CH. IV, 32

ur ci|ual sides. The same is true of the others, since they

tions of the iirst quadrilateral in diametral planes. Re-
placing tin.- arcs by their chords, we get the edges of an inscribed
cul >f. The radii to the vertices of the cube pass through the
mid-points of the faces of the octahedron; and the radii to the
of the octahedron pass through the mid-points of the
- cube.
of The two solids have the same planes of
the
svmnietry. Because of this fact (or we can show it directly as

before) the twenty -four reflections and twenty-four rotations of

preceding yroup transform the cube into itself.
/./,.

The remaining cases of III. furnish the groups of rotations

and reflections for the other regular solids. In (i.) |, -J, I/n, we
have the dihedron, or Ti-gon, a regular "solid" of two faces.
Kadi of the faces is a regular polygon of n sides. The two faces
coincide, and the figure has zero volume. In (ii.) i, 4, i, we have
the tetrahedron and in (iv.)
; J, 1, we have
,
the ikosahedron and
dodekahedron. In each case the solid is transformed into itself
by a group of N/2 rotations and N/2 reflections in planes of
symmetry, where N is the total number of triangles.*
*
In Klein's Lectures on the Ikosahedron will be found a treatment of all the
solids. Special attention is paid to case (iv.), the group of sixty rotations and sixty
reflections being used in investigating the general
equation of the fifth degree.
 CHAPTER V
XOX-EUCLI DEAX GEOMETRY

33. Absolute, or Cayleyan, Geometry. The system of geometry

to be described in this was introduced by Cay ley.*
section
"
It is a geometry in which the terms " distance and " angle
are shorn of their usual meanings and are given new definitions.
These definitions an: made to depend upon a chosen fixed conic
called the <.il>*<)lat<'.

Let P, Q be two points: and let the line PQ meet the

FIG. 25.

absolute in the points U, V (fig. 25). Then the distance PQ is

defined to be proportional to the logarithm of the cross ratio
of the four points P, Q, U, V ;
that is,

DUtPQ -*!<*.
Let L, M be two lines intersecting at O and let S, T be the ;

tangents from O to the absolute. The angle between the lines

*
Sixth Memoir on Qualities, Phil. Trans., 149, 1859, reprinted in Cayley's
Scientific Papers, vol. ii.

63
T,4 AUTOMOIMMIIC FL'NXTIONS [CH. V

L. M is defined to be proportional to the logarithm of the cross

rat i nt' the four lines L, M, S, T; that is,
sin LOS sin MOT
Angle LOM - k log
sin
. _ ..,- a
LOTp sm M<
,
.

>h
-

\Ye verity directly from the definitions that if R is a point

OD the line PQ, Dist. PQ +
Dist. = Dist. PR;
QR and that if N
line
through O, angle + angle = angle LON.
LOM MON We
can slmw, in fact, that all the axioms of ordinary Euclidean
netry hold with the exception of the parallel postulate.
:
We
note also from the definition that if P approaches the distance U
1\) becomes infinite, and that if L approaches S the angle LOM
becomes infinite. If O approaches the absolute, S and T approach
coincidence and the angle LOM approaches that is, the angle ;

between two lines meeting at infinity is zero. L and are then M

said to be parallel.

34. The Three Types of Geometry. We choose for the

lute a conic whose equation has real coefficients. There are
then three types of geometry, as follows:
(1) The absolute has a real locus, and divides the plane into
two parts. This is the case of the ordinary real hyperbola,
parabola, or ellipse. The geometry is called hyperbolic.
2 The absolute has no real locus. The geometry
( ) is called

elliptic.
(3) The absolute is a degenerate conic consisting of a real
double line. The plane is not divided into two parts. The
geometry is called parabolic. This type is the limiting case
pa rating the two preceding.
S.-

\Ve easily discover many of the fundamental facts of the

various types. If the geometry is hyperbolic, two lines can be
drawn through a given point parallel to a given line. Thus in
tli.-
given figure where we consider points interior to the ellipse
ilp- two lines through U,V and the
given point are parallel to
In parabolic geometry there is one
parallel to a given line
(
l' ..

through 11
point, since the two intersections with the
late enineide. In elliptic
geometry there are none, since a
n ival line has no real intersection with the absolute. It can
hnwii without difficulty that in
hyperbolic geometry the
the angles of a triangle is less than two right angles*
lei about a point being called four right angles.
33-35] XOX-EUCLIDEAX GEOMETRY 65

(the sum is zero if the vertices of the triangle lie on the

absolute) ;
in parabolic geometry the sum is equal to two right
angles ;
in elliptic geometry the sum is greater than two right
angles. The plane of elliptic geometry is characterised by the
fact that there are no infinite points, for the infinite points lie on
the absolute which is imaginary. It follows from this that eacli

straight line is closed and of finite length.

The homogeneous equation of the absolute can by means of
a projection be taken in the form

and the geometry is hyperbolic, elliptic, or parabolic according

as e is positive, negative, or zero. The Cayleyan distance
between two points P(x v y v 2^) and Q(# 2 y z , ,
z. )
2
we find, on
forming the cross ratio of the definition, to be
>-(x l ,v.2 + iw2 )-z&+\ -
z^o]
2 - [c(xf + ?/!") - zf][c(x + y 22 ) - 2
z.; ] \

+ 2/i?'2 -
", 2,z2

We shall now show that when e = we have the usual Euclidean

definition of distance. When e approaches zero, the formula just
found approaches zero unless /,-. which need be constant only for
a fixed value of e, is made a function of e. 'Faking /.:= 1/2 N /f,

wo get for the distance when e approaches zero

~
Lin log 1 I

~V" Vi-'/a)--!^

-
= Lim '
//,-)-
--^2

which is the ordinary Euclidean formula for distance. A

similar treatment of angle can be made from the tangential
equation of the absolute, the result being that the case e
=
yields the ordinary definition
of angle. When e = the absolute
is c- = 0. Knfliili-uii geometry is thus a,
special case of absolute
at infinity counted twice.
geometry, the absolute being the line

Representation of Non-Euclidean Straight Lines by

35.
Circles in the Plane. The Elliptic Case. It is our aim to
arrive at an interpretation of certain facts concerning Fuchsian
and other groups by means of the ideas of non-Euclidean
geometry. In pursuance of this plan it will now be shown how
 AT TOM OK I'll 1C FUNCTIONS [CH. V

i^-ht lines of non-Euclidean geometry can be represented

Circles in the plane. We shall work out the formulae for
distance only, merely stating the results in the case of angle.
The treatment of angle is very similar to that of distance, the
Mtial equation of the absolute being used.
The familiar example of a geometry in which the sum of the
i' a triangle is greater than two right angles is
geometry
MM the surface of a sphere. Angle here has its ordinary mean-
ing: and the distance between the points is the length of the
an- of the great circle joining them. We shall see how this
uetry arises from elliptic absolute geometry.
= be the equation of the absolute.
Let x 2 + y 2 + 1 Let
the plane of the absolute be placed tangent to the sphere
X- + Y--fZ- = l, at the point (0, 0, 1), the aj-and 2/-axes of the
plane being parallel to the X- and Y-axes of space. Now let the
plane be projected on the sphere, the centre of the sphere being
the centre of projection. [The projection from its centre of a
sphere on a plane gnomonic projection. The pro-
is called a

jection we have just made

the inverse of this transformation.]
is

The straight lines of the plane project into great circles on the
sphere, since the plane through the straight line and the centre
of projection cuts the sphere in a great circle. If now we trans-
form the formulae for distance and angle, replacing the co-
ordinates of points in the plane by their values in terms of the
coordinates of the corresponding points on the sphere, we shall
have a consistent system of elliptic geometry on the sphere in
which straight lines are represented by great circles.*
Let us see what these formulae become. If P(OJ, y)is a point
in the plane and P'(X, Y, Z) the
corresponding point on the
sphere, we have, since P, P', and the origin lie -in a line,
1 = X
/ : Y Z. X, Y, Z can then be used as the homogeneous
: :

of the point P. Let P(^, yj and Q(,/' //.,) be two ,

points in the plane, the corresponding points on the sphere being

Y
and </< X,, 2 Z 2 ). The absolute distance PQ is,
,

twin miila of the

preceding section
;

>
Y Y 2 + Z^.,) 2 - (X 2 +
1

V{ (X,X a + Y! Y 2 + ZA)* - (X 2 +
: y ++ V)( X
2
2
+ Y.,2 + Z.,2 }
2 )

''
x Y, V)' AY V> + Zf) }'
v
+

metry with onlinury spherical t^voinctry is due to-

:]5, 36] XOX-EUCLIDEAX GEOMETRY 67

Since (X p Y I} Z : ) and (X 2 Y 2 Z 2 ) are the non-homogeneous

, ,

coordinates of a point on the sphere, we have X 1 2 + Y 1 a + Z 1 2 =

X., + Y 2 + Z = 1.
2

Putting
spherical
2

arc
2

circle joining P'

formula then becomes
Let

I-.-
Dist.
i.
and

~nr\
Q'.

1
-
be the length of the arc of a great
Then

cos
cos

PQ = A7ilog cosQ +- isinO = K71log e

we have the Cayh'ynn distance equal to the

k=%i,
joining P'Q'. A similar treatment of angle will
i sin
=X X +
t 2

9l
-

fl
2i(?
Y^ + Z^.
= n -

2t#0.
7 />
The

show that tlif C<njhijan angle between two lines is equal to the
angle (in the ordinary sense) between the corresponding <jri1
circles on the sphere.
If now we
project the sphere stereographically on a plane,
the great circles become circles in the plane. If we let angle
have its ordinary meaning and define the length of a curve to
be that function of the curve which is equal to the length of its
stereographic projection on the sphere we do not need to derive
the function, we have a system of elliptic geometry in which
straight lines are represented by the circles just found.
It is easy to see what these circles are. If the projection is

made from the point (0, 0, 1) on the plane Z = 0, the great circle
common to the sphere and plane, X* + Y 2 = 1, is unchanged.
Each great circle on the sphere intersects this circle at opposite
ends of a diameter, and the same is true of the projection of the
great circle. The straight lines are then represented by the
two-parameter family of circles X-

36. The Hyperbolic Case. If the absolute is real, we can by

a projection take it to be the circle xL -\-y L = \. We shall now
project the plane on the sphere
2
+ -f
2
Z- = 1 as X
follows. PlaceY
the plane of the absolute in coincidence with the plane = 0> Y
the origin in the plane being at the point (0, 0, 0) and the x- and
T/-axes being parallel respectively to the X- and Z-axes. We
take as centre of projection the point in homogeneous coordinates
(0, 1, 0,the point where the Y-axis meets the plane at
0)
intinity. The projections of a point in the plane are the two
points in which a line through the given point perpendicular to
the plane Y = meets the sphere. The straight lines in the
plane project into circles on the sphere w hich
r
are orthogonal to
the plane Y = 0.
68 AUTOMORPHIC FUNCTIONS [CH. V

Now let this system of circles be projected stereographically

t'mni the point (0, 0, 1) on the plane Z = 0. The circle in which
the plane Y=
cuts the sphere (i.e. the absolute, which was
unchanged by the previous projection) is projected into the line
Y = 0: and the system of circles is projected into the circles
orthogonal to this line. Hence, if the definitions of distance
and angle be carried over, we have a representation of hyperbolic
geometry in the plane in which straight lines are represented
by circles orthogonal to a fixed line.
We shall now determine the formula for distance. The
equations connecting the points on the Cayleyan plane with
those on the sphere are evidently oj = X, y = Z. Let the co-
ordinates in the plane on which the sphere is stereographically
projected be By considering the projections on the
, v\.

coordinate axes of the line through a point ( 77) of the plane,

its corresponding point (X, Y, Z) on the

sphere and the point
(0, 0, 1), we derive the relations ^:^:1=X:Y:1 Z, where
X +Y +Z
2 2 2 =
1. From these we get the equations of the
transformation

It will be most convenient to

get the formula for distance in
the form of an integral. Let P(x, y), Q(x + Ax,
y + Ay) be
neighbouring points in the xy-pl&ne, P'( ^), Q'(+A i/ + Ai/)

being the corresponding points in the ^-plane. Substituting

2/ + Ay, 1) for (x lt y v %) and (x 2
(x, y, 1) and (x+Ax, , yz ,
z2)
respectively in the formula of Section 34, we have, e now
being 1

+ y 2 - 1 + xAx + y&y - V{ [x&x + yAijJ2 - [x2 + y2 -

'
2 2
'

1]
[(Ax) + Ay) 2 ] }
(

Dropping infinitesimals of higher order than the first, this can

rit ten

Dist.PQA;lc

1L- / / -
2
+ (Ay)- )

(?^-*^y\* (^)

"
"
\
\ 2 9 1
^~
T
~^" .
36,37] XOX-EUCLIDEAX GEOMETRY 69

From the form in which we have written the last equation

we verify at once from the equations of the transformation that
Dist. PQ = 2
If now we use the customary complex variable z = g+iij, this
becomes, putting 2k = 1 A: and this is the value to be used
, >/ ;

for the infinitesimal distance P'Q'. Tic Ic-mjtl of << curve in tl< :

gri-plane /x the ralue of the integral I

\dz /tj
taken along tli<>

curve * and the shortest distance between two points

;
A and B
/B
is the integral I
\dz\ />/, the path of integration being the circle

through the points orthogonal to the f-axis.

A consideration of angle shows that, as in the elliptic case,
anglefias its ordinary meaning.

37. Application to Groups of Linear Transformations. Let

us return to the Cayleyan plane. We shall call a transformation
of the plane by which both distances and angles are unchanged
a displacement. Since distances are to be unchanged, an
infinitely distant point must remain at infinity; that is,
the is transformed into itself.
absolute It follows also that
lines aretransformed into lines; for if P, Q, R lie on a line,
Dist. PQ-f Dist. t)H = Dist. PH. and \ve shall have for the trans-
(

formed points Dist. P'Q' + Dist. Q'R' = l)ist. P'H'. which is true
only if P', Q', R' are collinear. The displacement is then a
collineation
*
_ a^x + b^y + r }
,
_ a.,.r -f /'.,// + c.,

The invariance of the absolute imposes five conditions on the

coefficients. A collineation leaves invariant the cross ratio of
four points on a line or of four lines through a point and we ;

observe from the definitions of distance and of angle that the

collineations which transform the absolute into itself preserve
distance and angle save possibly I'or sign.
Xow let us consider the representation of the hyperbolic
plane in which the straight lines are represented by circles

orthogonal to the real axis. \Ye found that the absolute is

transformed into that axis. The displacements are the directly
*
Poinoiiiv, Ad". M'lt/ie/tiatica, vol. i.
pp. 7, 8.
 70 ATTOMORPHIC FUNCTIONS [CH. V

conformal transformations which leave the real axis invariant

ami transform circles orthogonal to the real axis into circles
orthogonal to that axis; that is, the linear transformations.
Tin- inversely conformal transformations leaving the absolute
invariant are the inversions in circles orthogonal to the real
axis. They preserve lengths but change the signs of angles.
Th- transformation amounts, to use the Euclidean analogy, to

rotating the plane through 180 about one of its lines.

In this hyperbolic geometry, the fundamental region of a
Fuchsian group is a rectilinear polygon. The transformations
of the group are all displacements. The various transforms of
the fundamental region which partition the plane are equal in
all respects, and any one can be brought into coincidence with

any other without alteration of shape or magnitude Certain

facts of previous proofs become obvious for example, that only
;

a finite number of regions will be contained within an area not

nding to the real axis; and that the regions will cluster
in infinitenumber along the axis.
In the elliptic case we projected the plane on the sphere in
such a manner that great circles play the role of straight lines.
The displacements are the rigid motions of the sphere; the
transformations which preserve lengths but change the signs
of angles are the reflections in diametral planes. [It is easy to
show that in the transformation we made, the absolute is
projected into the circle at infinity, which is unchanged by the
rotations and reflections mentioned.]
The three kinds of triangles that .arose in connection with
the triangle functions can be interpreted in the light of the
three kinds of geometry. In each case, all the triangles of a
network are equal to each other in magnitude when the
appropriate type of geometry is used. When X + a + ^ = l the
/

^i im of the
angles of the triangle is two right angles and the
Euclidean (fig. 19). When X + /* + < 1 the sum of
ii<-try is 1'

less than two

right angles, and the geometry is
tin- angles is

hyp.-rbolic (fig. 20). 'When \ + a + i/>l the sum of the angles

/

ber than two right angles and the

geometry is elliptic
<li--. 2:j) In each cast' a shaded can be brought into
triangle
coincidence with any other shaded triangle by a displacement
or into coincidence with
any unshaded triangle by a displace-
ment together -A-ith a turning. over.
37] XOX-EUCLIDEAX GEOMETRY 1

The Fuchsian groups furnisli us \vith discontinuous groups

of collineations which leave a real conic invariant. For, each
transformation of the group can be expressed as a transforma-
tion of the points in the Cayleyan plane by means of the

equations defining the trans-

formation from one plane to
the other. The circular-arc
fundamental region in the
complex plane is transformed
into a rectilinear polygon in
the Cayleyan plane, which is
a fundamental region for the
group of collineations. The
division of the plane for such a

group of collineations is shown

in tig. 2(5. The angles of each

triangle arc vr/S, Tr/3, 0. The

figure is. in fact, the division
for the modular group carried over to tin- (
'ayleyan plane by
means of the <Mjuatims of Section '-}(].
It is possible also to discover independently discontinuous

vroups of collineations which leave a certain conic invariant,

and from these to derive new discontinuous groups of linear
transformations. Studies <if the subject from this point of view
have been made by Poincaiv and by Vrlc
*
Poi: Fricke. M.iil,. Ann., 38 (1890),
50, 461.
 CHAPTER VI

UNIFOBMISATION

38. Riemann's Surfaces. of greatest simplicity The functions

functions; to such the simple
in analysis are the single-valued
theorems of the theory of functions can be applied with-
out reservation. In the case of functions which are many-
valued, there are complications owing to the fact that it is neces-
sary to deal with a particular one of the many branches of the
function. We shall recall here a device introduced by Riemann
in which the plane independent variable is replaced by a
of the

multiplicity of planes in such a manner that on the resulting

configuration the function is single-valued.
Consider first the function ?/;
2 = ar. Each value of x, excepting the points and
oo yields two values of w.
,
Let x = p e ie be a point in the K-plane, p being the
distance from the origin to x and the angle between Qx and the real axis. The
two values of w are iv l = Vpo e ^ oy/2 an d wz = -w l = \/p eie() ^ +7ri Let us fix our ()
.

attention on w l while x is varied. If x describes any closed path not encircling

the origin, both the distance from the origin and the angle return to their original
values; that is, w returns to the value wlt If, however, x passes once around the

origin, the angle is increased or diminished by 2-jr, and w l acquires the value iv 2 ;
another circuit of the origin brings the value back to Now let us superpose on the w^
ce-plane a second as-plane infinitesimally near it, each point of which represents the
value of x as the point immediately below it. Let the two planes be cut along
curv<^ extending from the origin to infinity, say along the negative half of the real
loin the part above the axis in each sheet to the
part below the axis in the
it, In the resulting surface a moving point a; passes from one sheet into
tin- other on crossing the negative half of the real axis. Now let us consider the
Lou "f //-, starting with the value w^ at x in the original plane. If a; makes a
'

i.
.muling the origin, either ./; remains always in the original plane, or
it crosses the negrtive part of the real axis into the second plane it recrosses into
it

the original plane, and returns to x with the If, however, x

original value w^.
the origin,
it returns to x in the second
rouit ul'
plane with the value ?/-2 .

path x makes returning to x the function has the value w 1 if

ii the original plane, and //-, if./: is in the second On this configuration,
plane.
UK-
function /* sini/le-m/itcd. The points and oo at
tin- line ,,T iiinetiire ,,f thi- sheds terminates are called branchpoints.
72
38, 39] UXIFOKMISATIOX 73

This simple example exhibits the essential features of the

Riemann's surface for the c5
general alofebi^aic
O function. Let
w be defined as a function of x by the irreducible* equation
$(w,x) = P (x)w +]' + P (itO = 0, wliere P,(.v) is
m
'+ . . .
M(

a polynomial in -x. For a given value of x the coefficients P. are

fixed and the solution of the equation gives values of w. ///

The fact is, as the preceding example would lead us to expect,

that w is a single-valued function of x on a Riemann's surface
composed of in sheets joined along certain lines terminating in
branch points. Similarly the same equation detines x as a
function of w single-valued on a Riemann's surface covering the
the number of sheets
i(;-plane, equalling the degree of the
J -

equation in ,''.

39. Genus. The study of tin- algebraic equation into which

a given algebraic equation <|>< //. .') = () can be transformed lias
led to the important conception of genus. If we make a projection

.
.'_
'
+c ]
+ bz x + f2
~a s w.+ b yc + Cs
there are certain numbers connected with the curve J which are

unchanged. For example, tin- d.-gn f the transformed curve,

its class, and the number of its double
points are the same as in
the original curve.
(
1

onsi<ler nowmost general kind of transformation by

tin 1

which each point of the given curve is transformed into one and
only one point of the transformed curve, and no two points of
the given curve are transformed into the same point. In other
words, the points on the Riemann's surfaces of the two curves are
to correspond in a one-to-one manner. It is known that the
most general such transformation is one in which lh- trans-
formation from the points on one curve to the points on the
other are expressible rationally /''= H,( \ .<). ./'= K./ </. .o anl
:

W= R,'( /''. .''). x = R.,'( '''. .''). Owing to its form this transforma-
tion is called birational.
*
/.'.. nr.t l)iv;ikii).,' up into factors of lower degree.
t There exists a Ricnuuin's surface for any function ic=f(x) but tin- nnn- ;

algebraic cases are characterised by the fact that the number of sheets, either for
'"'=/(*) or for the inverse function x=f- (u-) is infinite.
l
t

% It helpful to interpret our statements by

is means of plane curves :
although in
general the coefficients in the equation *(/', z) = arc complex and the curve has no
real locus.
 AITO.MOKI'IIK
1

FrXcTIOXS [CH. VI

:iat it be possible to solve the equations

?/:'
R,( =
rationally in term of w' t .-'. T>ut it must be possible to do so by
connected by the algebraic equation 4>(f.
ire = 0. The ,/.')

;s are not the general inverse transformation, but an

::iiati>ii which holds only for the points on the two curves. An
11 make this eleurer. Consider the curve iv'- = .>- - and the trans- ]

.-',

,'- which transforms it into ?/-' =

formation w'
4 - 1 The inverse
,
.''
,/'(/// )-. general
;i,,n ic=tc', .'' =
but if we make use of the equation
\J.i'' is not rational ;

of the curve the inverse transformation can be written rationally thus: w = w',

-1). The transformation is thus birational. An exceptional case occurs

for this transformation if the given curve is symmetrical with respect to the ay-axis,
Then two points on the given curve yield the same point on the trans-
formed curve. The transformation is not birational we are unable by using the
equation = .,- to get an inverse function involving only rational functions.
If the inverse is rational without using the equation of the curve the transforma-
tion is called a Cremona transformation ; for example, w' = x!w, x' = l/x, which has
the inverse w;=l/</-V, .'' = !/.//. The Cremona transformation is thus a special case
of the birational transformation.

\Ve readily find that in the case of the birational transforma-

tions neither the degree, nor the class, nor the number of double
points is
unchanged. There is, however, an important invariant
number. Let n, n be the degrees of the original and the trans-
formed curves, and d, d' the numbers of their double points.
In all cases (n - l)(n'-2)/2-d' = (n-l)(n -2)/2-d This in-
variant number, (n ~L)(n %)/2 d, is called the genus* of the
algebraic relation. It is usually represented by the letter p. The
maximum number of double points which an irreducible curve of
the //.tli
degree can have is (n l)(n 2)/2 hence the genus of a ;

curve is the number of double points by which it fails to attain

this maximum.!

Connectivity of Surfaces.
40. We can take an entirely
differentview of the genus of an algebraic equation, and one
more useful for our purposes, by considering the properties of
tip' liiwniann's surface quite apart from the equation connected
with it. It is necessary first to state some general results con-

Ing what termed the connectivity of a surface.

is

A surface said to be connected if we can join any point of

is

to any other of its points by a curve lying entirely

iin the surface. For example, the surface of a sphere, a
v it is
e^mm-mly called dendi'iicij. Tin- French term is <

''(. Genus was intn. lured in Uiemann's Tlir<>ri<' ilc-r AbeVschen

6 1
).IThe term ,/r/h'ic/ir// is due to Cayh-y.
!

!
.Multiple |i..ints higher i>rdr than simple double points are to be counted as
<.f

amber <>r double points,

39 40]:
UNIFORMISATION 75

plane surface bounded by a circle, a plane circular ring bounded

by two concentric circles are connected. We shall assume that
the surface has a boundary if it is a closed surface, such as the
;

sphere, we shall make a small hole in the surface to provide an

initial boundary.*
Now let us cut the surface along a curve beginning and end-
ing in the boundary. If, however this cut be made, the surface
is divided into t\vo pieces, the surface is said to be simply con-
nected ; otherwise it is multiply connected. The first and second
of the surfaces just mentioned are simply connected the third ;

is
multiply connected, since a cut extending from one boundary
circle to the other does not cut the
surface into two pieces. If one cut
renders the surface simply con-
nected, the surface is said to be
(Jut/h/i/ connected', otherwise we
can make a second cut beginning
and ending in the boundary (which
now includes the two sides of the
first cut). In general, if n cuts
an- necessary to make the surface

.simply connected, the surface is

(//+l)-ply connected, or is <>f con- FIG. 27.
'

it
'

n n-\-\.
It is clear that if a surface be continuously deformed in ,-my
way, or be transformed in a one-to-one and continuous manner
into any other surface, its connectivity is
unchanged; for the
cuts which render it
simply connected will be transformed into
cuts rendering the new surface simply connected.
can be proved without difficulty that a surface with a
It

single boundary, such as a closed surface with a small hole in it,

odd connectivity. We can then represent its connectivity
I*

by 2/;-}-l. where f> is an integer. 'The number p is called the

r/^x.s of the surface. It can be shown that a surface can be
transformed in a continuous and one-to-one manner into any
other surface of the same cla
*
\\V shall not consider in this section the so-called one-sided surfaces. The often-
quoted example of such a surface is that formed by a strip of paper whose ends are
joined after one of them 1ms been given a twist through 180.
t In fact, a surface can be transformed into any other of the same connectivity
if
they have the same number of boundaries.
76 AUTOMORPHIC FUNCTIONS [CH. VI

The simplest example of a surface of class p is that of a

clust-il surface with p holes through it. In tig. 27 is shown such
,-i surface of class 3. is at A
The and the six
initial boundary ;

cuts which render the surface simply connected are shown.

Now let us consider the Riemann's surface in the light of the
preceding remarks. To avoid considering the points at infinity,
let it be projected stereographically on the sphere. The surface
is a closed surface consisting of several spherical sheets ;
and it

can be shown that the class of this surface is equal to the genus
of the, algebraic relation connected with it* In view of this
fact it becomes evident intuitively why the genus is unchanged
by the birational, or one-to-one, transformations.

Genus of the Fundamental Region of a Group. In

41.
Section 21, Theorem F, we found that two functions, say
w and x, automorphic with respect to a group are connected
by an algebraic relation $(w, x) = 0. Each point z of the funda-
mental region yields a pair of values w(z), x(z) satisfying the
equation and in general no two points of the region yield the
;

same pair of values. In other words, each point of the funda-

mental region corresponds to one, and only one, point of the
Riemann's surface and conversely each point of the surface is
;

derived from one, and only one, point of the fundamental region.
Since w and x are automorphic, congruent values of z correspond
to the same point on the surface. Thus two congruent edges
of the fundamental region correspond to the same curve on the
surface.
Let us now form
the fundamental region into a closed surface
by bringing congruent edges together, each point being brought
into coincidence with its congruent
point. We thus get a closed
surface which corresponds point for point with the Riemann's
surface. Tlie genus (class) of this surface is the same as that
<>f the Riemann's surface. We define the genus of the funda-
mental region to be the genus of this closed surface formed by
In-! ng ing its
congruent edges together.
4
In fi^'. J8 w<; give an example of this process. In (I) is shown a fundamental
-'.rucnt sides being connected
by arrows. Bringing the sides B and H,

*
For proof of this fact in an important special case see Forsyth, Theory oj
tin;

'funs, 3;"'."'.
p. An excellent treatment of the connectivity of surfaces will be
I'MIIIM! in
Chapter XIV. of the same wrk.
40-42] UNIFORMLSATIOX 77

and D and F into coincidence, we get (II). A and E are now closed curves, and it is
necessary to bring the opposite ends of G and C together, as in (III), so that the
remaining sides can be joined without tearing. Making the juncture, we get the
surface in (IV), which is of genus 2. We note that the eight vertices form a cycle ;

and these are brought into coincidence on the surface.

Since the points of this closed surface formed from the

fundamental region correspond in a one-to-one manner with
the points of the Riemann's surface for $(?(;, a?) = 0, and since
neighbouring points on one surface correspond to neighbouring
points on the other, it follows that this new surface can be used
as one on which to represent the pairs of values (w, x) satisfying
the equation. In other words, it would serve as a Kiemann's
surface for the algebraic equation quite as well as the original

(01)

(D

-
cr (IV)

Fiu. -JS.

surface composed of several plane sheets. Viewed in this light,

\ve see tl^t the fundamental region />r/r<' hrini/ julncil into a

s-n,'fin.'f
is ////- Riemann's sv/rfaci dissected into a simply
region, or <it lt'<i*f partially dissected. [The funda-
mental region is not always simply connected. See fig. 12,
where some further cuts are necessary to make the region simply
connected.] This use of the fundamental region as a dissected
Iiiemann's surface is particularly useful because of the fact that
both iv and x are single-valued analytic functions of the variable
: of the fundamental
region.

42. The Problem of Uniformisation. The theory of single-

valued functions is both simpler and more developed than that
of many- valued functions and if we can express each of the
;

variables of the many-valued functional relation as single-

valued functions of a third variable a substantial contribution
 ATTOMOI-JMIIC FTXCTIOXS [CH. VI

ly of the function is made. This is the problem

<>!' uniformisation.
Let </ = /'(./) be a many-valued function.
7
\\ e wish to
express
./ and M; as single-. valued functions of a third variable, x = </>(:),
ty=s ;.-li tliat for each value of z for which both functions
di'tined the corresponding values of x and w satisfy the
equation "'=/(''), and each pair (w, x), satisfying the equation is
^ivi-n by one point at least. Looked at from the standpoint
urve theory, it is the problem of expressing the curve in
parametric form by means of single-valued functions. From
an"thrr point of view it is the problem of finding two single-
valued functions which map a portion of the s-plane in a
01 n - to-one and generally conformal manner upon the Riemann's
i

surface of the functions \vf(x).

We consider here particularly the algebraic curves. It is
well known that those of genus can be uniformised by means
of rational For example, the two-valued function
functions.
w =l 2
uniformised by the functions x = 2s/(l-}-z 2 ), w
x'
2
is

(I z-)/(l+z*). These functions map the whole 0-plane upon

the two-sheeted Riemann's surface.
The curves of genus 1 can be uniformised by means of
doubly periodic functions. We can employ functions of the
form x = (f>[p(z), $>'(z)], w = \ls[@(z), /(], where p(z) is the Weier-
strassian function and where $ and are rational functions i//-

of their arguments. The functions map each of the period

parallelograms on the Riemann's surface. [See fig. 7. We
rve that by joining congruent edges of the period parallelo-
gram the resulting surface is of class 1.]

Uniformisation by Means of Automorphic Functions.

43.

brought out in preceding sections lead us to hope that

Tin- facts
th- general algebraic curve can be uniformised
by means of
simple automorphic functions. For we found that between
i'-t ions,
ii !'i
w(z),x(z), there exists an algebraic relation
1 1

=
0; and this is obviously uniformised by the functions
But there was no indication as to whether
'
:
).

all al-vl,!- a ic relations could arise in this

way. It is the fact,
however, that there are no exceptions. It has been established,
eneral existence. theorem, that
.). -Aiii/ <i/ /rl>ric curve can be unifwmised by
;
42, 43] UXIFOKMLSATIOX 70

, t s <>f Fi ian
i. ucfi o ns,
<:!< .s c fwnda mental
ft. v</ o
>.
of tln>/ // i i ? i
,

group t<> which the function* bdon<j lying within the princ'tnal
circle.
Also a much more general theorem has been established.
Let iu = f(x) be a multiple-valued function of any kind. It has
been proved that it can be uniformised by means of automorphic
functions : that
THEOREM K.The
motf general many -valued function can
be uniformed by mean* of automorphic functions <>f ijrou i>
with a principal circle. The functions ha^c ofl/cr singu- i,

larities than poles within the principal andthecircl

a natural boundary for one at least of the functions.
Let us examine the statements in this theorem. If the
function to be uniformised is transcendental, it is clear that

the uniformising functions cannot be the simple automorphic

functions studied in Chapter III., for these functions are con-
nected by an algebraic relation. Since the uniformising functions
have only polar singularities within the region, it follows that
the fundamental region of the group contains portions of the
principal circle on its interior, or at least has vertices on the
principal circle, where one at least of the functions lias singu-
larities other than poles. It is necessary, in fact, in establishing
the theorem to admit the single transformation :' : as a
possible
group for this case, and the fundamental region for this group
is the whole plane.
The proofsof these theorems, while simple enough in con-

ception, involve considerable detail in the execution and they ;

cannot be given here. So far as I am aware, the only com-

plete and simple presentation is to be found in Osgood's
FunldionentJieorie.*
The automorphic functions of the types just mentioned are
not the only ones which can be used in the uniformisation of
functions. \\V shall presently give an example of the uniforini-
sation of algebraic functions by means of functions whose group
possesses a fundamental region not confined to the interior of
* The treatment is based on a series of papers
2nd
Vol. i. , ed., chapter xiv.
by Koebe in the and in the Mat.Jtemaiische Annalen,
G'Jt(in.i/i'.r X'-hri>'hten, 1907-9,
vols. 67 (1909) and 69 (1910). Further papers by Koebe have since appeared in
the latter publication. It was Klein who first saw the possibility of arriving at a

general existence theorem. Cf. Math. Ann., 21(1883), p. 162, and Abschnitt iii..

10.
 AUTOMORPHIC FUNCTIONS [CH. VI

the principal circle.The uniformisation can also be accomplished

l.v functions whose group preserves no circle invariant. But
functions of the kind mentioned in the theorem would seem to
!)> usually the most simple. There is one advantage in the fact
that a fundamental region within the principal circle is always
simply connected.
The advantages of uniformisation are many. In the study of
functions on a Riemann's surface functions of the form f(iv, ./),

where and x satisfy an algebraic equation &(w, x) =

tc the
are lacking in simplicity owing to the necessity of
investigations
considering the many exceptional points. The points at which
w or become infinite involve one kind of series expansion, the
,/

1tranch points another, both differing from that at ordinary points.

On the fundamental region in the plane of the uniformising
variable these exceptional points do not occur, for the funda-
mental region has no infinite points and no branch points.
Not least among the simplifications introduced by uniform-
isation is to be found in the treatment of Abelian integrals

integrals of the form IR(w, x)dx, where $(iv, x) = 0, and R is a

rational function. To take a simple example from the calculus :

the integral Idx/

1

J(l x2 ) that is, Idx/w, where w =lz

x2

is most easily found by putting x = sin z,w = cos z in other ;

words, by uniformising the curve u> = 1 x by means of auto-

2 2

morphic functions.
In case the genus of the surface is greater than the
integrals are always infinitely many- valued on the surface;
when certain closed paths are described, the integral is changed
by the addition of a constant. The integrals become single-
valued (provided they have no logarithmic singularities) when
the surface is dissected into a simply connected piece. But in
the plane of the uniformising variable where the fundamental
ion is simply connected these integrals are single-valued. A
path on the Riemann's surface by which the integral returns to
its original position with an added modulus
corresponds to a path
in the plane of the uniformising variable which terminates, not

in the point of departure, but in a

point congruent to it. The
int^ral is then single-valued function which changes by an
,-i

additive constant (which may be zero) in passing from

any point
 48, 44] UNIFORMISATIOX 81

to a congruent point. There is a marked gain in simplicity. It

seems probable that automorphic functions are destined to play
an important rule in the theory of algebraic functions and their
integrals.

44. We turn now to the question of

Whittaker's Group.
actually effecting the uniformisation of algebraic curves by
means of functions connected with specific groups. Groups of
the type discussed in this section were introduced by Whittaker.*
The groups are constructed from transformations S which
have the property that S?= 1 called .W/- verse transformations).
( /'//

Let S p S.,, S 3 S 4 S 5 S G be real sclf-invrrsr transformations satis-

, , ,

c,

the equation SgS =1. This equality gives much

freedom of choice, and it is not difficult todetermine such trans-
formations.! The trans formations are elliptic in type, the fixr<l

points being conjugate imaginaries. 15 .,

Let D . . D6 be the
points lying in the upper half-plane.
tixr.l Consider the trans-
forms of the point 6
D
when the transformations S 1? ., S 5 are
. .

made in succession. Since 8^ = 1, S x

rotates a short line through

Dj through an angle TT. Hence the circle orthogonal to the real

axis through l^Dy is transformed into itself, D 6 being trans-
formed into a point Cj on the opposite side of Dj (see fig. 29).
Similarly S., carries C x
into a point G>, the circle CjDgCg being

orthogonal to the axis. Continuing, we get C3 C 4 C 5 , ,

. But
*
Phil. Trans., 192 (1899), pp. 1-32.
t Whittaker gives as an example the transformations
- - - - - - 5)], [c, - -
[z, (5z 74)/(c 5)], [z,(2z 2)], [=, (5z
5)/(= 29)/(s (253.: 2061)/(33s 253)],
~-132)], [2, (281a-4786)/(17-281)].
6
 ATTOMOKl'HIC FUNCTIONS [CH. VI

- -;
i S 1
= S 6 ~ = S6 1
a transformation which leaves the point
,

D 6 unchanged. Hence C 5 coincides with D G and we have the ,

.1 circular polygon shown in the figure.

This polygon is in fact the fundamental region for the group
who>.- ^.'in-rating transformations are S 1} S 2 S 3 S 4 S 5 There , , ,
.

an- five pairs of congruent sides, DjDg and D^, DgC^ an(^ DC 2 2 ,

Tlie generating transformations satisfy the equations

S* = l, . .
., S 52 = l, and (S^S^SJ^S^I. From the last
equation we see that the sum of the angles at the congruent
vertices D Cv (Section 17).
6 , By bringing the
. .
., C4 is TT

congruent sides together to form a closed surface we see that

the genus of the fundamental region is 0.

Now let us consider the automorphic functions connected

with this group. It is known that when the genus of the funda-
mental region is there exist functions having only one pole in
the region and that in terms of any one of these functions all
;

the automorphic functions of the group are expressible rationally.

Let x be one of these functions, and we shall suppose for con-
venience that its pole does not lie on the boundary. [We can
place by using a function of the form
the pole arbitrarily
(ax+b)/(cx+d), which also has but one pole.] According to
Theorem E, Section 21, x acquires any given value once, and
only once, in the fundamental region. In other words, the
function x(z) maps the fundamental region on the whole #-plane.
Let us examine this mapping. Let e v e 5 be the values . .
.,

of x at the points D I} D 5 and e 6 its value at the congruent . .

., ;

points D G C v C 4 We shall start with the point D 6 and,

,
. .
.,
.
,

tracing the boundary of the fundamental region in a counter-

clockwise direction, find the corresponding locus in the #;-plane.
: moves from D 6 to D x
x moves along some curve from e
Q
to
'-,.
Now D 1
constitutes a cycle of angle 2?r/2 and, according to ;

Tin -on -in A of Section 21, x takes on its value twice there. Hence
.' -e = (z-D^f\^(z), where ^(D^^O, and
l the angle TT at D l is
mapped on an angle 2-rr in the cc-plane. Since x has the same
values along the side as along D 6 D 15 as z moves along C^ D^
curve just found from e 1 back to eQ
tin; C l also .

I..-IOHM-S to a cycle of
angle 2-7T/2, and the angle at Cj is mapped
"ii .-in ,-ni-lr twice as great at <>
n
. The side C^Cg
mapped on is
curve iroiu ,-
(
.
to e 2 and back again, and so Thus we find
on.
//,<
function x(/) nui/>s the fundamental region on tin
44] UNIFORMISATIOX 83

x-plane, the boundary of the region being mapped on jive curves

extending from the points e v ., e 5 to the point e

. .
6 (see fig. 30).
But this group, being of genus 0, will uniformise only the
curves of genus 0. We
get a group of greater genus in the
following manner. Adjoin to the region the region congruent
to it by one of the transformations, say S r Erase the common
side D 6 C P and the new region is a polygon of eight sides, the

opposite sides being congruent by the transformation S r The

side C 4 D 5 D 6 for example, is transformed into C 1 D/C (see fig. 29).
/

The transformation S 5 followed by S x transforms C 4 D 5 D 6 into

C'D/Cp and transforms the new polygon into one adjacent to
it along the side
C'D/Cp A similar treatment of the other
sides shows that the double polygon
is the fundamental region for the
group whose generating transforma-
tions are S 2 S 15 S 3 S 1} SJS^ S 5 S r The
eight vertices of the polygon form a
single cycle of angle 2-Tr. We find
that /// /N I'<'<I'K>II /N
of genus '2.

This group is a subgroup of the

preceding, since its transforma- all
tions belong to the former group. 30
Hene<- ": is an automorphic func-
tion of the new group; it acquires a ^iven value twice in the
double polygon. It maps the polygon on two planes joined
together along the branch lii: Joining congruent edges
of the new fundamental polygon to form a closed region amounts
to joining the two re-planes along the remaining branch lines.
Consider a function w(z) automorphic with respect to the
new group. The functions w and x are connected by an algebraic
relation 3>(iv, x) = Q, which is of the second degree in w, since to
each value of x correspond two points z of the region and hence
two values of iv. The two-sheeted surface just found is the
Riemann's surface for <fr(iv, x) = 0. Consider in particular the
function w=
J^x e^x e^x e^x e^x e^x e^)}. This
function is single-valued on the surface just found, and is hence
an automorphic function of z with respect to the second group.
Hence the automorphic functions of this group enable us to

uniformise the curve

?0
2 = (x - e^}(x - e 2 )(x - e )(x
3
- e )(x
4
- e b }(x -e
(] ).
 ArTMMolti'JIK' FUNCTIONS [CH. VI

\\V can of Theorem G, Section 21, that the

show by means
nil autoinurphic functions of the group are rational functions
and X.

The va! pend upon the function -x(z) and upon

ip.
and are not all at our disposal. The question that
If is: e1) If (x e 6 ) be
the function w = (x
z
. . .

arbitrarily chosen. does there exist a group of the type we have

di-cu^ing
i whose automorphic functions effect its uniformi-
aation As an indication that the uniformisation can always be
made bv these groups, it will be shown that the constants at our

disposal are equal in number to the conditions to be fulfilled.

Hach of the transformations S 15 ., S 6 contains three
essential . .

~
n-al The condition S = S l imposes one condition on
constants. i i

[namely, that a =
i d (Section 1)], leaving twelve constants.
The relation S 6 S 5 S 1 = l defines three of these in terms of
. . .

the Also this group is not essentially different from one

rest.

which is obtained by transforming it by any real transformation,

which shows that three more of the constants are non-essential.
So there are altogether six essential real constants at our dis-
posal in forming the group.
Now considering the #-plane, there are six points e v .,
e6 . .
;

and each of these is defined by two real coordinates, giving

>lve real constants. But if we make a linear transformation
of the .r-plane, we can transform any three points into the points
. <. . : which shows that six of the constants can be dis-
>rde(l as non-essential. So we have six essential constants
in the ./--figure. Hence, the number of essential constants is the
same the -.-figure as in the ^-figure, which shows that the
in

4>s have the generality requisite for the uniformisation.

is known that
It
any algebraic relation of genus 2 can be
deformed bi
rationally into a relation of the form w = (x
2
e1)
so that the groups in
question can be applied to
my <il</('l)rnic relation of genus 2.
Th.->e -mil ps can be generalised without difficulty. By
iderino -2y>4-2 self-inverse transformations and
forming the
double polygon as l.efore, we should be able to uniformise the
ion

,-

whic
44, 45J UXIFORMISATION 85

45. Weber's Group. In the group employed by Weber*

for purposes uniformisation an initial region bounded by
of

non-intersecting circles orthogonal to the real axis is considered.

For simplicity we shall take the case of three circles (C 1; C.,, C 3>
fig. 31); the region to be considered lies outside the circles.

Fio. 31.

From Schwarz's researches in the general theory of con-

formal representation f Weber shows that there is a function
with the following properties:
(1) x is analytic within the given region except at infinity.
win-re it has a simple pole and is continuous on the boundary
;
:

x is real on the circumferences of the circles;

(2)

(3) x acquires each value once in the region, and only once
save for points on the boundary.
It follows that the function ./( ;) maps the ivgion bounded by
the circles on the whole /'-plane. Let us see what the boundary
,

becomes. Since x is
real along a bounding
^' ^ ^ ^ ^ 5 ^
oii-flr. as ; traces the
circle
i

.'
iii
moves along the
* IG - <^'

ival axis, returning to the initial value on the completion of

the circuit. TJte {)''<>' region /'* //"/* n><i/>j>c</ on the ^-plane

bounded l>>/ flaw >///* e^.,, e 3 e 4 e-e 6 (See fig. 32.) [It is,
.

easy to see that x can trace each slit only twice once in each
direction and that the slits cannot overlap without violating
property (3) for points near the bounding circles.]
Now let us add to the given region the region arising by
an inversion in the circle C 3 By a well-known theorem on .

*
l-'.ln
Ilritrixj ::ur /'oinrare's Theorie der Fuchs'chen Functionen, Gottinger
Nuchrichten, 1886, pp. 359-370.
n'ffxbcrir/tten <l?r Berliner Akademie, 1870.
 MoKl'HIC FUX< I
[CH. VI

Mtinuation * a function which is n-al along a circle

will, if continued analytically aero- tin- same real part

Binary part of opposite sign at points inverse with

'
cle. The circles C v C 2
invert into C/, C 2

along which the function ./'lias its values unchanged.

one of these circles, the imaginary part of x
,

again. anl the result of the two inversions is that

I

3
unchanged. Now, an inversion in C 3 followed
h\- an inversion in a linear transformation, Sr say, carrying
C/ is
'
uto (_\'. Similarly inversions in C 3 and C 2 are equivalent
transformation S 2 carrying C 2 into C 2 S x and S 2 '.

rating transformations of a group with the real axis

fixed circle whose fundamental region is the combined region
found. And x is an automorphic function with respect to
this group, which acquires each value twice in the region once
ach of the combined portions. The function x(z) maps the
>n on two planes joined along the branch line e B e corre-
id
ing to C3
If we bring congruent edges of the region
.

ther,get a we
closed surface with two holes through it; and
genus of the region is thus 2. Closing the region corre-
j
tli

i<ls in the #-plane to joining the two sheets along the

remaining branch lines.

\\ see, exactly as in the preceding section, that if

w' 2 = (x- 6l )(x

- e 2 )(x - ea )(x
-e
4 )(x
- e )(x
5
- e6 }

-
an automorphic function of z. Hence the automorphic
i Ihr <jr<>n i> en able us to uniformise this algebraic
lius 2.
\V.- observe that the quantities e v ., e are real and we . .
{
.
;

show by tlie count of constants that this group is suffi-

ral to uniformise
any algebraic relation of the form
. . . (' ',;>.
where f
v . . .,<> (
. are real. As in the
three of
these quantities are essential.
only
quantities are real, this imposes three conditions on
'

tin- group.
nied from three
circles with heir centres on t

Each has (wo decrees of freedom its centre

cirri,-
;

may be chosen arbitrarily, save for certain in-

bo ensure tlmt the circles be exterior to one

v.-l. i., L'lid cd., ]]). 066 ct seq.

4o. 40] UXIFORMISATION 87

another. There are thus six real constants at our disposal. But
three of these can be regarded as non-essential since the group is
not essentially different when subjected to a linear transformation.
We are left with three essential real constants for the fulfilling
of the three conditions imposed, which shows that the groups
have the requisite generality to effect the uniformisation.
We have made no use of the fact that the circles are ortho-
gonal to the real axis except to conclude that the group is
Fuchsian. [Certain other results that we have not mentioned
do follow from this fact for example, that ,; has the values
;

ev e Q at the points where the circles intersect the axis.]

. .
., ,
If
we remove the condition of orthogonality. we get a Kleinian
group of the Schottky type (to be mentioned in the next section),
and the automorphic functions of this group effect the uni-
formisation of the relation wz =(xe ) l
. . .
(.' e 6 )> where
e l} . .
.,
CG are real.

46. The Work of Schottky and Burnside. In an article

Journal *
in Crelles Schottky deals with groups derived in the
manner of Weber's, "but where the circles are not restricted to
those orthogonal to a circle. Each circle has an added degree
of freedom since its centre is not restricted to lie on the axis,
which shows that we have tlnve more constants at our disposal
in groups of the Schottky type than in those of Weber (for the
case of three circles). There are thus cc 3 groups of this type
which can be used for tin- unil'onnisution of the function
,r- = (./ e 6 ), where c v c 6 are real.
<?j) (x . . . . .
.,

Schottky makes a study of the automorphic functions and

the Abelian integrals connected with the groups, and obtains
infinite products for their expression.
Similar groups are studied by Burnside,f who makes the
discovery that in a large class of groups, including those of
Weber, the Point-are theta series (Section 22) converge for m=l.
It is then possible to get the Abelian
integrals by integrating these
series. Burnside sets up the various types of integrals in this way,
and derives their properties in an exceedingly simple manner.

*
Vol. ci. (1837), pp. 227-272. Uebcr einc speciclle Function, welche bei einer
:

bestimmten linear en Trans formation ihres Argument's nnreraudcrt bleibt.

t On a Class of Automorphic Functions, Proceedings of the London Mathematical
Society, vol. xxiii. (1892), pp. 49-88.
 A BIBLIOGRAPHY OF AUTOMORPHIC
FUNCTIONS
NOTE. Many particular classes of functions, e.g. the circular functions, elliptic

functions, and triangle-functions (including the elliptic-modular and polyhedral

functions), had been studied extensively for many years before the general theory of
automorphic functions was created. It seems undesirable to attempt here a biblio-
graphy of these earlier researches, and the present list will therefore begin with the
papers in which Poincare created the general theory. Mention should, however, be
made of one or two memoirs of the earlier period in which the general theory was
notably foreshadowed especially ;

SCHWARZ, H. A., Uber diejenigen Fdlle, in welchen die Gaussische hypergeo-

metrische Reihe cine algebraische Funktion Hires vierten Elementes darstellt, Crelle's J.,
75 (1872), p. 292.
SCHOTTKY, F., Uber die conforms Abbildung mehrfach zusammenhdngender ebener
Fldchen, Crelle's J., 83 (1877), p. 300.
FUCHS, Uber eine Klasse von Funktionen mehrerer Variabeln, welche durch
L.,
Umkehrung der Integrate von Losungen der linearen Differentialgleichungen mit
rationalen Coeffizienten entstehen, Crelle's J., 89 (1880), p. 150.

POINCARE, H., Sur les fonctions fuchsiennes, C.R., 92 (1881), 333, 393, 859,
1198, 1274, 1484 ;
93 (1881), 301, 581.
SCARE, H., Sur lesfonctionsuniformes qui se reproduisent par des substitutions
l'<>i

lintaires,Math. Ann., 19 (1882), 553 20 (1882), 52. ;

KLEIN, F., Uber eindeutige Functionen mit linearen Transformation in sich, Math.
Aim., 19 (1882), 565 20 (1882), 49. ;

POINCARE, H., Sur les fonctions fuchsiennes, C.R., 94 (1882), 163, 1038, 1166 ;

95 (1882), 626 (1883), 1485. ; 97

I'oiNCAKK, H., Theorie des groupesfuchsiens, ActaMath., 1(1882), 1.
1 r ii WIT/, A.
1 Ueber eine Reihe neuer Functionen, welche die absoluten Invarianten
,

gewisser Gruppen ganzzahliger linearer Transformationen bilden, Math. Ann., 20

(1882), 125.
RA.U8BNBBBGER, 0., Ueber eindeutige Functionen mit mehreren nicht vertausch-
M.-it.il. Ann., 20 (1882) 187 ; 21 (1883), p. 59.
TRY, F., Ueber eindeutige Functionen mit linearen Transformationen in
i

rich, Math. Ann., 20 (1882), 29:3.'

! i
us, L., Ueber Functionen, welche durch lineare Substitutionen unvcramlert
'

"li. (1882), 81.

r<>ixr.\Ki':, II. (
Mtmuir,- ,sv/r Acta Math., 1 (1883), 193.
les fonctions fuchsiennes,
I'' 1

AI:II, K., Sur une, classe de groupes discontinus de substitutions lintaires ct *nr
''/es
independantes restant invariable* par ces substitutions,
(1882), 297.
'1
ali., 1
 BIBLIOGRAPHY 89

POIXCARE, H., Si> an tl>- theorie generate des functions, Bull, de la

i' 'it

Math, de France, 11 (1S83) 112.

Soc.

POINCARE, H., Memoir*.- sur 'cs groupes klein&ns, Acta Math., 3 (1883), 49.
DYCK, "W., Ueber die durch Gruppcn lin>:<tr<:i- Transriirnii'tioueii gegebeneu
reguMren Gabietscii'ti'ifun'ji'n des Raumes, Leipz. Ber. (1883), 61.
KLEIN, F., Neuc Bcitrii<j< zur Riemanri'schen Functionentheorie, Math. Ann., 21
(1883), 141.
PICARD, E., Me'inoire sur les formes quadratiques li /mires indefinics, Ann. de
1'Ec. Norm. (3), I. (1884), 9.
POINCARE, H., Sur les groupes des equations Untaires, Acta Math., 4 (1884),
201.
PICARD, E. Sur un groupe des tra/i-
,
s des print* <1>- Vc space situt's dn
m'.me rMe d'un plan, S.M.F. Bull., 12 (1884), 43.
POINCAUE, H., Memoir:' sur les function* :>'tafuchsicnnes, Acta Math., 5 (1884),
209.
PlCAKi), M. -

fintt.'i
l >rit;,
L ues ttrnaircs ind/jiiili'* a fnd^n-niinces con-
jiujt'es
>t sur les functions />;/p' rfu>'/ts<'> -
:
">n<.l<int>'*, Acta Math., 5 (1S84), 121.
POINCAKK, II., Sur an ti M. Fuc/ts, Acta Math., 7 (1885). 1.

BIEIIMANN, 0., Zu.r Theni'i>' d> r Fn<-h*scU>'ii Fiinctinnen, Wien Sitzungsb. ,

92
(1885), 1137.
VON MANGOLDT, H., Uebcr ein Verfulireii :ar Dnr*t>'llini<i Uiptlscher MnduJ-
fnnrfionen durch, u/tend/ir/ Producte iie'i^t finer Ausdehn/ (
,

I'l-rfaltren-

allijeincinerc Fundioncn, Giitt. Xai-h. (1886), 1.

'

WEBER, H., Eir . r Fur/iS*<-hcn Fuiidiunen, Gott.

Nach. (1886), 359.
POIXCAKE, H., Snr ; >'S et la r>

des w --'Hews, C.B., 102 (1886), 41.

POINCAUK, H., ,s'/'/-
lesfonctionsfitchs -mes quadratiques lerimii-'-s
1

'.R., 102 (1886)

JI I.MI;KI;T, G., Appli'-ntiun de /" ti. nctions fuchsiennes a I'etude des
bes a/>j<:i>i-i<iitr.i, Liuuville's J. (4), 2, 239.
SCHOTTKY, F. Ueber eine s/)e>'i>'ll>- Function, wclclie b?i cinrr bestimmten H,
,

rmiitiun iln-i's Aiyimvxt'-* nmrri'i'mfcrt bleiU, Crelle's J., 101 (1887), 227.
PICARD, E., i>in*fr"/ion d'u// thetoems general tur les fonctions wntformt
J><.
:

par /'//'- i-fltifion ti?g<'l>ri</u>', Acta Math., 11 (1887), 1.

HUMBERT, G., Sur /< t/n'ornn" d'Ab'-l <t <j>'> /,/,/

s-i/,tes de ses aj>p/iration.s
tri'iues, Liouville's J. (4), 3 (1887),
POINCARE, H., Les fonctions fuchsiennes >'t
rriihm<'ti<iu<', Liouville's J. (4), 3
(1887), 105.
STOUFK, X., s>/r In tritnsformafion des functions fuchsiewnes, Ann. de 1'Kc. Norm.
(3), 5 (1888), 219.
STAHL, H. Ueber die Darstellung d>r eindeutigen Functionen, die sif/t duvh
,

liiiru.rs. Sn.listitatiunt'n r> pr<! u'-inn. di/rc/i iniendiichc J'rod,ucf>', Math. Ann., 33

(1888), 291.
BIANCHI, L., Sidle super fi etc Fuchtunie, Lincei Rend. (4), 4 2 (1888), 161.
I'II',M:I>, E., Sur les formes IIUH/! ro-tiqu* * Unuins a indetermintcs conjugut-es et les

fonctiunsfuchsinmes, Am. J. Math., 11 (1888), 187.

ScHLEMNdEi;, L., Zur Theorie d<r Fucks'schen Functionen, Crelle's J., 105
(1889), 181.
UK BRUN, F. ,
ll> -ris for m^tra f<orem "f Poincare, Ofv. af K. Svenska Vet.-Ak.
Foih., 46 (1889), 677.
90 AUTOMOIM'HK.: ITXCTIOXS

Hri:\vii/. A., V Differentialgleichungen drittcr Ordnitu//, welchen die

*fch geiiiigcn, Matli. Ann., 33 (1889), 345.
[irmipcs fachsiens formes avec Ics ratines <f equations
.
uloose Ann., 4 (1890), P.
FI:ICKI , K , Ueb '>' /> Klassediskontinuierlicher
Gruppen reeller linearer
Math. Ann., 38 (1890), 50, 461.
KIJ.IN, K.. Xnr T/H'on'erfcr AbcMn'it Functionen, Math. Ann., 36 (1890), p. 1.

I:IKIIM\XX, 0. ,
Ucbcr die Durst el lung der Fuchsschen Functionen erster Fcunilie
lucte, Monatshefte f. M., 1 (1890), 49, 3 (1892), 143.

KLKIN, F., Zur Theorie der Larndschen Functionen, Gott. Nach. (1890), 85.
KI.KIN, F. ,
Ueber Xnrmirung der linearen Differentialghichungen zweiter Ordnung,
Math. Ann., 38(1890), 144.
L, G. Ofvcr en ofhandling afH.
; :
,
Weber med titel : JEin Beitrag zu Poincare"s ' '

FwUSwlmi Functional," Bihang till Sv. Ak., 16 (1891), Afd. L, No. 2.

CASSEL, G. tiur un problems de representation conforme, Acta Math. 15 (1891), 33.
, ,

PAiXLEVf:, P., Memoire sur les Equations differentialles du premier ordre, Annales
del'Ec. Norm. (3), VIII. (1891), 9, 103, 201, 267.

BTKNSIDK, AV., On a class of automorphic functions, Proc. L.M.S., 23(1892),

49, 281.
RITTEI:, E. Die eindeutigen automorphen Form-en vom Geschlecht Null, Gott.
,

Xach. (1892), 283, Math. Ann., 41 (1892), 1.

KLEIN, F., Ueber den Begriff des functionentheoretiachen Fundamentalbereichs,
Math. Ann., 40 (1892), 130.
FIUCKK, R., jyeitere U/ifersuchungen ilber automorpbe Gruppen soldier linearen
Hutionen einer
Substitutionskoeffizienten Quadratwurzeln
Fariabeln, deren
Zulden enthalten, Math. Ann., 39 (1891), 62.
FKICKK, R., Ueber discontinuirliche Gruppen, deren Substitutionscocfficienten
ganze Zahlen eines biquadratischen Korpers sind, Gott. Nach. (1892), 268.
FKICKE, R. Ueber ein allgemeines aritJimetisch-gruppentheoretisches Princip in
,

Thcoric der automorphen Functionen, Gott. Nach. (1892), 453.

SdiLKsiNGEit, L., Sur hi theoric des fonctions fuchsiennes, C.R., 114 (1892),
1100, 1409.
Snii.KsiNCKii, L., Ueber die bei den linearen Differentialgleichungen ziveiter
<'i I'uflntfcndi'ii J'rimformen, Crelle's J., 110 (1892), 130.
BUBCKHAEDT, II., Ueber die Darstellung einiger Fiille der automorphen Primformen
durch peddle Thetarcihen, Math. Ann., 42 (1893), 185.
lu.i. PK/XO, P., Kui
'jri'/>jii Kleineani a due variabili, Napoli Rend. (2), 7 (1893),
123.
KITIKI:, E., 1H<: dutomorphen Formen von belicbigem Geschlechte, Gott. Nacli.
1893), 121.
I-KKKK, R., Ziu- gruppentheoretischen Grundlegung der automorphen Functionen,
Math. Ann 42 (1893), 564. ,

.):, R Ui-bcr indefinite

,
quadratische Formen mitdreiundvier Vcrdnderlichen,
h. (1893), 705.
Au = ke u sur une
r<'-qut!ii. de Ricmann Liou-
surface fcrmte,
, 9(1893), 273.
,., Knhri,-k,-lun<jett :n, Transformation filnfter und siebenter Ordnung
nitu-ninr^hd' Functionen, Acta Math., 17 (1893), 345.
P., Ueber al^'brniyrhe (J/tirlt ,<ngen -irineJien cindcutic/cn Functionen,
Mionen vn M'C// zulassen, Crelle's J., 112 (1893)287.
[DI, w., Not, on linear substitutions, Mess. (2), 22 (1893), 190.
 mnLIOGRAPHY 91

W.. Xufc <>n (lie equation y = x(x*-l), Proc. L.M.S., 24 (1893), 17.
,
z

SCHOENFLIES, A., Uebcr Kreisbogcnpohjgone, Math. Ann., 42 (1S93) 377.

PICK, G., Ucber das Formcnsystem cincs Kreisbogenpolygons vom Gcschlccht Null,
Math. Ann., 42(1893), 489.
BlANCHI, L. Sopra alcu/ic classi di gruppi di sostituzioni lineari a coefficicnti
,

cnmplcssi, Math. Ann., 43 (1893), 101.

KAPTEYX, W., Ecu />aar tlellincjcn van Poincare, Handel, v. h. 4de Nederl.
Xat. Congres (1893), 149.
FORSYTH. A. R., Tli'-ory of Functions of a complex variable, Cambridge (1893),
{second edition, 1900).
KNFLIES, A., Ud:r Kreisbogendreicckc und Kreisbogcnvicrcclcc, Math. Ann.,
44(1894), 105.
D'OviDio. E., SuUcfnn-.io.d Ti ne, Torino Atti. 29 (1894), 741
Rri'TEii, E., Di'' muttiplicativen Fur men auf algebraiscken Gcbihle beliebigen
'"rifj anf die Theoric der un.tomorphen Formcn, Math. Ann.,
44(1894), 261.
FI:ICKE, R., r <formaiionstlicori'' dcr itufomor/Jioi Functions,
Matli. Ann., 44(1894), 97.
RITTEU, E. Di' der aitiomorphen Fum-tiousn bci stetiger Abanderuny
,

ndamentalbereicht, Matli. Ann., 45 (1894), 473, 46 (1895), 200.

FRICKE, R., Eine Anwendung der //7<'l>j,->/-i>- nnf <l/> Substitutionsgrupp*
'

net ion- H, (lutt. Xai-h. (1894), 106.

KF, X., St. /

<cs, Toulouse
Ann., viii. (1894) D.
EL, <r., K :
!er ofvcr teorin for dc aiUomorfa funktionerna, Diss.
la (1894).

FILICKE. R. ,
J>; Math. Ann.,
44(1894),. :

KLEIN, F. ,
Ueber linxiri' Differentialgleichungen >lcr zweiten "rJnuny, Auto-
graili. Vorlesuii-'. Littiu^-n, 1894
(
Math. Ann.. 46 :
;

395 . 77.

FuK'KE, R., Ueber die Discontit r '///';>// rcrlli-r Uninrcn Sub-

h. (1895), 300.

BAKEI:, H. F., On a <->ri_u( automorphic function, Canilir. Pruc., 8 (1895),

322.
RITTF.K, E.. U-lec J: '. 1.1,1' dm in lelii'litijt-n iiljcbra-
Qtbilde, Math. Ann., 47 (1895), 157.
'

1'.. f'fi" r <(/iC' ic/i nmj'n

<;<>I;I>AN, Tzu-cigt-
Math. Ann., 46 (1895).
.

IJA<;NEK.\. F., Sid (corema de/l' csi$ten-,<i dcllc fun:i'ii Fnchxi"nc, Rivista di

Mat., 5(1896).
r
KKMI-INSK r Ftirl*fi<-h>' FuncUmi- /< -./i-fier J-
ari".i>c/n, Matli. Ann., 47
;. 57:; : Hull, intern, de Cracovie (1895), 288.
FiurKE, R. , Notis, bcr die Discontinuity Collineationsgruppen, Math.
Ann., 47(1896), f.57
FIUCKE, R., Ucbcr die T :
<it(t<>TH'>r/ih>-n J/r-rf//A/''"/V"" < ; '"(
tt. Nach.
(1896), 91.
FKICKE, R. ,
Die Theoric der auto, nor/then F<iitrti<>/n.'it und die, Arithmetik,
Papers published by the Arner. Math. Soc., I. (1896), 72.
SciiLEsiNOKii, L. , Ha<tdl>ii.cli "'er T/< ">//'- d> r lincaren Differenlialgleichunyen,
2 Bde ., Leipzig (1896).
VITERBI, A., Lc c<ji'.a'.i/i/ti diffcrenziali lineari a coefficients algcbrici inteyrabili
 AUTOMORPHIC FUNCTIONS

cute, studiate in base alia teoria delle "funzione fuchsiane" del Poincare,
Giornale di Batt., 35 (1897), 150 ; 36 (1898), 55.
FIIICKE, R., and Vorlesungen iiber die Theorie der automorphen
KLKIN, F.,

Functioned, Leipzig, Teubner, 1897-1912.

FRICKE, R., Ueber den arithmetischen Character gewisser Nctze von unendlich
cniji-ti'-nten Vierecken, Braunschw. Festschr. (1897), 85.

BAKER, H. F., Abel's theorem Cambridge (1897). and the allied theory,

FRICKE, R., Die Discontinuitatsbereiche der Gruppen reeller linearer Substitu-

tioncn rincr (otnple.i'eii Veranderliehen, Deutsche Math. Ver. 4(1897), 151. ,

KLUYVER, J. C., A special case of DirichlePs problem for two dimensions, Acta
Math., 21 (1897), 265.
I'oi SCARE,H., Sur V integration algebrique des equations differenticlles du premier
du jiremicr dcgre, Palermo Rend., 11 (1897), 193.
et

POINCARE, H., Les fonctions fuchsiennes et V equation &u = c

u Liouville's
J., ,

(5), 4 (1898), 137 ; C.R.,126 (1898), 627.

SCHLESINGER, L., Sur un probleme de Riemann,
O.K., 126 (1898), 723.
FRICKE, H., Ueber die Beziehungen zwischen der Zahlentheorie und der Theorie
der automorphen Functional, Jahresb. d. Deutsch. Math. Ver., 6 (1898), 94.
WHITTAKER, E. T., On the connexion of algebraic functions v:ith automorphic
functions, Phil. Trans., 192 (1898), 1. A
BAGNERA, G., Un teorema relativo agli invarianti delle sostituzioni di un yruppo
a
Kfeinieno, Rendic. d. Lincei (5 ), 7 (1898), 340.
DIXON, A. C., Notes on the theory of automorphic functions, Proc. L.M.S., 31
(1899), 297.
LINDEMANN, F. ,
ZUT Theorie der automorplien Functional, Miinch. Sitzungsber. ,

29(1899), 423 : 30 (1900), 493.

WIRTIXGER, W., Zur Theorie der automorphen Functionen von n Vcranderlichen,
Wien Ber., 108 (1899), 1239.
FRICKE, R. Zur Theorie der Poincare"' schcn Reihen, Jahresb. d. Deutscheu
,

M. Ver., 9 (1900), 78.

FRICKE, R., Die automorphen Elementarformen, Gott. Nach. (1900), 303.
FRICKE, R., Die llitter'sche Primform auf einer beliebigen Riemann' schen Flache,
Gott. Nach. (1900), 314.
WKLLSTEIN, J., Ueber Primformen auf Riemann' schen Fldchen, Gutt. Nach.
(1900), 380.
DIXON, A. C., Notes on the theory of automorphic functions, Proc. L.M.S., 32
(1900), 353.
Dixos, A. C., Prime functions on a Riemann surface, Proc. L.M.S., 33
(1900), 10.
JIn.r.EKT, D., Mathematische Probleme, Gott. Nach. (1900), 253; Arch,
d. M. u. P. (3), 1 (1901), 44, 213.
FKICKE, R., Ueber die Poincare schen Reihen der (-l)-ten Dimension, Dedekind
chrift(1901).
llrniiiNMiN, .1.
I., On a class of automorphic functions, Trans. Amer. M.S., 3
(1902), 1.

AI,K/AIS, R. ,
,S'/'/- une classe de fonctions hyperfuchsiennes et sur certaines sub-
s'it"' res tful s'y rapport eat, Ann. de 1'Ec. Norm., (3), 19 (1902), 261.
WHUTAKKK, Iv T., Note on a function analogous to Weierstrass" sigm a- function,
1

.
M.-ith., 31 (1902), 14f>.
l'' |; i' Kl ,
R., Ueber die /// '//:/ T/i>'</rt>' d>-.r autouiorplien Funktionen auftretenden
<nun<, (Jilt i. N.-ioli. (1903), 331.
 BIBLIOGRAPHY 93

POINCAKE, H., St'.r V i nt t

:
[i
ration ahjcbriquc des tquatiuns lintaires et Ics periodes
a'.s, Liouville's J. (5), 9 (1903), 139.
BLU.MKNTHAL, 0., Ueber ModuJfunktionen von mchrcrcn Veriinder lichen, M. Ann.,
56 (1903), 509 ; 58 (1904), 497.
FUBINI, 0.. Xulla ti'oria dcllc forme quadrat ichc Hermitianc e dci sistemi di tali
Attid. Gioenia(4), 17 (1903).
.

BLUMEXTHAL, 0., Ueber Thetafunktionen und Modulfuriktionen mehrerer

Deutsch. Math. Vc%, 13 (1904), 120.
FUBINI. G., Sulla fun:.ioni auf<nn'>rie cl i^erfuchsianc (H piit variabili in-

enti, Annali di Mat. (3), 10 (1904), 1.

YOUNG, J. W., On the group or' tin- sign (0, 3 ; 2, 4, o ) and the functions belong-

ing to Trans. Anier. Math. Soc.,

it, 5 (1904), 81.
Fi'uiNi, G., Sulla t- Sinn, di Mat. (3), 11
(1904),

FKICKE, H., Beitrdge zum Contiiini' '>near-polymorpher

Fanldioncii auf Miemannschen F/d'hfii, Matli. Ann., 59 (1904), 449.
FinsiM, G., Una questione fondo :rtij>pi e delle fun:.i<:,ii
Rend. Lincei, 13 (1904), 590.
.

FUICKK, 11.. Neue Eiitn-i^klnng-'ii liber den Existenz-bfweis der polyninrpli>'n,

Funktioncn, Heidelberg dritt Math. Kon^n-ss (1904), 246.
HnvHiNsoN, J. I., On the au.t<imorj>/iii' of fh>' <jr<mp (0, 3; 2, 6, 6),
Trans. Anu-r. Math. Soc., 5 (1904), 447.
Bi.r.MENTHAL, 0., JJenwrkuiKjeti :nr T/teoric dcr (tuf<>inr/'/,t FunLtu.n, n, Gott.
Xacli. (1904), 92.
POIXCAKE, H., Sur lex iiirar'mnts Crelle's Journ., 129
ft/-if///>t<'ti<j//cs, (1905), 89.
HURWITZ, A., Zu.r Th>" -onwrpheii Funktioncn von belicbi<j vicleu
Vrmb,-ln> Math. Ann., 61 (1!"
MORIUS, R. <>n Hi' automorpfvic functions
, <>fthc group (0, 3 ;
l
} ,
?. 1
3 ), Tr. Am.
M.S., 7(1906), 425.
nuiiUKN, T., Zur Thcorle der iiichriletitiycn uutomorp/icn Funkt.ionm, Lunds
Univ :

-krift, 40(1904), 54.

JOHANSSON, S. .
Abbilhing cinf,-li ZH*H ,,,,,,, nJu'ing-
'/e k<infonn>'
) Flt'ich'Math. Ann., 62 (1906), 177. tokreis,

JOHANSSON, S., J!e"-,-is <!>T r-jH,lymni-i ,h-'r Funktionen rum

ui' si-hen Fti'ic/tcit, Math. Ann., 62
(1906), 184.
HnvHiNsoN, .1. I.. groups wl 'cuts are
'

int,-,jn's in a

ticfield, Trans. Ann. M.S., 7 (1906), 530.

LEVI, E. K. Ricerche svMa ti'orlo delle fumioni automorfe, Atti R. Ace. Lincei
.

(5), 15(1906),
Frr.iNi, (',.. Sull'i t, ni an.tnmvrfc, Atti R. Inst. Yen., 65 (1905),
T297.
HEIK. i.
(ny., L., C'-b.-rdie Gcstalt der auf algebraischen A"///-v//
nii'ijendxHi'mjuiiirm
Math. Ann., 62 (1906), 329.
tiii/>i/cic/"-' 'tiling,

Frr.iNi, <;., Sulla eostruzione dei compi fondamentali di un grujijm di*c<>ntinuo,

Ann. d. Mat. (3), 12 (1906), 347.
HUTCHIN>ON, J. I., A method for constructing the fundamental region of a dis-
continuous group of linear transformations, Tr. Am. M.S., 8 (1907), 261.
KOEBE, P., Ueber die Uniforniisivrung reeller algebraischcr Kurven, Gott. Nach.
(1907), 177.
KOKBK. P., Uti>r die Unifonni*i>'riing bdicbiger analytischer Kurven, Gott.
Nach. (1907), 191, 633.

DEPARTMENT OF MATHEMATICS
 AITn.MoRI'HIC FUNCTIONS

Fri;is.', <-.. tfuove riart a-i. alcune rl^si di gruppi discontinui,

Palermo liend., 21 (1906), 177.
les t>/n(iiiH i' n (forme* de certi>in>:s equations fonctionnelles, C. R. ,

143 1906), 546.

Sr.\in.. II., />' r>-'-/i
figi>ng ciiier Arl>:V von // rr/i. E. T. ll'liittaker, Arch. d. M. u. P.

(3), 10(1906), :

KLFJN, F., Hem-i-rknugeu. zur Theori*; der linea/en Differentialgleichungen

Inung, Math. Ann., 64(1907), 175.
'

1'iriv, G., einer-ebenen algebraischen Kurve gehorigen transzendenten

tialgleiehungen, Monutslieftt f. M. u. P., 18 (1907), 219.
1

K<>KI;K, P., Zcr Uniformisierung der a/ yeb raise/ten Kurven, Gott. Nach. (1907).
410.
FUBINI. G., Siitln ti'oria del/c funzioni automorfe, e dell? loro trans for mazioni,

Ann. di Math. 14(1907), 33. (3),

PoiN'CAKfe, H., Stir Vuniformisation des fauctions nnalytiques, Act. Mat. , 31

(1907), 1.

SCHOTTKY. F. Ueber Bezi<'hmi<ji'ii -:i<:ischen veraiider lichen Grossen, die auf

.

gegcbene Gebiete beschrankt sind, Berlin Sitzungsb. (1907), 919 and (1908), 119. ;

KLKIN, F., Ueber die Zusammenhang zivischen dem sogenannten Oszillations-

thevrem der linear Differcntialgleichungen und dem Fiindamenialtheorein der auto-
morphcn Funktionen, Deutsch. Math. Ver., 16 (1907), 537.
YOUNG, J. W. On a class of discontinuous ^-groups defined by normal curves of
,

fourth order, Palermo Rend., 23 (1907), 97.

th'.

YouNft, J. W., A fundamental invariant of the discontinuous ^-groups defined by

the normal curves of order n in a space of n dimensions Bull. Am. M.S. (2), 14 t

(1908), 363.
ROTHE, H., Ueber das Grundtheorem und die Obertheoreme der au'omorphen
Funklionen im Falle der Hermite- Lamtechen Gleichung mit vier singuldren Punkten t

Monat. Math. u. Phys. 19 (1908), 258. ,

KOEBE, P., Die Uniformisierung der algebraischen Kurven. Mitteilung eines

Grenziib'-r'.fiiiitf* durch itericrcndes Verfahren, Gott. Nach. (1908), 112.

IIu.is, E., Ueber Kleinsche Theoreme in der Theorie der linearen Differential-

gleichungen, Math. Ann., 66 (1908),

215 68 (1909), 24. ;

GARNIEK, R. Sur equations differentielles du troisieme ordre dont V integrate

,
les

rale est uniforme, C.R., 147 (1908), 915.

HILB, E., Neue Entwicklungai ilber lineare
Differentialgleichungen, Gott. Nach.
(1908), 231 (1909), 230.
;

Poi NCAKK, H. , Sur la reduction des integrates abeliennes et

lesfonctionsfuchsiennes,
Rend. d. Circolo d. Palermo, 27 (1908), 281.
\u, M., Ueber die Darstellbarkeit der homomorphen Formenscharen dvrch
r,,in'-iirt's<-ke Zc.tareihen, Diss. Tubingen (1908).
Fur.iNi, ('.. Su//if, d/Minfiittiitd propria dei gruppi discontinue, Atti iv. con-
.,. int.-ni. matcm. (Rome), vol. ii.(1908), 169.
FLMSIM. G., Introdimone alia tcoria dei gruppi discontinui e delle funzioni
1908).
KoKiiK, P., Ui.'.hi:r die Uniformisierung der algebraischen Kurven, Math. Ann.,
67(1909), 11.'..

Rn-ii.M<iNii, H. W., On < * u tumorphic functions in relation to the

general theory of
n. -Inurn. M., 40 (1909), 258.
\V., Uit /If /i^'nntini'.uui {-groups defined
by rational normal curves
ions, liull. Am. M.S. (2), 16 (1909), 363.
V
BIBLIOGRAPHY 95

HILB, E., Ueber Kleinsche Theoreme in der Theorie der linearen Differential-
Math. Ann., 68 (1910), 24.
gJ,-ichun>.icn,
POINCARE, H., Seeks Vortriig> ilbcr au&jeu-ahlte Gegenstdndf aus der reinen
j

Mathematik nnd mathematischen Physik, Leipzig (1910).

BIEBERBACH, L. Zar Theorie der automorphen Funktionen,
,
Diss. Gbttingeii

(1910).
(/RAIG, C. F., On a class of hyperfuchsian functions, Trans. Amor. Math. Soc.,
11 (1910), 37.
KOEBE, P., Ueber <?<e Unifnruiisi>-ruiuj der algebraischen Kurven durch
automorphe Funktionen mit i,, Substitution sijruppe, Gbtt. Xach. (1909),
68 ; (1910), 180.
KOEBE, P., Ueber die Uniformisierung der algebraischen Kurven, Math. Ann.,
69(1910), l.
KOEBK, P., Uf>'.- iforminerung beliebigw analytischer Kurven, Gbtt.
Nach. (1908), 337; (1909), 324.
HILBERT, D., Grn/vhii'je ein 'nen Theorie der linenren Intcgral-

gleichungen, Sechste Jfitiei/t'ng, Gbtt. Nach. (1910), 355.

KOEBE, P., Ueber die Httbertsche Uniformi8ieriingsmethode) Gott. Nach. (1910), 59.
HECKE, E., Ueber die Konstruktion der KJa 'iseher

Korpi r ni it Hilfe ron aiitonwrphm Funktionen, Gbtt. Xach. (1910), 619.

PRYM, F., and Ros i /
Pr>jmschen Funktionen crster Ordnung,
Leipzig (1911).
COURANT, R., Ueber a lung </* l>iricliletschcn I'ri. lie l'r<M^me
der konformen Abb iJ dung, Math. Ann., 71 (1911), 145.
CHAZY, Xm- J., ions uniformcs
f
au voisina<je <!

coupures, C.R.,152 (1911), 499.

.
LiLJESTRuM, A., Xur !>-s f- '

multiplicateurs runstunts, Arkiv

Matem. Astr. Fys., 6 (1910), 18.
KOEBE, P., Ueber die Uniformisierung beliebiger analytiseher l\ lie's J.,

138 (1910), 192 139 (1911), 251. ;

HECHE, E., //V;/< Modidfunktionen >'/ uivl ihre Am'-cndunt/ anf <!ie Z<ihb'i\:

Math. Ann. 71 (1911), 1.

FIIICKE, R., Zur Transformations!!* automorphen Funktiuncn, Gbtt.

Nach. (1911), 518.
BIANCIII, L., SuJ (iruppo uuti.nitorfo d>'ll<: fnrme tcrnarie qi>"drtir//,< suwftibil!
di rapprcscntare Atti Accad. Lincei (5) 21 (1912), 305.
lo zero,

JOHANSSON, S., Zur theorie d?r A' 'ncare'scken Reihen bei den

Hauptkrcix<jruppi:n, Ofversigt af Finska Vetenskaps-Soc. Fbrhand., 53 (1910),

15, p. 25.
JOHANSSON, S., <)m Jsonvergensen af de Poincar^'ska d-serirrnn i
hufvudcirkel-
fallet, -2d Congr. math. Scand. (1911), 181.
FRICKK, R. Zur Transformations'
, 'mtomorphen Funktionen, Gbtt.
Nach. (1912), 240.
BIEBERBACH, L. ,
Au = e u und die automorphen Funktionen, Gbtt. Nach. (1912),
599.
L. E. J., Ueber die topoloyischen Schwierigkeiten des Koniinuitdts
BROUWER,
beweises der E.cisten~theoreme eiiulcutig umkchrbarer polymorpher Funktionen
auj
lliemannschen Fldchen, Gott. Nach. (1912), 603.
OSGOOD, W. F., Lehrbuc/t der Funktionentfuorie, Leipzig (2nd edition) (1912).
BROUWER, L. E. J., Ueber 'fie Singularitdtenfreiheit der
Mudulmannigfaltigkeit,
Gott. Nach. (1912), 803.
 AUTOMORPHIC FUNCTIONS

KLKIN, F.. BI:OU\VEK, L. E. J., KOEBE, P., BIEBERBACH, L., und HILB, E.,
dliingeu letrcffend automorphen Funktionen, Dent. Math. Yer., 21
(1912),
KOKKI:. P., Ueber die Unifortnisicruny dcr algebr^i Kurven, Math. Ann.,
72 (1912), 437.
KMKBE, P., Zur Begrundung der Kontinuitdismethode, Leip. Ber., 64(1912), 59.

<;,VKNIKK, II.Sur des equations differentielles du troisieme

,
ordre dont Vinti'^ralc
ne, et sur une classe ^equations nourellcs d'ordre supiriewr
rule a S6S points critiques fixe*, Ann. de Pecole super. (3), 29

(1912), 1.

LILJESTKOM, A., fitudes sur la theorie du potentiel logarithmique, Arkiv Matem.

Astro. Fys., 7 (1912), 39, p. 56.
PLEMELJ, J., Die Grenzkreisuniformisicrung analytisclier Gebildc, Monats.
f. M. u. P., 23(1912), 297.
KOEBE, P., Jl^ijr'dndung der Kontinuitdtsmethode im Gebiete der konformen
.

und Uniformisierung, Gott. Nach. (1912), 879.

!

u,i(f

OSGOOD, W. F. Existenzbeiveis betre/end Funktionen, welche zu einer eigentlich

,

'nuicrUchen automorphen Gruppe gehoren, Palermo Rend., 35 (1913), 103.

OSGOOD, W. F., On the uniformisation of algebraic functions, Annal. Math.
Princeton (2) 14 (1913), 143.
FKICKE, R. , Automorphe Funktioncn mit Einschluss der elliptischen ModulfanJ:-
tionen, EucyclopJidie der math. Wiss., Band ii. 2, Leipzig (1913).

AND CO., LTD.

Edinburgh mathematical tracts
General Editor: Prof. E. T. WHITTAKBB, F.E.8.

I A COURSE IN DESCRIPTIVE GEOMETRY

AND PHOTOGRAMMETRY FOR THE
MATHEMATICAL LABORATORY.
By E. LINDSAY INCE, M.A., B.Sc.
Price 2s. 6d. net.

2. A COURSE IN INTERPOLATION AND

NUMERICAL INTEGRATION FOR THE
MATHEMATICAL LABORATORY.
By DAVID GIBB, M.A., B.Sc.
Price 3s. 6d. net.

3. RELATIVITY.
By Professor A. W. CONWAY, D.Sc., F.R.S.

Price 2s. net.

4. A COURSE IN FOURIER'S ANALYSIS AND

PERIODOGRAM ANALYSIS FOR THE
MATHEMATICAL LABORATORY.
By G. A. CARSE, M.A., D.Sc., and
G. SHEARER, M.A., B.Sc.
Price 3s. 6d. net.

5. A COURSE IN THE SOLUTION OF

SPHERICAL TRIANGLES FOR THE
MATHEMATICAL LABORATORY.
By HERBERT BELL, M.A., B.Sc.
Price 2s. 6d. net.

6. AN INTRODUCTION TO THE THEORY

OF AUTOMORPHIC FUNCTIONS.
By LESTER R. FORD, M.A.
Price 3s. 6d. net.
Other Tracts are in preparation.

L & SONS, Ltd,, York House, Portugal St., LONDON, W.C.

 14

QA Ford, Lester R.
351 An introduction to the theory
F68

Sd,

PLEASE DO NOT REMOVE

CARDS OR SLIPS FROM THIS POCKET

UNIVERSITY OF TORONTO LIBRARY