FOUNDATIONS OF A GENERAL THEORY

OF FUNCTIONS OF A VARIABLE COMPLEX

MAGNITUDE
Bernhard Riemann
1851, Göttingen

Typesetter’s preface – Caveat lector

This is the LATEX-ed version of an English translation of Bernhard Riemann’s
1851 thesis, which marked the beginning of the geometrical theory of complex
analysis, titled Grundlagen für eine allgemeine Theorie der Functionen einer
veränderlichen complexen Grösse. The translation is obtained from
http://science.larouchepac.com/riemann/page/31,
and the original German version, available on
http://www.emis.de/classics/Riemann/,
served as a handy and authentic reference. Full collection of English translations
of Riemann’s papers is also published by Kendrick Press. It should be noted
that this transcription is almost word-by-word copying the translation, and the
typesetter DOES NOT understand what all the sentences mean. (Only very
few minor changes are made and most of them are given explicit notes.) It
also should be noted that one page is missing in the translation and the original
German words are copied there for completeness. What’s more, I did not include
the notes appearing and the end of the translation because presumably they are
not Riemann’s original notes since they do not appear in the German version.
For any correction and suggestion please email the typesetter, whose personal
email address is [email protected].

Overview1
1
A variable complex magnitude w = u + vi is called a function of another variable
dw
magnitude z = x + yi, if the function so changes that is independent of dz. This
dz
definition is based on the definition that this always occurs when the slope (depen-
dency) of magnitude w(z) is given by an analytic expression.
1 This overview is almost completely based on Riemann.

1
2
Points O and Q on planes A and B represent the values of variable complex magni-
tudes z and w, and an image of one plane projected on to the other represents their
dependency (slope) on each other.

3
dw
If the dependency is of such a kind (chapter 1) so that is independent of dz, then
dz
the original and its image are similar down to their smallest segments.

4
dw
The condition that is independent of dz is identical with the following:
dz
∂u ∂v ∂u ∂v
= , =−
∂x ∂y ∂y ∂x
from it we get
∂2u ∂2u ∂2v ∂2v
+ =0 . + =0
∂x2 ∂y 2 ∂x2 ∂y 2

5
For the location of point O, we will replace plane A with a bounded surface T extended
through plane extended over plane A. Branch points of this surface.

6
On the cohesion (connectedness) of a plane.

7
∫ ( )
∂X ∂Y
The integral + dT which extends across all of surface T , is equal to
∫ ∂x ∂y
− (X cos ξ + Y cos η) ds throughout its entire boundary when X and Y are arbitrary,
continuous functions in all points on T for x and y.

8
The introduction of coordinates s and p of point O in regard to an arbitrary line. We
will established the mutual dependency (slope) of the signs ds and dp in such a way
∂x ∂y
that = .
∂s ∂p

9
∂X ∂Y
Application of the theorem in chapter 7, if + = 0 in all surface segments.
∂x ∂y

2
10
Those are the conditions, under which a function u, that is inside of a surface T
∂2u ∂2u
which simply covers A, and that is generally satisfying the equation 2
+ =0
∂x ∂y 2
is universally finite and continuous along with all of its differential quotients.

11
The characteristics of such a function.

12
The conditions under which a function w(z), that is inside of a surface T is simply
connected and that simply covers A, is universally finite and continuous together with
its differential equations.

13
The discontinuities of such a function in an interior point.

14
The extension of the theorems in chapter 12 and 13 to the points in the interior of an
arbitrary, level surface.

15
The general characteristics of the image of a surface T , which extends in plane A, onto
a surface S which extends in plane B. We can geometrically represent the value of a
function w(z) through this.

16
∫ [( )2 ( )2 ]
∂α ∂β ∂α ∂β
The integral − + + dT which extends throughout all of
∂x ∂y ∂y ∂x
plane T always has a minimum value for one function. This is caused by the changes
in α around continuous function, or around functions which are only discontinuous in
a couple of points, with these functions being equal to 0 at the margin. If we exclude
the discontinuities in isolated points through modification, then we will get a minimum
value for only one function.

17
This is the foundation, using the boundary method, of a theorem that was presupposed
in the previous chapter.

3
18
Assume that a function α + βi for x, y, is given in a level surface T , T being arbitrarily
connected and broken down into a simply connected T ∗ through cuts. Then the func-
∫ [( )2 ( )2 ]
∂α ∂β ∂α ∂β
tion is finite, and has − + + dT extending throughout
∂x ∂y ∂y ∂x
the whole plane. Then we can turn this function into a function of z only and always
through one method, through adding a function of µ + νi of x, y, which is conditioned
in the following ways
1. µ equals 0 at the margin, and ν is given for one point.
2. The changes µ undergoes are in T , and the changes ν undergoes are in T ∗ , are
∫ [( )2 ( )2 ]
∂µ ∂µ
only in isolated points, and are only so discontinuous that + dT
∂x ∂y
[
∫ ( )2 ( )2 ]
∂ν ∂ν
and + dT remain finite throughout the whole surface, and
∂x ∂y
the latter expression remains equal on both sides of the cut.

19
A rough calculation about the conditions that are necessary and sufficient to define a
function of a complex argument inside a given numerical domain.

20
The previous method of defining a function by numerical operations contains super-
fluous elements. As a result of the observations that we have carried out here, we can
trace the range of the parts that define a function back to the necessary standard.

21
Two simple connected surfaces can relate to each other so continuously that every
point in one surface corresponds to the point that is continuously progressing with it
in the other surface, in addition to their being similar down to their smallest parts.
Naturally, any one inteior point and any one boundary point can be arbitrarily given
a corresponding point. This is what defines the relation for all points.

22
Final remarks.

4
1
If we consider z to be a variable magnitude which can gradually assume all
possible real values, then we call w a function of z, when each of its real values
corresponds to a single value of undetermined magnitude such as w. If w also
constantly changes while z continuously goes through all of the values lying
between two fixed values, then we call this function within these intervals a
continuous or a continuirlich function.2
Obviously, this definition does not set up any absolute law between the
individual values of the function, because when we assign a determinate value
to this function, the way in which it continues outside of this interval remains
totally arbitrary.
We can express the slope function (dependence) of magnitude w(z) by a
mathematical law so that we can find the corresponding value of w for ev-
ery value of z through determinate numerical operations (Grössenoperationen).
Previously, people have only considered a certain kind of function (functiones
continuae according to Euler’s usage) as having the ability of being able to deter-
mine all the values of z lying between a given interval by using that same slope
function law; however. in the meantime, new research has shown that there are
analytic expressions that can represent each and every continuous function for a
given interval. This holds, regardless of whether the slope function of magnitude
w (magnitude z) is conditionally defined as an arbitrary given numerical oper-
ation, or as an determinate numerical operation. As a result of the theorems
mentioned above, both concepts are congruent.
But the situation is different when we do not limit the variability of magni-
tude√z to real values, but instead allow complex values of the form x + yi (where
i = −1).
Assume that x + yi and x + yi + dx + dy i are two infinitesimally slightly
different values for magnitude z, which correspond to the values u + vi and
u + vi + du + dv i for magnitude w. So then, if the slope function of magnitude
du + dv i
w(z) is an arbitrary given one, then generally speaking, the ratio
dx + dy i
changes for the values for dx and dy, because when we have dx + dy i = εeφi ,
2 Riemann’s original German for continuous is stetig or its variations, and the translation on

Internet reads constant here and occasionally after. The typesetter considers this translation
to be inappropriate, and changed the word here and after. Yet sometimes this change can
not be adequately made, as the typesetter does not have the time to proofread every single
sentence with understanding, and I’m sorry for that. – Typesetter

5
then
( ) ( )
du + dv i 1∂u ∂v 1 ∂v ∂u
= + + − i
dx + dy i 2∂x ∂y 2 ∂x ∂y
[ ( ) ]
1 ∂u ∂v ∂v ∂u dx − dy i
+ − + + i
2 ∂x ∂y ∂x ∂y dx + dy i
( ) ( )
1 ∂u ∂v 1 ∂v ∂u
= + + − i
2 ∂x ∂y 2 ∂x ∂y
[ ( ) ]
1 ∂u ∂v ∂v ∂u
+ − + + i e−2φi
2 ∂x ∂y ∂x ∂y

However, regardless of the manner in which we define w as a function of z

through compounding these simple numerical operation, the value of the dif-
dw
ferential quotient is always independent of the special values of differen-
dz
tial dz.3 Obviously, not every arbitrary slope function of complex magnitude
w(complex magnitude z) can be expressed in this manner.
We will base our following investigations on the characteristic that we just
emphasized and which belongs to all functions that are in a any way definable
by numerical operations. We will consider such functions independently of their
expressions, and will proceed from the following definition without proving for
now its universal validity and its adequacy for the concept of a slope function
that can be expressed by numerical operations.
We will call a variable complex magnitude w a function of another variable
complex magnitude z, if the first function changes in such a way in connec-
dw
tion with the second function, that the value of the differential quotient is
dz
independent of the value of the differential dz.

2
We can consider magnitude w, as well as magnitude z, to be variable mag-
nitudes, which can assume any complex value. Our comprehension of such
variability, which extends itself into a connected field of two dimensions, can be
substantially facilitated by acquaintance with spatial perception.
Assume that every value x + yi of magnitude z is represented by point O
on plane(Ebene) A, whose rectangular coordinates are x, y and that every value
u + vi of magnitude w is represented by point Q on plane B, whose rectangular
coordinates are u and v. Each slope function of magnitude w(z) will then show
up as a slope function of point Q’s position according to point O’s position.
Assuming that every value for z corresponds to a determinate value for w which
in turn is continuously changing itself in conjunction with z, or in other words,
3 Thisproposition is obviously justified in all cases where, by means of the rules of differ-
dw
entiation, expressing by z is permitted by expressing w by z. This proposition’s rigorous
dz
universal validity is valid from now on in. – Riemann’s original footnote.

6
that u and v are continuous functions of x and y, then every point on plane A
becomes a point on plane B, and generally speaking, each line corresponds to
one line, and each connected surface segment of the plane corresponds to one
other connected surface segment. We can thus represent this slope function of
magnitude w(z) as an image of plane A projected on plane B.

3
We will now investigate what properties this image has when w is a function of
dw
complex magnitude z, i.e., when is independent from dz.
dz
We will designate an indeterminate point on plane A in the vicinity of O
by o, and its image on plane B by q, in addition to designating the values of
magnitudes z and w at these points by x + yi + dx + dy i and u + vi + du + dv i.
We can then consider dx, dy and du, dv to be rectangular coordinates for points
o and q in reference to points O and Q as the points of origin. And when we
have dx + dy i = εeφi and du + dv i = ηeψi , then the magnitudes ε, φ, η, ψ
become polar coordinates for these points with the same points of origin. Now
if o′ and o′′ are any two determinate positions of point o within an infinitesimal
vicinity of O, and if we express the meaning of the remaining symbols that are
dependent on o′ and o′′ by corresponding indices, then the postulate

du′ + dv ′ i du′′ + dv ′′ i
′ ′
=
dx + dy i dx′′ + dy ′′ i
and consequently

du′ + dv ′ i η ′ (ψ′ −ψ′′ )i dx′ + dy ′ i ε′ (φ′ −φ′′ )i

= e = = e ,
du′′ + dv ′′ i η ′′ dx′′ + dy ′′ i ε′′

η′ ε′
from which ′′
= ′′ and ψ ′ − ψ ′′ = φ′ − φ′′ , i.e. the angles o′ Oo′′ and q ′ Qq ′′ are
η ε
equal in the triangles o′ Oo′′ and qQ′ q ′′ and the sides that include them are pro-
portional to each other. Thus we find that two infinitely small and corresponding
triangles are similar, and that this also universally holds for the smallest seg-
ments of surface A and their images on surface B. The only exception to this
theorem occurs in those special cases when the mutually corresponding changes
in magnitudes z and w do not occur in a finite relationship. This was tacitly
assumed in the derivation.4
4 One should see the following on this subject:

“Universal Solution to the Problem: Describing the segments of a given surface so that the
images of what are described are similar down to their smallest parts” by C.F.Gauss. (This
was published in Astronomische Abhandlungen, herausgegaben von Schumacher. Drittes Heft.
Altona. 1825, as the answer to the question in the contest set uo by the Royal Society of the
Sciences in Copenhagen of the year 1822.)(Gauss Werke Bd. IV, p. 189.) – Riemann’s original
footnote.

7
4
du + dy i
When we transform the differential quotients into the form
dx + dy i
( ) ( )
∂u ∂v ∂v ∂u
+ i dx + − i dy i
∂x ∂x ∂y ∂y
,
dx + dy i
then it is evident that we will get the same values for any two values of dx and
dy if
∂u ∂v ∂v ∂u
= and =−
∂x ∂y ∂x ∂y
These conditions are also necessary and sufficient in order to have w = u + vi
become a function of z = x + yi, and the following individual terms of this
function also comes from the conditions :
∂2u ∂2u ∂2v ∂2v
2
+ 2 = 0, 2 + 2 = 0
∂x ∂y ∂x ∂y
These two equations form the foundation for the investigation of the character-
istics that an individual term in such a function is considered to have. We will
allow the proof for the most important of these characteristics to be proceeded
by a thorough consideration of the entire function. But first, however, we will
discuss and define some points which belong to the universal domain in order
to smooth out the ground for this investigation.

5
In the following observations, we will limit the variability of magnitudes x and
y to a finite domain in which we consider the location of point O as no longer
on plane A itself, but a surface T , which extends over plane A. We have chosen
this wording, which makes it inoffensive to speak of planes lying on top of
one another in order to leave open the possibility that the location of point O
repeatedly occurs over the same segment of the plane. Nevertheless, we assume
in such a case that surface segments which are lying on top of each other are
not connected along a line, for this would cause convolutions in the planes, or
a fissure in the segments lying on top of each other.
We can then fully determine the number of plane segments that are lying on
top of each other in every surface segment when the boundary is given according
to location and direction (i.e., the boundary’s inner and outer side); however,
the actual course of the boundary can still develop differently.
In reality, if we draw an arbitrary line through the segment of the plane
covered by the surface, then the number of surface segments lying above one
another will only change when we cross the boundary. Naturally, when we cross
the boundary going from outside to the inside, the change is +1, while going

8
in the opposite direction the change is −1, and this holds everywhere. Every
bordering surface segment along the edge of this line continues to carry on in
a totally determined manner so long as the line does not touch the boundary,
because the only place any undeterminateness can occur at all is in an isolated
point, either in an isolated point on the line itself, or in an isolated point a
finite distance from the line. Therefore, when we limit our observations to a
passing section of line L that is inside the surface, and limit it on both sides to a
sufficiently small surface surface strip, we can speak of determinately contiguous
surface segments, the amount of which is equal on every side, and which we can
describe on the left as a1 , a2 , . . . an , and on the right as a′1 , a′2 , . . . a′n , if we give
the line a definite direction. Every surface segment a will then be continued in
a surface segment a′ , and of course, this will be universally the same for the
entire course of line L, even though it can change in a couple of its points for
special positions of L. Let us assume that above a certain point σ, (i.e., along
the anterior segment of L) the surface segments a1 , a2 , . . . an are connected in
succession to surface segments a′1 , a′2 , . . . a′n , but that below the same point
there are the surface segments aα1 , aα2 , . . . aαn , where α1 , α2 , . . . αn are only
differentiated through the series 1, 2, . . . n. Then a point above σ that stands for
a1 in a′1 will end up in surface segment aα1 if it crosses over to the left side under
σ. And when this same point encircles point σ going from left to right, then
the index of the surface segment in which this point finds itself will run through
the numbers 1, α1 , α2 , . . . , µ, αµ , . . . in series. In this series, as long as the term
1 does not repeat itself, it is inevitable that all the terms will differ from each
other because with any given arbitrary average term αµ it is imperative that µ
and all other previous terms back to 1 precede it in direct succession. However,
when the term 1 repeats itself in a series of terms which are evidently smaller
than n and equal to m, then the remaining terms have to follow in that very
same order. The point which was circling around σ will then, in conformity with
m, revert back to circulating in the same surface segment, and is limited by m to
the surface segments lying on top of each other, which are united to each other
at a single point above σ. We will call this point a branch point of the (m − 1)
order of surface T . By applying this same operation to the remaining (n − m)
surface segments, these surface segments, if they do not develop otherwise, will
break down into a system of m1 , m2 , . . . surface segments. In this case, there
would also be branch points of the (m1 − 1)th, (m2 − 1)th order in point σ.
[Wenn die Lage und der Sinn der Begrenzung von T und die Lage ihrer
Windungspunkte gegeben ist, so ist T entweder vollkommen bestimmt oder doch
auf eine endliche Anzahl verschiedener Gestalten beschränkt; Letzteres, in so
fern sich diese Bestimmungsstücke auf verschiedene der auf einander liegenden
Flächentheile beziehen können.
Eine veränderliche Grösse, die für jeden Punkt O der Fläche T , allgemein zu
reden, d. h. ohne eine Ausnahme in einzelnen Linien und Punkten5 auszuschliessen,
5 Diese Beschränkung ist zwar nicht durch den Begriff einer Function an sich geboten, aber

um Infinitesimalrechnung auf sie anwenden zu können erforderlich: eine Function, die in allen
Punkten einer Fläche unstetig ist, wie z. B. eine Function, die für ein commensurables x und
ein commensurables y den Werth 1, sonst aber den Werth 2 hat, kann weder einer Differen-

9
Einen bestimmten mit der Lage desselben stetig sich ändernden Werth annimmt,
kann offenbar als eine Function von x, y, angesehen werden, und überall, wo in
der Folge von Functionen von x, y die Rede sein wird, werden wir den Begriff
derselben auf diese Art festlegen.
Ehe wir uns jedoch zur Betrachtung solcher Functionen wenden, schalten
wir noch einige Erörterungen über den Zusammenhang einer Fläche ein. Wir
beschränken uns dabei auf solche Flächen, die sich nicht längs einer Linie spal-
ten.]6

6
When this possibility does not exist, we will consider these surface segments to
be separate.
Our investigation of the continuity (connectedness) of a surface is based on
cutting up the plane into transverse segments, i.e., through lines which simply
cut across the interior – not cutting one point more than once – going from
boundary point to boundary point. The latter boundary point can also lie in
a segment that is added to the boundary, or thus, in an earlier point on the
transverse cut.
We say a surface is connected, when every transverse breaks it down into
pieces so that they are either simply connected ot multiply connected.

Pedagogical Theorem 1 A simply connected surface A is broken down by any

cut ab into two simple connected pieces.

Assume that one of these pieces is not partitioned by cut cd. We can then
obviously see that although none of this piece’s endpoints, nor endpoint c, nor
both endpoints fall on line ab, we can get a connected surface by cutting A that
is contrary to our postulate by establishing contact along all of line ab, or along
part of cb, or along part of cd.

Pedagogical Theorem 2 When we break surface T down into a system of

T1 of m1 simply connected surface segments by using an amount n1 of cuts 7 q1
and when we break it down into a system T2 of m2 surface segments by using
an amount n2 of cuts q2 , then n2 − m2 cannot be larger than n1 − m1 .

If any line q2 does not entirely fall into the q1 cut system, then at the same
time it also becomes one or more of the cuts q2′ across surface T1 . We can
consider the endpoints of cut q2′ to be:
tiation, noch einer Integration, also (unmittelbar) der Infinitesimalrechnung überhaupt nicht
unterworfen werden. Die für die Fläche T hier willkürlich gemachte Beschränking wird sich
später (Art. 15) rechtfertigen.
6 These paragraphs are missing in the translation, and the typesetter therefore has to put

the original German text here for completeness. – Typesetter

7 Dividing up a surface through various cuts always means a successive division, i.e., that

kind of division, where the planes that result from a cut get partitioned again by a new cut.
- Riemann’s original footnote

10
 1. the 2n2 endpoints of cut q2 , except when their ends coincide with a seg-
ment of the line system q1 ,
2. any average point on cut q2 , where it joins up with any average point on
line q1 , except when the former point is already on another line q1 , i.e.,
when it coincides with one end of cut q1 .
We shall now: define µ as how frequently lines from both systems meet or
cross in their course (we will count a single common point twice), define ν1
as how frequently an end section q1 coincides with a middle section q2 , and
define ν2 as how frequently an end section q2 coincides with a middle section
q1 . Finally, we will define ν3 as how frequently an end section q1 coincides with
an end section q2 . Given the above Nr. 1 2n2 − ν2 − ν3 , Nr.2 µ − ν1 produce the
endpoints for cut q2′ . But if we take both cases together, then they contain all
the endpoints, and each endpoint only once. Therefore, the number of cuts is:
2n2 − ν1 − ν3 + µ − ν2
= n2 + s
2
We can get the number of cuts q1′ of surface T2 by a totally similar deduction,
which is based on the lines q1 ,
2n1 − ν1 − ν3 + µ − ν2
2
thus = n1 + s. Surface T has now been obviously transformed, through the
n2 + s cuts q2′ , into that very same surface in which T2 is broken down by n1 + s
cuts q1′ . However, what we get out of T1 as a result of m1 are simply connected
pieces, which break down according to Theorem 1, and through n2 + s cuts, into
m1 + n2 + s8 surface segments. From this it would have to follow that if m2
were smaller than m1 + n2 − n1 , the number of surface segments T2 produced
by the n1 + s cuts would have to be more than n1 + s, which is absurd.
According to this theorem, if n does not define the number of cuts, then
m describes the number of pieces, n − m being constant for all partitions of
a plane into simply connected pieces. For if we observe any two determinate
partitions by n1 cuts into m1 pieces, and by n2 into m2 pieces, then if the
former pieces are simply connected, n2 − m2 ≤ n1 − m1 , while if the later pieces
are simply connected, then n1 − m1 ≤ n2 − m2 . When both conditions occur,
n2 − m 2 = n1 − m 1 .
We can appropriately call this number the “degree of connection” of a plane;
it is
– according to definition – decreased by 1 with every cut,
not changed by a line simply cutting from an interior point through the
interior to a boundary point or to an earlier point on the cut, and
increased by 1 through an interior cut that is universally simple and that
has two endpoints,
8 Here the English translation reads m + n and s, changed according to the German
1 2
version. – Typesetter

11
 because the first case can be changed by one cut, but the last case can only
be changed by having two cuts in one cut.
And last of all, we can obtain a degree of connection from a surface consisting
of several pieces if we add the degrees of connection of these various pieces
together.
In the following section, we will generally limit ourselves to a surface con-
sisting of one section (piece) and we will suit ourselves by using the artificial
description of a simple, twofold, etc. connection for its connection, so that what
we mean by an n-fold connected surface is one which is divisible by n − 1 cuts
into a simply connected surface.
When we consider the slope function of a boundary’s connectedness in rela-
tion to the connectedness of a surface it is readily apparent that:

1. The boundary of a simply connected surface necessarily consists of one

encircling line.
If the boundary consisted of fragmented pieces, then cut q, which links
a point in section (a region, piece) a with a point in another section b,
would only be separating connected surface segments from each other.
This would be so because inside the surface along a, a line would lead
from one side of cut q to the opposite side, and therefore q would not
partition the surface, which is contrary to the supposition.
2. Every cut either increases the number of sections in the boundary by 1,
or decreases it by 1.
Cut q either connects a point on a boundary section a with a point on
another boundary section b, – and in this case, all of these together form the
series a, q, b, q forming a single boundary from one encircling piece –
or cut q connects two points on one boundary piece – and in this case the
segment breaks down into two pieces through both of the end points of this cut.
Both of these pieces now form, together with the cut, a section of the boundary
that circles back into itself.
or finally, cut q ends at one of its earlier pints and we can consider it as
composed of one line o that circles back into itself, and of another line L which
connects a point on o with a point on boundary segment a, – in which case,
o forms one part of, and a, L, o, L form another part of a boundary piece that
circles back into itself.
So there are either – in the first place, only one boundary piece in place of
two, – or in both of the latter examples two boundary portions in place of one,
from which our proposition comes.
Therefore, the number of pieces comprising the boundary of an m-fold con-
nected surface segment is either = n or is smaller by a precise number.
We can even produce a corollary from this:

Corollary 3 If the number of boundary pieces that an n-fold connected surface

12
has is = n, then this surface breaks down into two separate pieces with every cut
in the surface’s interior that circles back into itself.
The degree of connection is not changed as result of this, and the numbers of
pieces in the boundary are increased by two; so if the surface were a connected
one, it would have n-fold connectedness and n + 2 boundary pieces, which is
impossible.

7
Assume that X and Y are two continuous functions of x, y which are in all
points of the surface T , which in turn is extended over A. Then the integral
that extends to all the elements dT in this surface
∫ ( ) ∫
∂X ∂Y
+ dT = − (X cos ξ + Y cos η) ds
∂x ∂y

if we describe ξ as the inclination against the x axis of a straight line drawn

from the boundary towards the interior, for every point on the boundary, and
if we describe η as the inclination against the y axis. And finally, this integral
equals the other one if this integration covers collectively all of the elements ds
that are on the boundary line. ∫
∂X
In order to transform the integral dT we will partition the segment of
∂x
plane A that is covered by surface T into primary bands(Elementarstreifen) by
means of a system of lines parallel to the x axis. And we will do this in such a
way that every one of surface T ’s branch points falls on one of these line. As a
result of this precondition, we get one or more differentiated trapezoidal shaped
pieces developing from every one of these surface T segments that falls of one
of the lines. Given then any undetermined primary band which segregates ∫ the
∂X
element dy our of the y axis, this band’s contribution to the value of dT
∫ ∂x
∂X
will obviously be = dy dx, if this integral is extended through this or
∂x
these straight line belonging to surface T , these straight lines falling on a normal
proceeding from a point dy. If we describe the lower endpoints of these lines
(i.e., which correspond to the smallest values of x) as O′ , O′′ , O′′′ , the upper end
points as O′ , O′′ , O′′′ , the x-value in these points as X′ , X′′ , . . . X ′ , X ′′ , . . . , the
matching elements which are segregated by the planar bands out of the boundary
as ds′ , ds′′ , . . . ds′ , ds′′ , . . ., and the values of ξ in these as ξ′ , ξ′′ , . . . ξ ′ , ξ ′′ , . . .
then, ∫
∂X
dx = −X′ − X′′ − X′′′ . . . + X ′ + X ′′ + X ′′′ . . .
∂x
It is evident that angle ξ becomes acute at the lower end points, and obtuse
at the higher endpoints. Therefore
dy = cos ξ′ ds′ = cos ξ′′ ds′′ . . . = − cos ξ ′ ds′ = − cos ξ ′′ ds′′ . . .

13
 ∫ ∑
∂X
Through substitution this value results in ds dx = − X cos ξds
∂x
where the summation relates to all the boundary elements which have dy as a
projection in the y axis.
We can obviously exhaust all of the elements in surface T and all of the ele-
ments in the boundary by the integration of all dy that comes into consideration.
Considering this environment, we get,
∫ ∫
∂X
dT = − X cos ξds
∂x
And we get as a result of totally similar conclusions
∫ ∫
∂Y
dT = − Y cos ηds
∂y
and consequently
∫ ( ) ∫
∂X ∂Y
+ dT = − (X cos ξ + Y cos η) ds
∂x ∂y

Q.E.D.

8
Consider a boundary line proceeding from an established starting point cut
into a direction that will be defined later. We will describe the length of this
boundary up to an undefined point O0 by s. Next, consider the distance of a
normal set up from point O0 to an undefined point O which we will call p and
which we will consider to be positive on the inside of the boundary. Then we
can consider the values that x and y have in point O to be functions of s and
p, and the partial differential quotients
∂x ∂y ∂x ∂y
= cos ξ, = cos η, = ± cos η, = ∓ cos ξ
∂p ∂p ∂s ∂s
in the points of the boundary line. Turn out so that in these differential quotients
the upper notation shows in cases the direction, in which we consider magnitude
s to be growing, includes an equal angle in with p, just as the x axis includes
the angle with the y axis, when one is counter-posed to the other, the lower.9
We will assume this direction to be such in all segments of the boundary so that
∂x ∂y ∂y ∂x
= and consequently =−
∂s ∂p ∂s ∂p
which does not at all essentially infringe upon our result’ universality.
9 The English translation here is incomprehensible, and the typesetter made some modifi-

cation based on the original German version. – Typesetter

14
 We can also expand these determinations to lines inside of T . And in or-
der to determine the signs for dp and ds, if we want to continue their mutual
dependency (slope function) as it was previously, we can add on a statement
which will determine the signs for dp or ds. In creating such an encircling line,
naturally we will indicate which of the surface segments separated by such a
line also serves as this line’s boundary. It is through this that we determine the
sign for dp, not with an encircling line, but at its beginning point, i.e., at the
endpoints where s assumes the smallest value.
When we introduce the values we got for cos ξ and cos η from the proven
equations in the previous chapter we will then get, to the same extent as we got
in the previous chapter,
∫ ( ) ∫ ( ) ∫ ( )
∂X ∂Y ∂x ∂y ∂y ∂x
+ dT = − X +Y ds = X −Y ds
∂x ∂y ∂p ∂p ∂s ∂s

9
When we apply the theorem from the conclusion of the previous chapter to the
∂X ∂Y
situation where + = 0 in all the segments of the plane, then we get the
∂x ∂y
following theorem:
1. If X and Y are finite and continuous for all the points in T , and if they
∂X ∂Y
provide satisfactory functions for the equation, + =0
∂x ∂y
then if we expand through the whole boundary for T ,
∫ ( )
∂x ∂y
X +Y ds = 0
∂p ∂p
If we can now imagine an arbitrary surface T, that is stretched our over A,
breaking down into pieces T2 and T3 in an arbitrary manner, ∫ (then in rela- )
∂x ∂y
tions to the boundaries for T2 , we can consider the integral X +Y ds
∂p ∂p
to be the difference of the integral in relation to the boundary for T1 and
T3 , while in the case where T3 runs right up to T1 ’s boundary, both inte-
grals conceal each other out. However, all the remaining elements corre-
spond to an element in the boundary of T2 .
Through this transformation, we can get the following out of Theorem 1:
∫ ( )
∂x ∂y
2. The value of the integral X +Y ds, which covers the entire
∂p ∂p
boundary of a surface that is extended over A, will remain constant during
arbitrary expansions and contractions only if it does not gain of lose any
surface segments as a result of this. If this were to happen, the precondi-
tions for Theorem 1 would not be fulfilled.
If the functions X and Y suffice for every surface segment of T in the
differential equation that we just described, but if they are afflicted with

15
 discontinuity in isolated lines or points, then we can encapsulate every one
of these lines and points in an arbitrarily small plane segment, like a seed
pod. We then get the following by applying Theorem 2:
∫ ( )
∂x ∂y
3. In reference to the entire boundary of T , the integral X +Y ds
∫ ( ) ∂p ∂p
∂x ∂y
is equal to the sum of the integrals X +Y ds in relation to the
∂p ∂p
encapsulation of all discontinuities. Naturally, this integral also has the
same value for every one of these discontinuities, no matter how compact
the boundaries are that encircle them.
This value is necessarily equal to null for a simple discontinuous point, if
the distance of point O from the discontinuous point ϱ becomes infinitely
small at the same time that ϱX and ϱY do too. We can then introduce the
polar coordinates ϱ, φ in reference to such a point as a starting point and in
reference to an arbitrary initial direction. Finally, in order to encapsulate
these polar coordinates, we can choose to draw a circle around them that
has the radius ϱ, so that the integral that relates to this is
∫ 2π ( )
∂x ∂y
X +Y ϱdφ
0 ∂p ∂p
Consequently, it cannot have a value for κ different than null, because just
as we can always assume κ to be ( small, we can also
) assume ϱ to be too,
∂x ∂y
so that irrespective of the symbol X +Y , ϱ can become smaller
∂p ∂p
κ
than for every value of φ. Consequently,
2π
∫ 2π ( )
∂x ∂y
X +Y ϱdφ < κ
0 ∂p ∂p

4. Let
∫ (us take a simple) connected ∫surface
( extended)over A. If the integrals
∂x ∂y ∂x ∂y
Y −X ds = 0, and X +Y ds they being integrals
∂s ∂s ∂p ∂p
that cover the whole boundary of every surface segment, then these inte-
grals will have the same value for any two fixed point O0 and O in relation
to all lines going from O0 to O in these integrals.
The pair of lines s1 and s2 which connect the points O0 and O form
together a line s3 that circles back into itself. This line in turn either
has the property of being unable to cut across any point more than one,
or the property of being capable of partition into several totally simple
lines that circle back into themselves. It has this second property because
when we want to go back to an earlier point, we can eliminate the seg-
ment which has become continuous in the meantime from an arbitrary
point on these very same continuous lines. We can then consider what
follows as a direct continuation of what went on before. However, every

16
 one of these lines partition the plane into a simply and into a twofold
connected segment, and therefore it necessarily follows that one of these
lines form
∫ the
( the entire ) boundary for one of these segments, while the
∂x ∂y
integral Y −X ds extends through this plane equals zero in
∂s ∂s
accordance with the proposition. This also holds true for the integral that
extends through all of line s3 , if we consider magnitude s to be increasing
everywhere in the same direction. Therefore, the integrals that extend
through lines s1 and s2 must cancel each other out, if this direction re-
mains unchanged, i.e., if it goes in one direction from O0 to O, and in the
other direction from O to O0 . So if the latter direction is changed, the
integrals become equal.
If somewhere there is now an arbitrary surface T , in which, generally
∂X ∂Y
speaking, + = 0, when we can next exclude the inconsistencies if
∂x ∂y
this is necessary, so that in the remaining surface sections
∫ ( )
∂x ∂y
Y −X ds = 0
∂s ∂s

for every surface segment. This is then partitioned by cuts into a sim-
ple connected surface T ∗ . ACCORDINGLY, our integral has the same
value for every line that goes from a point O(0 to another O)inside surface
∫ O
∂x ∂y
T ∗ . This value, for which the notation Y −X ds suffices as
O0 ∂s ∂s
shorthand, holds O0 to be fixed and O to be moving. We can also consider
it to be a determinate function for every one of O’s positions, regardless
of the course of the connecting lines. Consequently, we can consider it to
be a function of x, y.
We can express the change that occurs( in this function
) by displacing O
∂x ∂y
along an arbitrary linear element ds by Y −X ds and the change
∂s ∂s
in this function in continuous for T ∗ everywhere, as well as being equal
along both sides of a cut across T .

5. Therefore, when we consider O0 to be fixed, the integral

∫ O( )
∂x ∂y
Z= Y −X ds
O0 ∂s ∂s

forms a function of x, y, which is continuous everywhere in T ∗ . However,

when this function passes beyond the cuts in T , it changes around a func-
tion along the cut from being a branch point to being another constant
magnitude. The partial differential quotient for this is
∂Z ∂Z
=Y , = −X
∂x ∂y

17
 The changes which we brought about by passing beyond the cuts are de-
pendent on having the same number of cuts as there are magnitudes that
are independent of each other. For when we go this system of cuts back-
wards – doing the later segments first – this change is generally determined
when its value is given at the beginning of every cut. However, the later
values are independent of each other.

10
If we replace the functions that have been described by X( up to now with )
∂u′ ∂u ∂u′ ∂u ∂X ∂Y ∂ 2 u′ ∂ 2 u′
u − u′ and u − u′ for Y , then + =u + −
∂x
( 2 ∂x ) ∂y ∂y ∂x ∂y ∂x2 ∂y 2
∂ u ∂2u
u′ + 2 and if the functions u and u′ satisfy the equations
∂x2 ∂y

∂2u ∂2u ∂ 2 u′ ∂ 2 u′
2
+ 2 = 0, 2
+ =0
∂x ∂y ∂x ∂y 2
then
∂X ∂Y
+ =0
∂x ∂y
and we can∫find( the application) of the theorem in the∫ previous
( chapter) in the
∂x ∂y ∂u′ ∂u
expression X +Y ds which is equal to = u − u′ ds.
∂p ∂p ∂p ∂p
Now if in relation to function u we make the hypothesis that this function,
together with its first differential quotient, does not tolerate any possible kind
of discontinuity along a line, and if we also assume that function u becomes
infinitely small for every discontinuous point as the distance ϱ of point O from
∂u ∂u
those very same ϱ and ϱ does at the same time, then we can conclude from
∂x ∂y
the notes to section III of the previous chapter that we can keep on disregarding
the discontinuous in u.
Therefore,we can assume a value R of ϱ for every straight line that proceeds
from an discontinuous point, so that
∂u ∂u ∂x ∂u ∂y
ϱ =ϱ +ϱ
∂ϱ ∂x ∂ϱ ∂y ∂ϱ

ϱ(u − U ) and thus ϱu along with ϱ will become infinitely small at the same
time. And according to the proposition, the same goes for which always finite
at its lower end. We can also describe U as the value of u for ϱ = R, M , for
∂u
every interval, regardless of the signs of the greatest value for the function ϱ .
∂ϱ
Then, following the same interpretation, u−U will always be < M (log ϱ−log R),
∂u ∂u
and consequently, ϱ and ϱ . Consequently, if u′ is not burdened with any
∂x ∂y

18
discontinuities the same also goes for
( ) ( )
∂u′ ′ ∂u ∂u′ ′ ∂u
ϱ u −u and ϱ u −u ;
∂x ∂x ∂y ∂y

the cases discussed in the previous chapter also making their appearance in this.
We will even assume further, that surface T , which forms the site for point
O, has extended over A everywhere, and that an arbitrary fixed point O0 , where
u, x, y have the values of u0 , x0 , and y0 , is in this same extended surface. If we
1 ( )
consider the magnitude log (x − x0 )2 + (y − y0 )2 = log r to be a function of
2
x, y, then it has the characteristic that

∂ 2 log r ∂ 2 log r
+ =0
∂x2 ∂y 2
so that it is only subjected to a discontinuity when x = x0 , y = y0 . Thus in our
case, this only occurs for one point on surface T . ∫ ( )
∂ log r ∂u
According to Article 9, theorem III, when we replace u′ with log r, u − log r ds
∂p ∂p
the entire boundary around T is equal to this integral with regard to an arbi-
trary encirclement of point O0 . So when we want to select the periphery of a
circle in this case, where R has a constant value, and where by starting out
from one of the points on the periphery and proceeding in a fixed arbitrary
direction, we can describe the arc up to O in terms
∫ 2π of segments of the radius
∫ by
∂ log r ∂u
then the integral directly above is equal to − u rdφ − log r ds
∂r ∂p
∫ ∫ 2π 0
∂u
or therefore to ds = 0 , = − udφ, whose value for an infinitely small
∂p 0
r crosses over into −u0 2π when u is continuous in point O0 .
Therefore, in regard to the propositions we established for u and T , when
we have an arbitrary point O0 in which u is continuous inside the surface
∫ ( )
1 ∂u ∂ log r
u0 = log r −u ds
2π ∂p ∂p

in relation to the entire boundary itself and

∫ 2π
1
= udφ
2π 0
in relation to a circle drawn around O0 . We can draw the following conclusions
from the first expression in this paragraph:
Pedagogical Theorem 4 If a function u, which is inside of a surface T that
∂2u
itself simple covers plane A everywhere, generally satisfies the differential equation: +
∂x2
∂2u
= 0 so that,
∂y 2

19
 1. The point in which this differential equations is not fulfilled are not surface
segments,
∂u ∂u
2. The points in which u, , become discontinuous, do not continuously
∂x ∂y
satisfy any line,
∂u ∂u
3. The magnitudes ϱ ,ϱ become infinitely small for every discontinuous
∂x ∂y
point as well as for the distance ϱ of point O from that same inconsistent
point.
4. u excludes any discontinuity that can be cancelled out by changing its value
in isolated point.
Then this function u is necessarily finite and continuous along with all of
its differential quotients for all of the points inside this surface.
In reality, however,
∫ ( we will consider
) point O0 to be movable so that in
∂u ∂ log r ∂ log r ∂ log r
the expression, log r −u ds only the values log r, ,
∂p ∂p ∂x ∂p
change. However, these magnitudes are also finite and continuous functions of
x0 , y0 , for every element in the boundary, so long as O0 remains inside of T ,
and in addition to all of their differential quotients. These finite and continuous
functions can be expressed by the broken rational functions of these magni-
tudes, the functions that only have powers of r in their denominators. And this
also holds for the value of our integral, and consequently for function u0 itself,
because under our earlier propositions, function u0 could only have a value dif-
ferent from the value of our integral in those isolated points in which it would
be discontinuous. And this possibility has been eliminated by proposition 4 of
our theorems.

11
Using the same preconditions that we applied to u and T at the end of the last
chapter, we get the following theorems:
∂u
I. When u = 0 along a line, and = 0, then u = 0 everywhere.
∂p
∂u
Next we can prove that line λ, where u = 0 and = 0, cannot form the
∂p
boundary of surface segment a, where u is positive.
Given that this occurs, then we can cut a piece out of a. This piece’s bound-
ary is partially formed by λ, and partially by a circumferential line. In addition,
this piece does not contain point O0 which is centre for the circumferential line,
and this whole construction is possible. Then when we describe O’s polar coor-
dinates in relation to O0 by r, φ, we get
∫ ∫
∂u ∂ log r
log r ds − u ds = 0
∂p ∂p

20
expanding through this piece’s entire boundary. As a consequence,
∫ ∫
∂u
udφ + log r = 0,
∂p
our assumption for all of the arcs that also belong to the boundary, would be,
∫ ∫
∂u
ds = 0 or udφ = 0
∂p
which is irreconcilable with our presumption, that u is positive in a’s interior.
∂u
In a similar manner, we can also prove that equations u = 0 and =0
∂p
cannot occur in a boundary segment belonging to s surface piece b where u is
negative.
∂u
So if u = 0 and = 0 on a line in surface T , and if u were to be different
∂p
from null in any one of surface T ’s segments, then such a surface segment would
obviously have to be bounded either by this line itself, or by a surface segment
where u would be equal to 0. So in any case it would be bounded by a line
∂u
where u and would be equal to 0 and this would necessarily return us to
∂p
one of the assumptions we negated a few lines back.
∂u
II. When we are given the values for u and along a line, then this defines
∂p
u in all segments in T .
If u1 and u2 are any two determinate functions which satisfy the conditions
that we imposed on function u, then these conditions also hold for their differ-
ence, u1 − u2 , and we can show this right away by substituting this difference
into these conditions. And if u1 and u2 as well as their first differential quo-
tients, converge towards p when they are on a line, but they do not do so in
∂(u1 − u2 )
another surface segment, then u1 − u2 = 0 and = 0 along this line,
∂p
without being equal to 0 every else. This would then be contrary to Theorem I.
III. The points inside of T where u has a constant value necessarily form lines
if u is not constant everywhere. These lines then divide those surface segments
where u is larger from the surface segments where u is smaller.
This theorem is composed of the following conditions:
• u cannot have either a maximum or a minimum value in a point inside of
T.
• u cannot be constant in only one section of the plane.
• the lines in which u equals a cannot bound both sides of the surface
segment where u − a has the same symbol.
As we can easily see, theorems which always have to lead to violating the
equations we proved in the last chapter;
∫ 2π ∫ 2π
1
u0 = udφ or (u − u0 )dφ
2π 0 0

21
are therefore impossible.

12
We will now return to considering a complex variable magnitude w = u+vi, and
we will consider it generally (i.e., without excluding the exceptions in isolated
lines and points.) This magnitude has a determinate value for every point O in
surface T that continuously changes with point O’s position, and in conformity
with the equations
∂u ∂u ∂u ∂u
= , =−
∂x ∂y ∂y ∂x
we will characterise this property of w according to the way we did earlier and
so we will call w a function of z = x + yi. In order to simplify what is coming,
we will pre-establish that a discontinuity that can be eliminated by changing its
value in an isolated point cannot occur in a function z.
First of all, we will attribute surface T with a simple connectedness and with
simple expansion everywhere over plane A.

Pedagogical Theorem 5 If function w(z) does not have any break in its con-
tinuity anywhere along a line, and furthermore, if w(z − z ′ ) becomes infinitely
small as it approaches point O for any arbitrary point O′ in the surface where
z = z ′ , then this function is necessarily finite and continuous for all points
inside the surface and for all of its differential quotients.
The preconditions which we set up for the changes in magnitude w break down
when we substitute z − z ′ = ϱeφi for u and v in
∂u ∂v
1) − =0
∂x ∂y
and
∂u ∂x
2) + =0
∂y ∂x
for every segment in surface T ; and when

3) function u and v are not discontinuous along a line;

4) ϱu and ϱv become infinitely small along with the distance from point O to
O′ for any point O′ .
5) function u and v exclude any discontinuities that can be eliminated by chang-
ing their values in isolated points.
∫ ( )
∂x ∂y
As a result of preconditions 2,3 and 4, integral u −v ds which ex-
∂s ∂s
tends to the boundaries of surface T , ends up being equal to 0 for every segment
in surface T , according to chapter 9, III. According to chapter 9, IV, integral

22
∫ O ( )
∂x ∂y
u −v ds has the same value for every line going from O0 to O.
O0 ∂s ∂s
Additionally, when we consider O0 as fixed, this integral forms function U (x, y)
which is necessarily continuous up to isolated point, and for which the differ-
∂U ∂U
ential quotient = u and = −v for every point, (according to 5). But
∂u ∂y
by substituting these values for u and v, preconditions 1,2 and 3 change over
into the conditions of the pedagogical theorem at the end of chapter 10. In
this case therefore, function U , along with its differential quotients is finite and
continuous for all points in T , and this also holds for the complex function
∂U ∂U
w= − i and its differential quotients according to z.
∂x ∂y

13
We will now investigate what happens when we assume, still retaining chapter
12’s special preconditions, that (z − z ′ )w = ϱeφi w no longer becomes infinitely
small for a determinate point O′ as we infinitely converge on point O. In this
case, as point O converges infinitely close to point O′ , w becomes infinitely large.
We can assume that when magnitude w does not remain with ϱ1 in the same
series, i.e., if both of their quotients approach a finite boundary, then at least the
order of both magnitudes will be in such a finite ratio to each other, that a power
of ϱ will result whose product in w for an infinitely small ϱ will be either infinitely
small or remain finite. If µ is the exponent of such a power, and if n is the
next largest whole number, then magnitude (z − z ′ )n w = ϱn enφi w will become
infinitely small with ϱ, and therefore (z − z ′ )n−1 w is a function of z (because,
d(z − z ′ )n−1 w
da is independent from dz) which satisfies the preconditions in
ds
chapter 12 for these surface segments. Consequently, (z − z ′ )n−1 is also finite
and continuous in point O′ . If we describe its value in point O′ by an−1 , then
(z − z ′ )n−1 w − an−1 is a function which is continuously at this point, and which
= 0. Therefore, it becomes infinitely small through. From this we can conclude
an−1
according to chapter 12 that (z − z ′ )n−2 w − is a continuous function at
z − z′
′
point O . By continuing this procedure we can see that w gets turned into a
function which remains continuous and finite at point O′ through subtracting
a1 a2 an−1
an expression from the form + + ... + .
z − z′ (z − z ′ )2 (z − z ′ )n−1
Therefore, when this change occurs according to the preconditions in chapter
12 so that function w becomes infinitely large as O converges infinitely on a
point O′ inside of surface T , then this infinite quality’s order (when we consider
a magnitude that is increasing in reverse relationship to the distance as an
infinite magnitude of the first order) when it is infinite, will necessarily be a
whole number. And if this number = m, then function w can be changed into
a function that is continuous at this point O′ by the addition of a function that
contains 2m arbitrary constants.
Note: We consider a function as containing an arbitrary constant if the

23
possible varieties that it agrees with encompass a continuous domain of one
dimension.

14
The limitation which we established in chapter 12 & 13 for surface T are not
essential for the validity of the results we achieved. It is plain that we can
surround any point in the interior of an arbitrary surface with a piece of the
same surface. This piece will have the same properties that were presupposed
for that surface with the sole exception being the case where this point is a
branch point in the surface.
In order to investigate this case, we will assume that we can draw surface T ,
or an arbitrary piece of it which contains a branch point of the (n − 1)th order
of O′ where z = z ′ = x′ + y ′ i, by means of the function ζ = (z − z ′ ) n , onto
1

a different plane A. I.e., we can imagine the value of the function ζ = ξ + η i

at point O by a point Θ, whose rectangular coordinates are ξ, η and which is
represented in this latter plane. So we can consider Θ as the image of a point
O. This means that we get a connected surface extended over A as an image
of this segment of surface T . And as we will show very soon. this new surface
which has the image of point O′ in point Θ does not have any branch point.
In order to get a grasp of this mental image,we should think of a circle
around point O with a radius R on plane A. We will also draw a chord parallel
to the x-axis, where z − z ′ becomes a real value. Then the piece of surface T
which surrounds the branch point, and which we have cut out of the area by the
circle, will then separate into scattered half circle shaped surface segments on
both sides of the diameter in n, if R is kept sufficiently small. We will describe
these surface segments by a1 , a2 , . . . , an on the side of the chord where y − y ′ is
positive, and those surface segments on the other side by a′1 , a′2 , . . . , a′n . We will
also assume that a1 , a2 , . . . , an is the series associated with negative values of
z −z ′ and that a′1 , a′2 . . . a′n is the series associated with the positive values which
is connected to a′n , a′1 , . . . , an−1 . This way, a point that encircles point O′ (in
the required sense) runs through the series of surfaces a1 , a′1 , a2 , a′2 , . . . , an , a′n
and succeeds in getting back to a1 through a′n , which is an obvious assumption.
Next, we will introduce polar coordinates for both planes by setting φ
up z = z ′ =
′ n1 1
ϱe , ζ = σe and we will select that value of (z − z ) = ϱ e for depicting
φi ψi n n i

surface segment a1 whose expression comes under the assumption of 0 ≤ φ ≤ π.

1 π
So σ ≤ R n and 0 ≤ ψ ≤ will hold for all points in a1 and the image of
n
these points will all collectively be in plane A, in a sector stretching from ψ = 0
π 1
to ψ = of a circle drawn around Θ with a radius of R n . Naturally, every
n
point in a1 immediately corresponds to a point in this sector that is constantly
advancing along with it, and the reverse also holds. What follows then is that
the image of surface a1 is a simply connected surface extended over this sector.
π
In a similar manner, the image for surface a′1 is a sector stretching from ψ =
n

24
 2π 2π
to ψ = , the image for surface a2 is a sector stretching from ψ = to
n n
3π 2n − 1
ψ= , and the image for surface a′n is a sector stretching from ψ = π
n n
to ψ = 2π if we select φ for every point on this surface in the series between π
and 2π, 2π and 3π . . . (2n − 1)π and 2nπ which is always possible, and which
is only possible, in one way. These sectors also connect up with each other in
the very same manner as do surfaces a and a′ so that the points adjoining one
another in one sector correspond to points adjoining one another on another
sector. Therefore, we can combine these sectors into a connected image of one
of the pieces of surface T that includes point O′ . Obviously, this image is a
surface that is simply extended over plane A.
A variable magnitude that has a determinate value for every point O also
has a determinate value for every point Θ and the reverse also holds, because
every O corresponds to only one Θ, and every Θ only corresponds to one O.
Furthermore, if this variable magnitude is a function of z, then it is also a
dw dw
function of ζ, for when is independent of dz, is also independent of dζ.
dz dζ
The reverse also holds, and we get from this that we can apply the theorems
from chapter 12 and 13 ro all function w(z), even to the branch point O′ if we
consider them to be functions of (z − z ′ ) n . This gives us the following theorem:
1

When function w(z) becomes infinitely small through the finite convergence
of O to a branch point (n − 1)th order of O′ , then this infinite magnitude nec-
essarily has the same order with a power of distance, as that whose exponent is
1 m
a multiple of . If this exponent is = − , then this infinite magnitude can be
n n
changed into a function that is continuous at point O′ through adding an expres-
a1 a2 am
sion of the form 1 + 2 + ... + m , where a1 , a2 , . . . , am
′
(z − z ) n ′
(z − z ) n (z − z ′ ) n
are arbitrary complex magnitudes.
This theorem contains a corollary stating that function w is continuous at
point O′ when (z − z ′ ) n w becomes infinitely small as a result of the infinte
1

convergence of point O towards O′ .

15
We will now consider a function of z, which has a determinate value for every
point O on a surface T that arbitrarily extends over A, and which is not constant
everywhere. Picture it geometrically to that is value w = u + vi at point O is
represented by a point Q on plane B, whose rectangular coordinates are u, v.
We then get the following:
I. We can consider the totality of point Q as forming a surface S, in which
every point corresponds to a determinate point O that continuously keeps ad-
vancing in T as the point in S does.
In order to prove this, it is obviously only necessary to prove that the position
of point Q always (and of course, generally speaking, continuously) changes

25
along with the point O. This is contained in the theorem:
A function w = u + vi of z cannot be constant along a line unless it is
constant everywhere.
Proof: if w were to have a constant value a + bi along a line, then u − a and
∂(u − a) ∂v
, which is equal to = − would be equal to zero for this line and for
∂p ∂s
∂ 2 (u − a) ∂ 2 (u − a)
+ generally. And then according to chapter 11, I, u − a
∂x2 ∂y 2
∂u ∂v ∂u ∂v
and v − b too, (because of = , = − ) would also have to be equal
∂x ∂y ∂y ∂x
to 0 everywhere, which is contrary to our presuppositions.
II. As a result of the precondition we established in section I, there cannot
be any connection between the segments of S without connection among the
corresponding segments of T . The reverse is universal too, for where connec-
tion occurs in T and w is continuous, the surface S also has a corresponding
connection.
If we presuppose this, then S’s boundary corresponds on one hand to T ’s
boundary, and on the other hand to discontinuous positions. However its inner
segments, excluding isolated points, extends smoothly (schilicht) over B every-
where, i.e., there is neither a fissure in the segments lying on top of each other,
nor is there a fold anywhere either.
Because T is correspondingly connected everywhere, the first condition could
only occur if T underwent a fissure – which is contrary to our assumption. We
can prove the second condition is the same way.
dw
Next of all, we will prove that point Q′ , where is finite, cannot lie in a
dz
fold on surface T .
In reality, what we would do is surround point O′ , which corresponds to Q′
with a piece of surface T ′ that is of arbitrary form and indeterminate dimensions.
We could have to assume this piece’s dimensions to be so small (according to
Chapter 3) that the form (Gestalt) of the corresponding segment of S will
deviate in an arbitrarily small way, so that its boundary will exclude a piece
including Q′ from plane B. But this is impossible if Q′ lies in a fold on surface
S.
dw
So now if we consider as a function of z, according to I, it can only be
dz
equal to 0 in isolated points. And because w is continuous in the points of T
dw
that are under consideration, can only become infinite in the branch points
dz
of this surface. Therefore, Q.E.D.
III. Surface S therefore is a surface which satisfies the preconditions we
established in chapter 5 for T , and the indeterminate magnitude z has one de-
terminate value for every point Q on this surface. This one determinate value
dz
continuously changes with the position of Q in such a way that is inde-
dw
pendent of the change in location. Therefore, in the sense that was established
earlier, what we get forming is a continuous function of the variable complex

26
magnitude w for the entire magnitudinal field (Gebiet) presented by S.
What follows is :
Let O′ and Q′ be two corresponding interior points on surface T and S, and z =
z ′ and w = w′ in those same surface. Then if neither of these points are a branch
w − w′
point, will converge towards a finite limit, as O infinitely converges on
z − z′
O′ , and the image here will be similar down to the smallest segments. However,
if Q′ is a branch point of the (n − 1)th order, and O′ is a branch point of
(w − w′ ) n
1

the (n − 1)th order, then 1 approaches a finite limit as O infinitely

(z − z ′ ) m
converges on O′ . We can easily get a method of depicting the adjoining surface
segments from chapter 14.

16
Pedagogical Theorem 6 Let α and β be two arbitrary functions of x, y, for
∫ [( )2 ( )2 ]
∂α ∂β ∂α ∂β
which the integral − + + dT has a finite value as
∂x ∂y ∂y ∂x
it expands through all the segments of surface which is arbitrarily extended above
A. Then when we alter around continuous functions, or around functions which
are only discontinuous in isolated points, (both kinds of functions being = 0 at
their margin) the integral will always have a minimum value for one of these
functions. And if we exclude the discontinuities that occur by making changes
in isolated points, then we would only get a minimum value for one function.

We will define λ as being an indeterminate, continuous function or as a function

that is only discontinuous in a couple of points. It will be = 0 at its margin
∫ (( )2 ( )2 )
∂λ ∂λ
and the integral L = + dT which extends over the entire
∂x ∂y
surface will have a finite value for this function. Additionally we will define ω
as [an indeterminate of the function] α + λ and we will define Ω as the integral
∫ ( )2 ( )2
∂ω ∂β ∂ω ∂β
− + + dT , which extends over the entire surface.
∂x ∂y ∂y ∂x
The totality of the λ functions form a cohesive, self-contained domain in which
each one of these functions continuously change into others. However, these
functions themselves cannot infinitely discontinuously converge on a line, with-
out having L become infinite (chapter 17). This is so because when we assume
ω = α + λ for every λ, Ω becomes a finite value that becomes infinite along with
L, and that continuously changes with the form (Gestalt) of λ, but that can
never sink below null. Therefore it follows that Ω has a minimum for at least
one form (Gestalt) of the function ω.
In order to prove the second part of our theorem, let u be one of the functions
of ω which gives Ω a minimum value. Let h be a constant magnitude that is
indeterminate on the entire surface,so that u + hλ satisfies the preconditions set

27
up for function ω. Then the value of Ω for ω = u + hλ which
∫ [( )2 ( )2 ]
∂u ∂β ∂u ∂β
− + + dT
∂x ∂y ∂y ∂x
∫ [( ) ( ) ]
∂u ∂β ∂λ ∂u ∂β ∂λ
+ 2h − + + dT
∂x ∂y ∂x ∂y ∂x ∂y
∫ (( )2 ( )2 )
2 ∂λ ∂λ
+h + dT
∂x ∂y
= M + 2N h + Lh2

must therefore be greater than M fr every λ (according to the concept of the

minimum), as long as we assume h to he sufficiently small. But this then requires
2N
that every λN = 0 for otherwise 2N h + Lh2 = Lh2 (1 + ) would become
Lh
2N
negative when is counter-posed to N , irrespective of the signs < . Therefore
L
the value of Ω for ω = u+λ which is the form that obviously contains all possible
values for ω, becomes = M + L. Consequently, because L is essentially positive,
Ω cannot have a smaller value for any form (Gestalt) of function ω than ω = u.
Then if there is a minimum value M ′ of Ω for another u′ of the functions
ω, the same obviously holds for this. We will get M ′ ≤ M and M ≤ M ′ , and
consequently M = M ′ . But if we introduce u′ into the form u + λ′ then we
get the expression M + L′ for M ′ , as long as L′ describes the value of L for
λ = λ′ , and the equation M = M ′ gives L′ = 0. This is only possible when
∂λ′ ∂λ′
= 0, = 0 in all surface segments. Therefore, as long as λ′ is continuous,
∂x ∂y
this function is necessarily continuous. And because it is = 0 at its margin, and
it is not discontinuous along a line it can only have a value different from null, at
the most, in some isolated points. So then, two of the functions of ω, which give
Ω a minimum value, can only be different from each other in isolated points.
And if we put aside the discontinuities in function u that crop up by making
changes in isolated points, then this function is totally determinate.

17
We will now supply the proof that λ cannot infinitely converge on a discontin-
uous γ located on a line without prejudicing L’s infiniteness. I.e., if function λ
is subjected to the condition of agreeing with γ outside of a surface segment T ′
that includes the line of discontinuities, then we can always assume T ′ to be so
small that L must become larger than an arbitrarily given magnitude C.
Assuming s and p as having their usual relation in relation to the line of
discontinuity, we will define κ as the curvature of an indeterminate s, a curvature
which is convex on the side of the positive p, and which we will consider as
positive. We will define p1 as the value of p at the boundary of T ′ on the

28
positive side, and on the negative side by p2 . We will define the corresponding
values of γ as γ1 and γ2 . So if we now consider a continuity curved segment
of this line, and if the segment of T ′ that is contained between the normals
in the endpoints does not reach to the middle point of the∫ curvature,
∫ then
this segment of T ′ contributes the following expression to L: ds pp1 p2 (1 −
[( ) ( ) ] d
2
∂λ ∂λ 1
κp) + ; however, we find the smallest value of the
∂p ∂s (1 − κp)2
∫ p2 ( )2
∂λ
expression (1 − κp)dp at the fixed boundary values γ1 and γ2 of λ
p1 ∂p
(γ1 − γ2 )2 κ
to be equal to, according to well-known rules, to = .
log(1 − κp2 ) − log(1 − κp1 )
Therefore, we will have to∫ necessarily assume that every contribution, as
(γ1 − γ2 )2 κds
well as λ inside T ′ , to be > . Function γ would
log(1 − κp2 ) − log(1 − κ1 )
be continuous for p = 0 if the greatest value which could contain (γ1 − γ2 )2
for π1 > p1 > 0 and π2 < p2 < 0 were to become infinitely small through
π1 − π2 . Therefore we can assume that for every value of s there exists a
finite magnitude m so that no matter how small π1 − π2 is assumed to be,
m will always be contained inside the boundary value of p1 and p2 , which
are expressed by π1 > p1 > 0 and π2 < p2 < 0 (in which their equal-
ity is mutually excluded), and for which (γ1 − γ2 )2 > m. Furthermore, if
we arbitrarily assume a form (Gestalt) for T ′ in accordance with the earlier
limitations, we will give p1 and p2 the ∫ determinate values of P1 and P2 and
mκds
define a as the value of the integral which ex-
log(1 − κP2 ) − log(1 − κP1 )
tends through the segment of the ∫ line of discontinuities that we are considering.
(γ1 − γ2 )2 κds
Then we can obviously make > C to the ex-
log(1 − κP2 ) − log(1 − κP1 )
tend that we so assume p1 and p2 for every value of s so that the inequalities
a a
1 − (1 − κP1 ) C 1 − (1 − κP2 ) C
p1 < , p2 > and (γ1 −γ2 )2 > m will suffice. But
κ κ
this leads to the consequence that we assume that the segment of L that comes
from the piece of T ′ that we are considering, and therefore even to a greater
degree L itself, are larger than C, just as we would assume to be inside T .
Q.E.D.

18
According to chapter 16, the∫function
[( u which
) we established
( there]is = 0, as
)
∂u ∂β ∂λ ∂u ∂β ∂λ
are any of the functions N = − + + dT which
∂x ∂y ∂x ∂y ∂x ∂y
extend throughout all of surface T . We will now draw some further conclusions
from this equation.
Let us take a piece T ′ , that includes the discontinuous points, u, β, λ and

29
cut it out from surface Y . We can then find segment N , which is(based on ) the
∂u ∂β
remaining pieces T ′′ , with the aid of chapter 7 and 8, if we replace − λ
( ) ∂x ∂y
∂u ∂β
for X and + λ for Y ,
∂y ∂x
∫ ( 2 ) ∫ ( )
∂ u ∂2u ∂u ∂β
=− λ + 2 dT − + λds
∂x2 ∂y ∂p ∂s
As a consequence of the boundary ∫ ( conditions ) that have already been imposed
∂u ∂β
on function λ, the segment of + λds relating to the joint bound-
∂p ∂s
ary piece that T ′′ has with T is∫equal ( to 0. We)can then consider N to be
∂2u ∂2u
composed out of the integral − λ 2
+ 2 dT relative to T ′′ , and of
∫ [( ) ( ) ∂x
] ∫∂y( )
∂u ∂β ∂λ ∂u ∂β ∂λ ∂u ∂β
− + + dT + + λds relative to T ′ .
∂x ∂y ∂x ∂y ∂x ∂y ∂p ∂s
∂2u ∂2u
So now it is obvious, that if + were to be different from zero in any
∂x2 ∂y 2
segment of surface T , N would likewise behave a value different from 0 so long
as λ, which is(free, is equal)to zero inside of T ′ , and so long as we choose λ inside
∂2u ∂2u
T ′′ so that λ + 2 would have the same sign everywhere. However, if
∂x2 ∂y
∂2u ∂2u
+ are = 0 in all segments of T , then the component segment N which
∂x2 ∂y 2
is based on T ′′ vanished for every λ. The result of the condition N = 0 will
then be that component segments relating to discontinuous points = 0.
∂u ∂β ∂u ∂β
Concerning functions − , + therefore, what we get when we
∂x ∂y ∂y ∂x
have the first one = X and the latter = Y , if we just do not want to speak
∂X ∂Y
generally, is the equation + = 0. But, to extent that this equation
∂x ∫ ( ∂y )
∂x ∂y
really has a determinate value, X +Y ds = 0.
∂p ∂p
So (according to chapter 9,V) if surface T has the property of multiple
connection, we divide it by cuts
) into a simply connected T #. As a result,
∫ O(
∂u ∂β
the integral − + ds has the same value for every line in T #’s
O0 ∂p ∂s
that goes from O0 to O. And when we consider O0 to be fixed, this integral
will then also form a function for x and y that undergoes a continuous change
and a change that is equal on both sides of a cut in T #. When we add this
function ν to β we get a function v = β + ν whose differential quotient is
∂v ∂u ∂v ∂u 10
=− and = .
∂x ∂y ∂y ∂x
Therefore we have the following:
10 Here, again, the English translation doesn’t agree with the original German, according to

which a modification has been made. – Typesetter

30
Pedagogical Theorem 7 Assume that a complex function is given a con-
nected surface T which in turn is divided by cuts into a simple connected surface
∫ [( )2 ( )2 ]
∂α ∂β ∂α ∂β
T #. In terms of this function − + + dT extends
∂x ∂y ∂y ∂x
throughout the entire surface and has a finite value, so that it can always be
changed and can only be changed into a function of z through the addition of a
function which satisfies the following conditions:
1. is = 0 at the margin, or is only different from it in isolated points, while
v is arbitrarily given for a point.
2. the changes made by in T , and changes made by v in T #, only occur
∫ [( )2 ( )2 ]
∂µ ∂µ
in isolated pints and are so discontinuous that + dT
∂x ∂y
[
∫ ( )2 ( )2 ]
∂ν ∂ν
and + dT remain finite as they extend through the
∂x ∂y
entire surface, and the latter expression remains equal on both sides along
the cut.
These conditions’ adequacy in determining µ + ν i stems from having µ,
through which we determined ν up to an additive constant, always furnish a
minimum for the integral Ω at the same time. This is so because given u = a+µ,
N will obviously be = 0 for every λ: a property which can only belong to one
function, according to chapter 16.

19
The principles which are the basis for the pedagogical theorem at the conclusion
of the previous chapter open up the path for investigating the determinate
functions of a variable complex magnitude (independent of an expression for
the same.)
A quick review of the range of the conditions that are necessary for the deter-
mination of such a function inside a given numerical domain (Gr´’ossengebiets)
will serve us as an orientation to this field.
First of all, we will pause at a specific case: If the surface which is extended
over A (which is how we will represent this numerical domain) is a simply
connected surface, then the function w = u+vi of z will be determined according
to the following conditions:
1. a value is given for u in all the boundary points, and when this value
undergoes an infinitely small change of position, it changes by an infinitely
small magnitude of the same order. Otherwise, the value will change
arbitrarily.11
11 In themselves, the changes in this value are only subject to the limitation that they are not

discontinuous along a part of the boundary. We have only imposed the additional limitation
in order to avoid formal difficulties which are unnecessary here. – Riemann’s original footnote

31
 2. the value for v ar any point is arbitrarily given.

3. the function should be finite and continuous at all points. The function is
totally determined by these conditions.
In reality, this does follow from the pedagogical theorem in the previous
chapter, if we so define α + βi so that α at the margin is equal to the given
value, and if the change in α + βi is infinitely small and of the same order
for every infinitely small change of location in the entire surface. It is always
possible for us to define α + βi this way.
Generally speaking, therefore, we can have u at the margin be as a totally
arbitrary function of s, and we can also define v anywhere through this. We can
assume the reverse, too, for if v is arbitrarily given for all boundary points, then
the value for u follows from this. So the full range for the choice of values for
a w at the margin encompasses a one-dimensional manifold for every boundary
point. In order to totally define this manifold, what we need for every boundary
point is an equation for which it is not essential that every one of the equations
is solely based on the value of one term in one boundary point. Our definition
can also turn out in such a way, so that what we get for every boundary point if
an equation containing both terms that continuously changes its form (Form)
as the position of this boundary point changes. Or, what can happen simulta-
neously to several segments of the boundary is that every point defined as an
(n − 1)point of this segment gets matched to one point in such a way that for
every n amount of such points, we collectively get an n amount of equations
that continuously change with their locations. However, these conditions, whose
totality constitutes a continuous manifold, and which are expressed by compar-
ison (equations) between arbitrary functions, generally speaking, still require
either limitation of amplification by means of isolated conditional equations –
equations for arbitrary constants – in order to get a reliable and adequate defi-
nition for a function that is continuous everywhere inside a numerical domain.
These conditions require this, that is, to the extend that the accuracy we used
in our evaluations does not reach up to this level.
Our observations will not have to undergo any essential modifications of the
situation where magnitude z’s domain of variability is represented by a multiply
connected plane because the application of the theorem in chapter 18 creates a
function constituted as before, excepting the changes that occur in overstepping
the cuts – changes, which can be made = 0 if the boundary conditions contain
an amount of disposable constants that are equal to the number of cuts.
The situation in the interior, where we have relinquished all claims for conti-
nuity along a line, organises itself like the previous situation if we consider this
line to be a cut on the surface.
And finally, if we allow continuity to be violated at an isolated point, then,
according to chapter 12, this is how a function becomes infinite. So, by retaining
the special preconditions that we made in the first case for this point, function z
can be arbitrarily given after its concurring function becomes continuous. How-
ever, as a result of this, function z becomes completely defined. For if we assume

32
the magnitude which is in an arbitrarily small circle drawn around the discon-
tinuous point to be equal to the given function, and, moreover, to also conform to
∫ (( )2 ( )2 )
∂α ∂β ∂α ∂β
the earlier formulae, then the integral − + + dT =
∂x ∂y ∂y ∂x
0 when it is extended over this circle, and equals a finite magnitude when it is
extended over the remaining segment. And so we can apply the theorem from
the previous chapter, through which we get a function with the desired proper-
ties. From this we can generally derive, with the aid of the theorems in chapter
13, that when a function can become infinitely large to the nth order in a dis-
continuous point, then a number of 2n constants become available.
According to chapter 15, let us assume a function w of a variable complex
magnitude z that is inside of a given magnitude domain of two dimensions.
Then when we represent this function geometrically, we will get an image S
covering B that is similar down to its smallest segment to a surface T covering
a given A. The only exception to this are isolated points. And, we will not base
the value of the conditions that are necessary and sufficient for defining this
function on either boundary points or on discontinuous points. Consequently,
(according to chapter 15) the conditions that define this function all turn out
to be the conditions for the position of S’s boundary, and naturally, they give a
conditional equation for every boundary point. So if every one of these condi-
tional equations only relates to one boundary point, then we can represent them
by a group of curves, each one of which forms the geometrical location for each
boundary point. If we then jointly subject two boundary points that keep in
step with each other continuously to two conditional equations, what we then
get is such a dependency (slope) between the two boundary segments that when
we arbitrarily assume a position for one point, we can derive the position of the
other point from it. In like manner, we can also get something of geometrical
importance out of the other forms of the conditional equations, but we do not
want to pursue this further here.

20
The origin and the immediate purpose for the introduction of complex number
into mathematics is the theory of creating simpler12 dependency laws (slope
laws) between complex magnitudes by expressing these laws through numerical
operations (Grössenoperationen). And, if we give these dependency laws an
expanded range by assigning complex values to the variable magnitudes, on
which the dependency laws are based, then what makes its appearance is a
harmony and regularity which is especially indirect (versteckt) and lasting. Of
course, up until now the situation in which this occurs have encompassed a small
12 Here we will consider Addition, Subtraction, Multiplication and Division as elementary

operation, and we will consider a dependency law (slope law) to be all the more simpler –
the fewer are the elementary operations that determine the dependency. In reality, all of the
functions that have been used up to now in this analysis can be defined by a finite number of
these operations. – Riemann’s original footnote

33
domain – we can almost totally trace these situations back to those very laws
covering the dependency between two variable magnitudes, where one function
is either an algebraic function of the other 13 or is that kind of function whose
differential quotient is an algebraic function. – But in almost every step that we
have taken here, we have not just simply given a simpler, more consistent Gestalt
to our results without any help from complex magnitudes. Our steps have also
pioneered the way for new discoveries, and the account of our examination of
algebraic function, circular – or exponential function, elliptical and Abelian
functions furnished the evidence for this.
We will now briefly indicate what the theory of these functions has gained
through our examination.
The previous methods that were used to deal with these functions always
has, as the basis of their definition, and expression of the function through
which the function’s value would be given for every value in the argument. Our
examination has shown as a result of the general character of a function of a
variable complex magnitude, what we get in a definition of this kind is that any
one segment of the pieces making up the definition is a direct consequence of the
remaining segments, and of course, we can trace the range of pieces making up
the definition back to those pieces that are necessary for the definition, which
essentially simplifies our treatment of the definition. For example, in order to
prove that two expressions of the same functions are equal, we would have had
to previously show that both agree for every value of the complex magnitude.
But now, the evidence of their agreement in a considerable smaller range is
sufficient.
A theory of these functions that is based on the foundations that we have
supplied here would define the function’s configuration (Gestalting)(i.e. its value
for every value in the argument), independent of the method of determining this
through numerical operations(Grössenoperationen). For in this new definition,
we would only add the features that are necessary to define the function to the
general conception of a function of a variable complex magnitude. And only
then would we add these features to the various expressions which the function
is capable of undergoing. We can then express the common characteristic of a
species of function, which could be expressed in a similar manner by numerical
operations, in the form of the boundary – and discontinuity conditions that are
imposed on the functions.
Assume, for example, that magnitude z’s domain of variability extends either
simply, or multiply over all of infinite plane A, and that inside this same plane
our function is discontinuous only in isolated points. We will also only tolerate
a function that is becoming infinite and whose order is finite. As a result, we
will consider this magnitude itself to be an infinite magnitude of the first order
1
for an infinite z ′ , but we will consider to be an infinite magnitude of the
z − z′
′
first order for every finite value of z . So, the function is necessarily algebraic,
and conversely, every algebraic function fulfils this condition.
13 I.e., where an algebraic equation occurs between both. – Riemann’s original footnote

34
 In our paper, we have abstained for now from realising this theory, because
as we remarked, this realisation would be characterised by bringing simple de-
pendency (slope) laws that are conditional on numerical operations out into the
light of day. We have not done this so far because we have rule out considering
the expression of such a function for the present.
And for these very same reason, we also did not concern ourselves here
with our theorem’s usefulness as the foundation of a general theory of these
dependency (slope) laws. What we would need for this is a proof that the
concept of a function of variable complex magnitude, which is our basis here,
is in complete agreement with a dependency (slope) that is expressible by a
numerical operation. 14

21
Nevertheless, a detailed example of its application can be of use in illustrating
our general theory.
The application of our theory which was described in the previous chapter is
only a special application, even though it was intended to be our first example.
Assume a dependency is conditioned by a finite number of the numerical oper-
ations that we considered to be elementary operations in the previous chapter.
Then its function contains only a finite number of parameters that succeed in
having no arbitrary determinate conditions at all occur under them along a line
at any point. This is so regardless what the form is of the system of mutually
independent boundary – and discontinuity conditions that are adequate to de-
fine the function. Therefore it seems better suited for our present purpose if
we do not select an example that comes from that situation, but if we instead
take an example where the function of the complex variable is dependent on an
arbitrary function.
In order to make an assessment, and to get a more comfortable framework
we will give our example the geometrical form that we used at the end of chapter
19. What we will then appear to have is an investigation of the possibility of
producing an analogous image, connected down to its smallest segments, of a
given surface. The image’s Gestalt is given in the form that was expressed
above, where there is locational curve for every boundary point in the image,
and where the locational curve is given for all these boundary points, with the
exception of the boundary and branch point as given in chapter 5. We will
limit ourselves to solving this problem for the situations where every point in
one surface will only correspond to one point in the other surface, and where
14 This includes every one of the dependencies that can be expressed by a finite or infinite

number of the four simple methods of calculation: addition and subtraction, multiplication
and division. In terms of magnitude operations (Grössenoperationen, usually tran. previously
as numerical operations. – Note from the translator ), in contrast to counting operations
(Zahlenoperationen), their expression themselves should indicate those methods of calculation
which do not bring these magnitudes’ commensurability into question. – Riemann’s original
footnote

35
the surfaces are simple connected. This situation is contained in the following
pedagogical theorem.
Pedagogical Theorem 8 Two simply connected surfaces can always relate to
each other in such a manner that every point on one surface corresponds to the
point on the other surface that is steadily progressing with it, and so that their
smallest corresponding segments are similar. Naturally, we can arbitrarily give
corresponding points to the interior points in one surface, and to the boundary
points on another, but this is what determines the relationship for all points.
If two surfaces T and R are so related to a third surface S that their smallest
corresponding segments are similar to S’s, then a relation develops out of this
between surface T and R which is obviously the same as the first relationship.
We can trace our task, which consists of relating two arbitrary surfaces to each
other so that they are similar in their smallest segments, back to portraying
every arbitrary surface through another surface which we define as similar down
to its smallest segments. According to this, when we draw a circle K with the
radius 1 around the point in plane B where w = 0, all we have to prove in order
to follow our pedagogical theorem is that: we can portray an arbitrarily, simply
connected surface T that covers A, as a continuously connected surface, and one
that is similar down to its smallest segment, by circle K in such a manner, and
only in such a manner, that an arbitrarily given interior point O0 corresponds to
the circle’s middle point, and an arbitrarily given boundary point O’ on surface
T corresponds to an arbitrarily given points on the circle’s periphery.
We will describe the meanings defined for z, Q for point O0 , and O′ by
corresponding indices, and we can ascribe the middle point of an arbitrary circle
Θ, which does not reach up to T ’s boundary, and which does not have any branch
points, as being around O0 in T . With the introduction of polar coordinate to
the extent that we have z − z0 = reφi then the function log(z − z0 ) = log r + φi.
As a result of this, all the real values in the entire circle change continuously,
except for point O0 , whose value becomes infinite. But wherever we select the
smallest possible value for φ among all possible values, the imaginary value has
0 value on the one side, and the value of 2π on the other side along the radius
where z − z0 assumes real positive values. In all other points, however, the
imaginary numbers change continuously. Obviously, this radius can be replaced
by a totally arbitrary line L drawn from the middle point to the periphery, so
that function log(z − z) undergoes a sudden diminution of about 2πi when point
O crosses over from the negative side of this line (i.e., where p becomes negative
according to chapter 8) to the positive side. Elsewhere, however, the function
continuously changes with the position O has in circle Θ. If we also assume
that in circle Θ complex function α + β i of x, y = log(z − z0 ), except when we
arbitrarily expand l up to the margin, then the function
1. will become totally imaginary at the margin of T , and on Θ = log(z −z0 )’s
periphery,
2. will change by approximately −2πi in crossing over from the negative to
the positive side of L, and otherwise, it will change by an infinitely small

36
 magnitude of the same order with every small change in location, all of
which becomes increasingly more possible.
∫ (( )2 ( )2 )
∂α ∂β ∂α ∂β
Therefore, integral − + + dT , has a value of null
∂x ∂y ∂y ∂x
when it expands across Θ, and when it extends across the remaining segments
it has a finite value. Therefore, we can change α + βi into a function t = m + ni
of z through the addition of a continuous function of x, y which is continuously
determinate with the exception of a totally imaginary constant remainder, and
which is totally imaginary on the margin. The real segment m of this function
will be = 0 on the margin, will be = −∞ at point O0 , and will continuously
change in all the rest of T . Therefore, for every value a of m that lies between 0
and −∞, T , disintegrates as the result of a line where m = a, disintegrates into
segments where m is smaller than a and where O0 is contained on the inside,
and disintegrates into segments on one side and the other side where m > a
and where these segments’ boundaries are partially formed by T ’s margin, and
partially through lines where m = a. As a result of this disintegration: either
the order of surface T ’s connection does not change, or it is reduced. And so,
because this order is equal to −1, the surface disintegrates into two pieces, or
into more than two pieces. But this latter situation is impossible because then m
would have to be finite and continuous everywhere in at least one of these pieces,
and constant in all the segments of the boundary. As a result, either there would
have to be one constant value in each surface segment, or there would have to
be a minimum or a maximum value anywhere – in a point or along a line, which
is contrary to article 11,III. So then the points where m is constant form simple,
self-encircling lines everywhere, lines which bound a piece including 0, m will
necessarily decrease going towards the interior, which results in n increasing
continuously, as long as it is continuous in a positive range, (where s increases
according to chapter 8). And if once again we disregard multiples of 2π then
every value between 0 and 2π becomes equal because n only undergoes a quick
change of about −2π 15 in crossing over from the negative side of line L to its
positive side. If we then have et = w then em and n will become the polar
coordinates of point Q in relation to the middle point of circle K. The totality
of point Q will then obviously form a surface s that extends simply over K
everywhere; point O0 itself will then be at the middle point of the circle, and
point Q′ can be backed into an arbitrarily given point on the periphery with
the help of the constants that are still available in n.
Q.E.D.
For the case where point O0 is a branch point of the (n − 1) order, and if we
1
replace log(z − z0 ) with log(z − z0 ), then we will use every similar conclusions
n
15 Because line l leads from a point lying in the interior to one lying outside, then it must

go one more time from the inside to the outside then it goes from the outside to the inside,
if it crosses the boundary several times. Therefore, the sum of the sudden changes in n in a
positive range will always be −2π.

37
to reach the goal whose further exploitation we can easily fill out from chapter
14.

22
We will not completely carry out the investigation of the general case in the
last chapter, where one point in one surface should correspond to several points
in other surfaces, and where we do not make the prerequisite that these points
just have simple connections. We will not carry this out completely because
our entire investigation has had to lead to a general Gestalt, if we comprehend
it from a geometrical viewpoint. For this reason, it was not essential that
we limited ourselves to level, smooth (schlicht) surfaces with the exception of
isolated points; rather, our task has been to portray one arbitrarily given surface
onto another arbitrarily given one so that they are similar down to their smallest
segments, or to give it very similar treatment. We will content ourselves here by
referring to two Gaussian treatises which are cited in chapter 3 and the general
inquiry about surfaces in chapter 13.

38