Vous êtes sur la page 1sur 56

1

Collected Papers

Bernhard Riemann

Translated from the 1892 edition by Roger Baker,Charles Christenson


and Henry Orde

Kendrick Press

Preface
Bernhard Riemann(1826-1866)was one of the greatest mathematicians of the
modern era. He aimed very deliberately at new ways of thinking about existing
problems and concepts in mathematics,often with starting success.Looking for a
quantitative way to support the assertion that he inuenced twentieth century
mathematics more than his contemporaries, I turned to the collection Development of Mathematics 1900-1950, edited by J-P.Pier. Riemann is mentioned in
the index as many times as Gauss, Cauchy, Weierstrass and Dedekind combined.
Over the years, I have often talked to mathematicians and students who
would like to read Riemanns papers, but nd this dicult or time-consuming in
the original German.This is the audience I had in mind in organizing the present
volume. A few of the papers have been translated before,but the translations
in this book are all new.
Rieman was fortunate enough to attend the lectures(on the least squares
method) of the legendary C.F.Gauss(1777-1855) as a young student in Gottigen.
He came to idolize Gauss later,but in 1847 he traveled to Berlin for greater stimulus. There he attended the lectures, and took part in the discussions, of
C.G.J.Jacobi(1804-1851) and P.G.L.Dirichlet(1805-1859). Both these scholars
became important inuences on Riemann. He cited Jacobis great memoir Fundamenta nova theoriae functionum ellipticarum in many places in his own work,
and it suggested possibilities to him that went well beyond Jacobis beautiful calculations.Dirichlet, on the other hand,liked to take an abstract approach to each
topic, and Riemann much preferred this way of attacking problems. Dirichlet
was also an enthusiastic proponent of mathematical models of physical problems, and Riemann was to carry Dirichlets approach further in both his writings
and his lectures. Returning to Gottigen,he also bacame red with enthusiasm
for experimental physics and construction of theoretical explanations for the
new phenomena that were being observed.Here his great mentor was Wilhelm
Weber(1804-1891).
As a young researcher, Riemann was drawn in many directions.However, his
doctoral thesis of 1851 on the foundations of the theory of functions of a complex
variable took on great importance in suggesting to him several further lines of
research that needed to be carried through. The thesis also became a watershed
in the subject, and in 1951 was celebrated with a centennial conference. There
were man inspired ideas in the thesis. Riemann surfaces appear there for the rst
time. Analytic functions are viewed as conformal mappings, and the Riemann
mapping theorem is proved and given a proof, albeit unsatisfactory. Riemanns
researches later in the 1850s on hypergeometric functions, Abelian functions and
the Riemann zeta function demonstrate his commitment to complex function
theory as a tool of exploration. Yet Dieudonne(1985) writes of Riemanns great
paper VI on Abelian functions as an epoch in algebraic geometry. It took a very
long time before all the mathematical ideas such as A.Clebsh(1833-1872) felt
somewhat defeated by Riemanns memoir, although Clebsh took up the potent
idea of genus of an algebraic curve from it. The paper VII on the Riemann
zeta function, treasured by number theorists as a gem, is the source of the most

4
famous unsolved problem in mathematics,the Riemann hypothesis.
Simply in preparing for his habilitation in 1854,Riemann wrote two works of
geniushis habilitation thesis of trigonometric series, and his trial lecture on
the foundation of geometry. The methods in the thesis can still be found intact
in modern works such as Zygmunds Trigonometric Series. The trial lecture
became the inspiration for a new era in dierential geometry,and Einsteins
theory of general relativity is its (not very indirect) descendant. Yet Riemann
never published either the thesis or the trial lecturethis was left to his friend
and colleague Richard Dedekind (1831-1916) after Riemanns death. In 1854,
these works simply qualied Riemann to be a poorly paid instructor at Gottigen.
It is hard to believe, yet while working on these wonderful ideas, Riemann
still had a great deal of time to think about problems of physics. Of the nine
papers that he published during his lifetime, four are on problems of mathematical physics (Riemanns total of publications up to 1866 is brought up to
eleven by announcements of his papers on hypergeometric functions and the
propagation of sound waves.). Much of his writing on physics from the 1850s
was not submitted for publication.
It is dicult to guess the directions in which Riemann might later have
focused his extraordinary abilities. He succeeded Dirichlet as a professor at
G
ottingen in 1859, but his health deteriorated from 1862 onwards. He remained
dedicated to scholarship to the last; this touching story is told here in an essay
by Dedekind.Among his contemporaries,Dedekind was the one who most fully
appreciated Riemanns mathematics.
After Riemann died lately in 1866, Dedekind and others oversaw the publications of seven posthumous papers. The early death of Clebsch delayed the
appearance of collected papers, but eventually Heinrich Weber(1842-1913)was
able to send the rst edition to the press in 1876. Weber was ably assisted by Dedekind and by H.A.Schwarz(1848-1921). A dozen papers in the rst
edition were assembled by poring over Riemanns Nachlass, the mass of materials left behind at his death. The 1892 edition,which is the source for the
present translations,contained some modest additions and corrections, and its
numbering I-XXXI of the papers is preserved here for the readers convenience.
However, I omitted three non-mathematical items: XVIII,The mechanism of
the ear, and two fragments on philosophy that can be found on pages 509-525
of Weber(1892).
Footnotes in the text of the papers are mostly Riemanns.However,some
footnotes making Riemanns reference to the literature more precise were added
by Webber. These are indicated by a W.
There is a far more in Riemanns work than any one person can comprehend
if the inuence of the papers is to be properly considered. Nevertheless, I provide
notes ant the end of the book containing basic information about each paper
and suggestions for further reading.
A natural next step would be the book of Laugwitz(1999), which gives a unied account of Riemann as a mathematician and natural philosopher. Riemann
also had a very important role as teacher and expositor through the revision
and publication of his lectures by Hattendor, Stahl and others.

5
Readers who would like to suggest corrections or alternative readings within the papers,or additional remarks for the notes,should send these to baker@math.byu.edu.I will maintain a web page that takes these suggestions into
account, at http://www.math.buy.edu/baker/Riemann/index.htm
During the nal stages of the preparation of this book, Henry Orde, one
of my fellow translators, died in Kent, England. Henry was born in 1922. He
served in the British armed forces throughout the second world war, and then
studied mathematics at Cambridge.After working for a rm of merchants in
Malaya for several years, he joined the nascent computer industry in England.
A heart attack forced him into retirement at age 50, and he was then able
to devote his energies to pure mathematics, book collecting and music.Number
theories will recall that he gave an elementary proof of the class number formula
for quadratic elds with negative discriminant. See Orde(1978).
Roger Baker

7
Cotents
I.Foundations for a general theory of functions of a complex variable.
Contents
II.On the laws of distribution of electric charge in a ponderable body that is
neither a perfect conductor nor insulator,but may be viewed as resisting the
holding of electric charge with nite power.
III.On the theory of Nobilis color rings.
IV.Contributions to the theory of the functions etc.
V.Authors announcement of the preceding paper.
VI.The Theory of Abelian Functions.
VII.On the number of primes less than a given magnitude.
VIII.On the progagation of planar air waves of nite amplitude.
IX.Authors announcement for VIII.
X.A contribution to the study of the motion of a homogeneous uid ellpsiod.
XI.On the vanishing of -functions.
XII.On the representation of a function by a trigonometric series.
Contents.
XIII.The hypotheses on which geometry is based.
XIV.A contribution to electrodynamics.
XV.A proof of the theorem that a single-valued periodic function of n variables
cannot be more than 2n-fold periodic.
XVI.Extract from a letter written in Italian on January 21,1864 to Professor
Enrico Betti.
XVII.On the Surface of least area with a given boundary.
XIX.An attempt to generalize integration and dierentiation.
XX.New theory of residual charge in apparatus for static charge.

XXI.Two general theorems on liner dierential equations with algebraic coecients.


XXII.A mathematical work that seeks to answer the question posed by the most
distinguished academy of Paris.
Webers remarks on XXII.
XXIII.On the representation of the quotient of two hypergeometric series as an
innite continued fraction.
XXIV.On the potential of a ring.
XXV.Diusion of heat in an ellipsoid.
XXVI.Equilibrium of electricity on cylinders with circular cross-section and parallel axes.Conformal mapping of gures bounded by circles.
XXVII.Examples of surfaces of least area with a given boundary.
XXVIII.Fragments concerning limiting cases of the elliptic modular functions.
Commentary on XXVIII.
XXIX.Fragment on analysis Situs.
XXX.Convergence of p-fold innite theta-series.
XXXI.On the theory of Abelian functions.
Natural Philosophy.
The life of Bernhard Riemann.
Notes on the Individual Papers.
Bibliography.

10
Foundations for a general theory of functions of a complex variable
(Inaugural dissertation,G
ottigen 1851;second printing (unchanged),G
ottigen 1867.)
1.
Denote by z a variable that can take successively all possible real vales.
When there is a unique value of the variable w corresponding to each z, we say
that w is a function of z. If w varies continuously when z runs continuously
over all values between two given points, we say that the function is continuous
in this interval.
This denition clearly enforces no law between individual values of the function. For when the function is specied in a certain interval, the method of
continuing it outside the interval remains entirely arbitrary.
The dependence of the quality w on z can be expressed by a mathematical
law, so that denite operations at each value of z yield the corresponding w.
The possibility of a single law of dependence for all values of z in a given interval
was formerly ascribed only to a certain class of functions(functiones continuae
in Eulers terminology). More recent researchers have shown, however, that
there are analytic expressions that represent each continuous function on a given
interval. Hence it is one and the same thing to say that w depends on z in some
arbitrary given manner; or that w is given by denite operations. Both notions
are equivalent, in view of the results mentioned.
This is not the case, however, if z is not restrict
real values, but varies over
complex numbers of the form x + yi(where i = 1).
Let
x + yi, x + yi + dx + dyi
be two values of the quantity z with an innitely small dierence, and let
u + vi, u + vi + du + dvi
be the corresponding values of w. If the dependence of w on z is dened arbitrary, the ratio
du + dvi
dx + dyi
varies, generally speaking, according to the values of dx and dy. Indeed, let
dx + dyi = ei ; Then
1 u v
1 v
u
du + idv
= (
+
)+ (

)+
dx + idy
2 x y
2 x y
1 u v
v
u du idv
[

+ i(
+
)]
2 x y
x y dx + idy
1 u v
v
u i
1 u v
+
)+ [

+ i(
+
)]e .
= (
2 x y
2 x y
x y

11
In whatever way w is determined from z by a combination of a simple operations,
1
the value of the derivative dw
dz will always be independent of the particular
value of the dierential dz. Obviously we cannot obtain arbitrary dependence
of complex w on complex z in this way.
The above characteristic, common to all functions obtained via any operations, will be fundamental to the following investigation, where such a function
will be treated independently of its expression. Without proving the general
suciency and validity of the denition of dependence via operations, we take
the following denition as starting point.
A complex variable w is said to be a function of another complex variable
z, if w varies with z in such a way that the value of the derivative dw
dz is independent of the value of the dierential dz.
2.
Both quantities z and w will be treated as variables that can take every
complex value. It is signicantly easier to visualize variation over a connected
two-dimensional domain, if we link it to a spatial viewpoint.
We represent each value x + yi of the quantity z by a point O of the plane
A having rectangular coordinates x, y; and every value u + vi of the quantity w
by a point Q of the plane B, having rectangular coordinates u, v. Dependence
of w on z is then represented by the dependence of the position of Q on the
position of O. When w corresponds to z in a way that varies continuously with
z, or in other words when u, v are continuous functions of x, y, then to every
point of the plane A corresponds a point of the plane B; generally speaking, to
every line corresponds a line, and to every connected piece of surface there is a
corresponding connected piece of surface. Thus we can think of this dependency
of w on z as a mapping of the plane A on the plane B.
3.
We now investigate the properties that this mapping has when w is a function
of z, that is, when dw/dz is independent of z.
We denote by o a general point of the plane A in the neighborhood of O,
and the image of o in the plane B by q. Further let x + yi + dx + dyi and
u + vi + du + dvi be the value of z and w at these points. We may view dx, dy
and du, dv as rectangular coordinates of the points o and q with respect to the
points O and Q taken as origins. If we write dx+dyi = ei and du+dvi = ei ,
then the quantities , , , become polar coordinates of these points relative
to these origins. Now let o and o be any two points innitely close to O.
For quantities depending on o , o , we use the above notations with appropriate
accents. By hypothesis
du + dv i
du + dv i
=
dx + dy i
dx + dy i
1 This assertion is obviously justied in all the cases where one can obtain from the expressions of w in terms of z, using the rules of dierentiations, an expression for dw
in terms of
dz
z. The rigorous general validity of the assertion is left aside for now.

12
Consequently
du + dv i
( )i
dx + dy i
=
e
=
du + dv i

dx + dy i

= e( )i

and so

=
, =

That is, in the triangles o Oo and q Qq the angles o Oo ,q Qq are equal, and
the corresponding sides are proportional.
This yields the similarity of two corresponding innitely small triangles;and
consequently, in general, similarity of the smallest parts of the plane A and their
images on the plane B. An exception to this result occurs only in the special
case where the corresponding variations of the quantities z and w are not in
nite ratio.We tacitly excluded this exception from our deduction2 .
4.
If we write the dierential quotient du+dvi
dx+dyi in the form
( u
x +

v
x i)dx

v
+ ( y

u
y i)dyi

dx + dyi
it is plain that it will have the same value for any two values of dx and dy,
exactly when
u
v v
u
=
,
= .
x
y x
y
Hence this condition is necessary and sucient foe w = u + vi to be a function
of z = x + yi. For the individual terms of the function, we deduce the following:
2u 2u
2v
2v
+
=
0,
+
= 0.
x2
y 2
x2
y 2
This equation is the basis of for the investigation of the properties of the individual terms of such a function. We give a proof of the most important of
these properties before undertaking a deeper treatment of the complete function. However, we rst establish some points concerning more general matters,
in order to smooth the ground of the investigation.
5.
2 On

this subject, see:General solution of the problem of mapping the portions of a surface to that the image is similar to the original in the smallest parts,C.F.Gauss.(Response
to the 1822 prize question proposed by the Royal Society of Sciences in Copenhagen. Astronomische Abhandlungen,edited by Schumacher,vol.III,Altona 1825;Gauss, Collected works,vol.IV,p,189.)

13
For the following treatment we permit x, y to vary over a nite region. The
position of the point O is no longer considered as being in the plane A, but
in a surface T spread out over the plane. We choose this wording since it is
inoensive to speak of one surface lying on another, to leave open the possibility
that the position of O can extend more than once over a given part of the plane.
However, in such a case we suppose that the portions of surface lying upon one
another do not connect along a line. Thus a folding of the surface, or a splitting
of the surface into superimposed parts, does not occur.
Now the number of pieces of surface superimposed in each part of the plane
is completely determined, when we give the boundary of the region and its
direction(that is, its inner and outer sides). A transit of these pieces can take
dierent forms.
Indeed, if we draw a line through the part of the plane covered by the surface, the number of superimposed surfaces only changes on crossing boundary.
Indeed, the number changes on moving from outside to inside by +1, in the
opposite -1. Thus the number is determined everywhere. Along the edge of the
line, each bordering portion of surface on continues in a denite way,as long
as the line does not meet the contour, For indeterminacy can only occur at
an isolated point, and consequently occurs either at a point of the line itself,
or at a nite distance from it. If we conne ourselves to part of the line in
the interior of the surface, and both sides of it on a suciently narrow strip of
surface, we may speak of denite bordering parts of the surface,whose number
is equal on either side. Specifying a denite direction of the line, denote the
parts of the surface on the left side by a1 , a2 , , an and those on the right by
a1 , a2 , , an .Each part a will continue into one of the parts a . Indeed, this will
in general be the same part along the length of the line , except that for certain positions of it can change at one point. Suppose that above such a point
(that is,along the preceding part of ) the surface portions a1 , a2 , , an
are connected respectively with a1 , a2 , , an ,while below , a1 , a2 , , an
connect respectively with a1 , a2 , , an .Here 1 , 2 , , n is a permutation of
1, 2, , n. A point of a1 that moves into a1 ,above , will pass to a1 if we go
back to the left side below . If a point moves around from left to right, the
index of the part of surface in which it lies takes successively the values
1, 1 , 1 , , , ,
In this sequence, as long as the term 1 do not recur, the terms are necessarily
distinct. For an arbitrary intermediate term is necessarily preceded by ,
and in direct succession by all preceding terms up to 1. But when after a certain
number of terms, say m, a number evidently less than n, the term 1 reappears,
the order terms must then recur in the same order.The point moving around
comes back after every m circuits into the same part of the surface and is
restricted to m superimposed surface parts, the meet above at a unique point.
We call this point a branch point of order m 1 of the surface T . Repeating the
procedure, the n m remaining surface parts, provided they are not isolated,
divide into systems of m1 1, m2 1, are located at .

14
When the position and direction of the boundary of T , and the position of the
branch points, are given, then T is either completely determined, or restricted
to one of a nite number of distinct forms. The latter occurs, in so far as these
determining portions can lie on dierent parts of the superimposed surfaces.
A variable which in general, that is after excluding isolated lines or points3 ,
takes a denite value at each point O of the surface T that varies continuously
with position,can obviously be considered as a function of x, y. When we refer
to functions of x, y in what follows,the term will always be employed in this
fashion.
Before passing to the treatment of such functions, we introduce some clarications on the connectivity of a surface. We restrict ourselves to surfaces that
do not split apart along a line.
5.
We regard two parts of surface as connected, or being to a single piece,
if from any point of one part to any point of the other, a line can be drawn
interior to the surface. We regard two parts as separate, when this procedure is
not possible.
The study of the connectivity of a surface is based on its decomposition via
transverse cuts, that is lines which cut through the interior from one boundary point simply(no point occurring multiply) to another boundary thereby
adjoined,and thus at an earlier point of the transverse cut.
A connected surface that is split apart by each transverse cut, is said to be
simply connected; otherwise it is multiply connected.
Theorem 1 A simply connected surface is divided by each transverse cut ab
into two simply connected pieces.
Suppose that one of the piece is not split apart by a transverse cut cd. There
are three possible cases: neither or the endpoints c, d is situated on ab; only c is
on ab. Rejoining the surface respectively along the whole line ab; along the part
cb; or along the part cd of the line, we obtain a connected surface that arises
from a transverse cut of A, contrary to hypothesis.
Theorem 2 Suppose that a surface T divides via n1 transverse cuts4 q1 into a
system T1 of m1 simply connected pieces, and via n2 transverse cuts q2 into a
system T2 of m2 pieces.Then n2 m2 n1 m1 .
3 This restriction does not arise from the denition of a function, but is needed for the
application of innitesimal to it. As an example of a function discontinuous at all points of
a surface, take the function whose value is 1 for rational x and y, and otherwise is 2. We
cannot apply to the function either dierentiation or integration, and so cannot directly use
innitesimal calculus. The arbitrary restriction placed here on the surface T will be justied
later(section 15).
4 By a decomposition through several transverse cuts we understand a successive decomposition. That is, the surface obtained by one transverse cut is decomposed further by a new
transverse cut.

15
Every line q2 , not completely contained in the system of transverse cuts q1 ,
yields together with the system of cuts q1 one or several transverse cuts q2 of
the surface T1 .As endpoints of the transverse cuts q2 ,we have:
1) The 2n2 endpoints of the transverse cuts q2 , except for those whose endpoints meet part of the system of lines q1 .
2) Every intermediate point of a transverse cut q2 , at which this cut meets
an intermediate point of a line q1 , except for the case where the point is already
on another line of q1 , that is, when an end of a cut q1 coincides with this point.
Denote now by the number of times that lines of both systems join or
separate(where in consequence an isolated common point is to be counted twice);
by 1 ,the number of times an end part of a line q1 coincides with an intermediate
part of a line q2 ; by 2 , the number of times an end part of a line q2 coincides
with an intermediate part of a line q1 ; and nally, by 3 the number of times
an end portion of a line q1 coincides with an end part of a line q2 . Then case 1)
yields 2n2 2 3 endpoints, case 2) yields 1 endpoints, of the transverse
cuts q2 . Both cases taken together yield all endpoints, each one only once. The
number of these transverse cuts is thus
2n2 2 3 + 1
= n2 + s.
2
In an entirely analogous way we conclude that the number of transverse cuts
q1 of the surface T2 , formed by the lines q1 , is
2n1 1 3 + 2
= n1 + s.
2
Now the surface T1 will obviously be transformed by the n2 + s transverse
cuts q2 into the same surface obtained from T2 by the n1 + s transverse cuts q1 .
But T1 comprises m1 simply connected pieces and consequently, by Theorem 1,
is decomposed by n2 + s transverse cuts into m1 + n2 + s pieces. Consequently,if
m2 < m1 + n2 n1 , the number of pieces of the surface T2 would be increased
by more than n1 + s by the eect of n1 + s transverse cuts, which is absurd.
As a consequence of this theorem, if the indeterminate number of transverse
cuts is denoted by n, and the number of pieces by m, then n m is constant
for all divisions of the surface into simply connected pieces. For consider any
two divisions, by n1 transverse cuts into m1 pieces and by n2 cuts into m2
pieces.Since the rst set of pieces are simply connected, we have
n2 m 2 n1 m 1 .
Since the second set are simply connected, we have
n1 m 1 n2 m 2 .
and since both inequalities hold, we have n2 m2 = n1 m1 .
This number can properly be designated the connectivity of a surface. By
denition, it will be diminished by 1 by a transverse cut. I will be unchanged by

16
the eect of a simply cut, starting from a point of the interior and ending either
at a point of the boundary, or at a point of a previous cut. It will be increased
by 1, by a simply cut in the interior of the surface with two endpoints. For in
the rst case, the cut becomes a transverse cut, if we make a new transverse
cut. In the latter case, two new transverse cuts are needed.
Finally the connectivity of one surface formed several pieces is obtained by
adding the connectivity of these pieces.
In the following, we usually restrict ourselves to surfaces that have only one
pieces, and for their connectedness we simply speak of simply, double, and so
on. Here we understand, by an n-fold connected surface, a surface that can be
decomposed by n 1 transverse cuts into one that is simply connected.
As to the dependence of the connectedness of the boundary on the connectedness of the surface, it is clear that:
1) the boundary of a simply connected surface necessarily comprises a single
closed line.
Suppose the boundary comprised two separate pieces. A transverse cut q
that joins a point of one piece a to a point of another piece b, merely divides
connected parts of the surface from each other. For one can form a line interior
to the surface along a, starting from one side of the transverse cut q and ending
on the opposite side. So q does not split the surface, contrary to hypothesis.
2) Each transverse cut either decreases by 1, or increases by 1, the number
of pieces of the boundary.
For a transverse cut q there are three cases: Either q joins a point of one
piece of the boundary a, to a point of another piece b. In this case, all these lines
taken in the sequence a, q, b, q form a single closed piece of the boundary. Or,
q joins two points of a single piece of the boundary. In this case the boundary
falls into two pieces via the two endpoints, each of which, taken together with
the transverse cut, gives a closed piece of the boundary.
Or nally,q tends at one of its own preceding points. In this case, q can be
considered as comprising a closed line o and a line joining a point of o to a
point of a boundary piece a. In this case o on the one hand, and a, , o, on the
other hand, are closed paths each forming a piece of the boundary.
Instead of two boundary pieces we obtain one, in the rst case; and two
instead one, in the last two cases. This yields our theorem.
The number of pieces comprising the boundary of an n-fold connected portion of surfaces, is thus either n, or less than n by an even number.
From this we obtain a corollary: If the number of boundary pieces of an nfold connected surface is n, the surface is split into two separate parts by every
simple closed cut in its interior.
For the connectivity is not altered by this cut and the number of boundary
pieces is increased by 2. Consequently, if the new surface were connected, it
would be n-tuply connected with n + 2 boundary pieces, which is impossible.
7.
Let X and Y be two functions of x, y continuous at all points of the surface
T spread out over A. With the surface integral extended over all elements dT

17
of this surface, we have

X
Y
(
+
)dT = (X cos + Y cos)ds
x
y
Here, at each point of the boundary, denotes the inclination of the interior
normal to the x-axis, and denotes its inclination to the y-axis. On the right
side, the integral extends over all elements
ds of the boundary line.

In order to transform the integral X


dT
, we divide the part of the plane A
x
covered by T into strips, via a system of lines parallel to the x axis, in such a way
that each branch point of T falls on one of these lines. With this assumption,
each part of T corresponding to one of these strips
is formed of one or more

separate trapezoidal pieces.The contribution to X


dT
of one of these surface
x
strips, cutting o an element dy of the y axis, will obviously be

X
dx
dy
x
Here the integration is taken over the (one or more) straight lines belonging to
the surface T that fall on a normal issuing from some point of this element dy.
Now let the lower endpoints of these lines be O1 , O2 , O3 , , (that is, corresponding to the smallest value of x); the upper points O(1) , O(2) , O(3) , ,and
denote by X1 , X2 , , X (1) , X (2) , the value of X at these points. Let ds1 , ds2 ,
, ds(1) , ds(2) , be the corresponding elements cut o from the boundary of
the surface strips, and 1 , 2 , , (1) , (2) , the value of at these elements.
Then

X
dx = X1 X2 X3 + X (1) + X (2) + X (3) +
x
The angles are evidently acute at the lower endpoints, obtuse at upper endpoints; consequently
dy = cos1 ds1 = cos2 ds2 = = cos (1) = cos (2) =
Substituting these values yields

X
dy
dx =
X cosds
x
the summation extending over all boundary elements whose projection onto the
y-axis is dy.
By integration over all the elements dy that occur, it is clear that all elements
of the surface T and all elements of the boundary will be exhausted. Accordingly
we obtain, with the integration taken over the perimeter,

X
dT = X cosds
x

18
By entirely analogous reasoning, we obtain

Y
dT = Y cosds
y
so that

X
Y
+
)dT =
x
y

(X cos + Y cos)ds

as we wished to prove.
8.
On the boundary line, we denote by s the distance of a general point Oo ,
from a xed initial point, in a direction to be a xed below. On the normal at
Oo let p denote the distance of an indeterminate point O from Oo ; the distance
is taken to be positive for interior points of the normal. The values of x and y
at the point O can clearly be considered as functions of s and p. At the points
of the boundary line we then have the partial derivatives
x
y
x
y
= cos ,
= cos ,
= cos ,
= cos
p
p
s
s
Here we take the upper sign if the direction in which s increases makes the
same angle with p as the x-axis with the y-axis. In the opposite case, we take
the lower sign. We take the direction of increasing s throughout the boundary
in such a way that
x
y
y
x
=
, and so
=
s
p
s
p
This does not essentially restrict the generality of our results.
Obviously we can extend this rule to lines interior to T . Here, in determining
the signs of dp and ds, if their mutual dependence is xed as above, we must still
indicate how to x the sign of dp or of ds. For a closed line,we shall indicate
one of the two parts into which it cuts the surface, and take the line as its
boundary, which determines the sign of dp. For a line that is not closed, we
indicate instead the origin of the line, that is, the endpoint where s takes the
least value.
Substituting the values obtained for cos and cos into the equation proved
in the previous section, with the domains of integration as before, we have

X
Y
x
y
(
+
)dT = (X
+Y
)ds
x
y
p
p

y
x
= (X
Y
)ds
s
s

9.

19
We apply the result at the end of the previous section to the case where,
throughout the surface,
X
Y
+
=0
x
y
This yields the following theorems.
I.Let X and Y be two functions nite and continuous throughout T satisfying
Y
X
+
=0
x
y
Then, integrating over the entire boundary of T ,

x
y
(X
+Y
)ds = 0
p
p
Consider an arbitrary surface T1 spread over A and divide it in arbitrary
fashion into two pieces T2 , T3 . The integral

y
x
+Y
)ds
(X
p
p
over the boundary of T2 can be regarded as the dierence between the integrals
over the boundaries of T1 and T3 . For where T3 borders T1 the two integrals
cancel out; every other element corresponds to an element of the boundary of
T2 .
Using this transformation, we deduce from I:
II.The value of the integral

x
y
(X
+Y
)ds
p
p
taken over the boundary of a surface spread over A, remaining constant when
the surface is increased or diminished in size in a manner that does not cross
any part of the surface where the hypotheses of Theorem I are violated.
When the functions X, Y satisfy the above dierential equation throughout
the surface T , but have discontinuities at isolated lines or points, we can surround each of these lines or points with an arbitrary small region of surface.
Applying Theorem II, we obtain:
III.The integral

x
y
(X
+Y
)ds
p
p
over the boundary of T is equal to the sum of the integrals

x
y
(X
+Y
)ds
p
p
taken around all discontinuities. The contribution to the integral from each
discontinuity is constant, however narrow the boundary that encloses it.

20
The contribution from a point of discontinuity is necessarily zero if, denoting distance of O from the discontinuity, X and Y become innitely small
with . Take such a point, as origin, and an arbitrary direction as initial line,
for polar coordinates , ; choose as the boundary curve a circle of radius and
center at this point. Then the integral in question is expressed by

(X
0

y
x
+Y
)d
p
p

This cannot take a nonzero value , since whatever the value of , we can take
y

so small that the absolute value of X x


p + Y p is smaller than 2 for every
, giving
2
x
y
(X
+Y
)d <
p
p
0
IV.Suppose that for a simply connected surface spread over A, we have

x
y
(X
+Y
)ds = 0
p
p
when we integrate around an arbitrary part if the surface; or

x
y
(Y
X )ds = 0
s
s
Then this integral, taken between any two xed points Oo and O, has he same
value for any line from Oo to O.
Any two paths s1 and s2 joining Oo to O, taken together, form a closed line
s3 . Either this line itself passes through no point more than once, or one can
divide it into several simple closed lines as follows. Start from an arbitrary point
and describe the contour. Each time we meet a point already traversed, separate
o the intermediate part, and consider the following part as the immediate
extension of the preceding portion of the curve. Each such line divides the
surface into a simply connected piece and a doubly connected piece, and forms
the entire boundary of one of these pieces. The integral

x
y
(Y
X )ds
s
s
corresponding to the simply connected piece will be 0 by hypothesis. Consequently the same property holds for this integral taken over the whole path s3 ,
with the quantity s treated as increasing in the same direction through-out. It
follows that the integrals along the lines s1 and s2 cancel out, if there is no
change in this direction; that is, one line passes from Oo to O and the other
from O to Oo . If the alter the sense of the latter integral, they become equal.
Consider any surface T in which, generally speaking,
Y
X
+
=0
x
y

21
We remove the discontinuities if necessary, so that for any portion of the
remaining pieces of the surface

y
x
(Y
X )ds = 0
s
s
and reduce the remaining surface via transverse cuts to a simply connected
surface T . For each path in the interior of T , from a point Oo to another
point O, the above integral has the same value, which for brevity we denote by
O
y
x
X )ds
(Y
s
s
Oo
The integral, where Oo is xed and O is arbitrary, is independent for each
O of the path joining the points; so it can be treated as a function of x, y. The
variation of this function when O moves along an arbitrary line element ds is
(Y

x
y
X )ds
s
s

this is continuous everywhere in T and equal on both sides of a transverse cut


of T .
V.The integral

Z=

(Y
Oo

y
x
X )ds
s
s

thus represent for xed Oo a function of x, y continuous throughout T , but


varies by a constant on crossing anywhere along a transverse cut of T from one
branch point to another. The function has partial derivatives
Z
Z
= Y,
= X
x
y
The variations on crossing transverse cuts depend on certain quantities, independent of each other, equal in number to the number of transverse cuts. For
when one runs over the system of transverse cuts in a retrograde sense-the later
parts rst-this system variation is everywhere determined if its value at the start
of each cut is known. However, the latter values are independent of each other.

10.
For the functions so far denoted by X and Y , take
u

u
u u
u
u , u
u
x
x y
y

respectively. Then
Y
2 u
2 u
2u 2u
X
+
= u( 2 +
) u ( 2 + 2 )
2
x
y
x
y
x
y

22
Thus if u, u satisfy the equations
2u 2u
2 u
2 u
+
=
0,
+
=0
x2
y 2
x2
y 2
we have

X
Y
+
=0
x
y

We can apply the results of the previous section to

y
x
+Y
)ds
(X
p
p
which is equal to

(u

u
u
u )ds
p
p

Suppose now that u and its rst order partial derivatives never have discontinuities along a line. Suppose further that at each point of discontinuity,
u
denoting distance from O to the discontinuity, u
x and y become innitely small with . According to the remark on III of the previous section, the
discontinuities of u can be disregarded.
For, given a straight line from one of these discontinuities, take a value R of
such that
u
u x
u y

=
+

x rho
y rho
remains nite for < R. Let u = U for = R, let M be the largest absolute
value of u
in the interval of the line with 0 < < R. Then, again disregarding
sign,
u U < M (log log R)
so that (u U ), and indeed u, become innitely small with . By hypothesis,
u

this holds also for u


x , . Consequently, if u does not become discontinuous,
the same holds for
(u

u
u
u
u
u ) and (u
u )
x
x
y
y

so that the case considered in the previous paragraph applies here.


Suppose now that the surface T formed from the positions of O is simply
wherever it covers A. Take an arbitrary xed point Oo of T where u = u0 , x =
x0 , y = y0 . The quantity
1
log((x x0 )2 + (y y0 )2 ) = log r
2
considered as a function of x and y, has the property that
2 log r 2 log r
+
=0
x2
y 2

23
and has a discontinuity only for x = x0 , y = y0 ; and so in our case for only one
point of the surface T .
Consequently, from Section 9, III with u = log r, the integral

log r
u
(u
log r )ds
p
p
has the same value for the whole boundary of T as it does for an arbitrary
circuit around the point Oo . Take for this circuit the circumference of a circle
on which r is constant, and denote by the angle of a radius with endpoint
O, measured from an arbitrary point of the point of the circumference in some
denite direction. The integral is
2

log r
u

u
rd log r
ds
r
p
0
which, since

reduces to

u
ds = 0
p

ud = 0
0

This value becomes u0 2 for innitely small r if u is continuous at Oo .


Under the above hypotheses on u and T , then, we have, for an arbitrary
point Oo interior to the surface, where u is continuous,

1
u
log r
u0 =
(log r
u
)ds
2
p
p
with the integral taken over the boundary; and
2
1
u0 =
ud
2 0
with the integral taken over a circle around Oo . The rst of these expressions
yields the following
Theorem 3 Suppose that, in integral, the function u satises
2u 2u
+ 2 =0
x2
y
inside a simply surface T spread over the plane. Suppose further that
1) the points where this dierential equation is violated do not ll a region
of the surface;
u
2) the points in which u, u
x , y are continuous do not ll any line;
u
3) for each discontinuity the quantities u
x , y become innitely small along
with the distance of O from the discontinuity, and,

24
4) a discontinuity of u, that can be removed by modifying the value of u at
isolated points, is excluded.
Then u and all its partial derivatives are continuous and nite at all points
interior to the surface.
For consider Oo as a variable point. The only quantities that vary in the
expression

log r
u
u
)ds
(log r
p
p
r log r
are log r, log
x , y . However, these quantities, along with their partial
derivatives, are nite and continuous functions of x0 , y0 at every point of the
boundary while Oo remains in the interior of T ; the partial derivatives may be
expressed as rational functions of these quantities, containing only powers of r in
the denominator. The continuity property persists for the value of our integral
and consequently for the function u0 . Now u, by the earlier hypotheses, can
only dier in value from our integral ar isolated points where u is discontinuous.
This possibility is excluded by hypothesis 4) of our theorem.

11.
Under the hypotheses on u and T at the end of the previous paragraph, we
have the following theorems:
I.If u = 0 and u
p = 0 along a line, then u = 0 everywhere.
we show rst of all that a line where u = 0 and u
p = 0 cannot be the
boundary of a region a of the surface, where u > 0.
For suppose that this did take place. From a take a piece bounded on one
side by and on the other side by a circle; the center of the circle Oo being
excluded from this piece. This construction is always possible. With the integral
taken over the boundary of this piece, and with r, denoting polar coordinates
of O with respect to Oo , we have

u
log r
log r ds u
ds
p
p
By hypothesis it follows that, with the integral taken only over the circle,

u
ud + log r
ds = 0
p
Since

we obtain

u
ds = 0
p

ud = 0

which is incompatible with the hypothesis that u > 0 in the interior of a.


In a similar way we show that the equations u = 0 and u
p = 0 cannot hold
on the boundary line of a region of the surface b, where u is negative.

25
Suppose now that u and u
p are 0 on a line in the surface T , and there is a
region of T where u = 0. Clearly such a region is bounded either by this line or
by a region of the surface where u = 0. Consequently, it is always bounded by
a line where u and u
p are 0, which leads necessarily to one of the hypotheses
rejected above.
II.If u and u
p have given values along a line, then u is determined throughout
T.
Let u1 and u2 be any two functions that satisfy the above conditions on u.
These conditions also apply to the dierence u1 u2 . Suppose that u1 , u2 are
identical, along with their rst order partial derivatives with respect to p, on a
line, but dier in some region of the surface. Then along this line u1 u2 = 0

and p
(u1 u2 ) = 0, while u1 u2 is not 0 everywhere, contrary to Theorem I.
III.If u is not a constant in T , then the points in T where u takes a constant
value are necessarily lines which divide regions where u is larger from regions
where u is smaller.
This theorem is obtained by combining the following.
u cannot have a minimum or maximum at an interior point of T ;
u cannot be a constant in just a part of T ;
The lines in which u = a cannot bound portions of the surface on both sides
where u a has the same sign.
These are all propositions whose negation, as we readily see, would lead a
violation of the equation
2
1
u0 =
ud
2 0
or

(u u0 )d = 0
0

proved in the previous section. The negation is consequently impossible.


12.
We now turn back to the treatment of a complex variable w = u + vi which,
generally speaking(that is, excluding isolated lines and points) has a unique
value for each point O of the surface T that varies continuously with the position
of O. Moreover we suppose that the equation
u
v u
v
=
,
=
x
y y
x
hold except at the excluded values. As indicated earlier, we say that w is a
function of z = x + iy. To simplify the following, we suppose that a function of
z has no discontinuity that can be removed by changing its value at an isolated
point.
The surface T will at rst be supposed to be simply connected, and to be
spread out simply over the plane A.
Theorem 4 Suppose that the set of discontinuities of the function w of z do
not contain a line. Further suppose that at an arbitrary point O of the surface,

26
where z = z , (z z )w becomes innitely small as O tends to O . Then w,
along with all its derivatives, is nite and continuous at all points interior to
the surface.
Let z z = ei . The hypotheses on the variation of the quantity w have
the following consequence for u and v.
v
1) u
x y = 0 and
u
v
2) y + x
= 0 through T ;
3) the functions u and v are not discontinuous along a line;
4) for every point O , the quantities u and v become innitely small with
the distance of O from O ;
5) neither u nor v has discontinuities that can be removed by changing the
value at an isolated point.
As a consequence of the hypotheses 2),3),4), for any portion of the surface
T we obtain

y
x
v )ds = 0
(u
s
s
the integral being taken over the boundary, from III, Section 9. Further the
integral
O
x
y
(u
v )ds
s
s
Oo
(by IV, Section 9) has the same value along every line from Oo to O, and for xed
Oo forms a continuous function U of x, y except for isolated points.Recalling 5),
U
the partial derivatives satisfy U
x = u and y = v at every point. By substituting these values for u and v, the hypotheses 1),3),4) become the condition of
the theorem at the end of the Section 10. Accordingly the function U is nite
and continuous, along with all its partial derivatives, at every point of T . Thus
U
the same holds for the complex function w = U
x y i, and its derivatives with
respect to z.
13.
We now investigate what transpires when, retaining the other hypotheses of
Section 12, we suppose that, for a specic point O interior to the surface,
(z z )w = ei w
does not be innitely small when O tends to O . In this case w becomes innitely
large when O tends to O . We suppose that, if the quantity w is not of order
magnitude 1 , that is, their quotient does not have a nite bound, then at least
the order of the two quantities is in nite ratio. That is, there is a power of
whose product with w either becomes innitely small , or remains nite. If
is the exponent of such a power and n is the smallest integer with n > ,
then the quantity (z z )n w = n eni w becomes innitely small with . Now
d
((z z )n1 w) is independent of dz).
(z z )n1 w is a function of z (since dz
This function satises, in this region of the surface, the hypotheses of Section

27
12, and consequently is nite and continuous at the point O . Denote its value
O by an1 . Then (z z )n1 w an1 is a function that is continuous at O and
vanishes there, and consequently becomes innitely small with . By Section 12,
n1

(z z )n2 w azz
is continuous at the point O . Continuing this procedure,
it is obvious that on subtracting an expression of the form
a1
a2
an1
+
+ +
z z
(z z )2
(z z )n1
w becomes a function that remains nite and continuous at O .
Accordingly, if we vary the hypotheses of Section 12, by permitting w to
become innitely large as O tends to a point O in the interior of T , then the
order of this nite quantity, if nite, is necessarily an integer. (A quantity
inversely proportional to the distance from O to O is considered to have order
1.) If this integer is m, then on attaching to it a function containing 2m arbitrary
constants, w becomes a function continuous at O .
Remark We consider a function to contain one arbitrary constant, when the
possible determinations of the constant comprise a continuous one-dimensional
domain.
14.
The restrictions on the surface T of Section 12 and 13 are not essential for the
validity of the results obtained. Obviously one can surround each point interior
to an arbitrary surface with a piece of the surface which has the properties
assumed there. The only exception is the case where this point is a branch
point of the surface.
To investigate this case, consider the surface T , or an arbitrary piece of it,
containing a branch point O of order n1, where z = z = x +y i, and mapped
by the function = (z z )1/n onto another plane . That is, we represent
the value of the function = + i at the point O as a point of with
rectangular coordinates , , and treat as the image of O. The image of the
region T obtained in this way is a connected surface spread over , without a
branch point at the image and O , as we now show.
To x ideas, draw a circle of radius R around the point O in the plane A and
draw a diameter parallel to the x-axis, so that z z is real on this diameter.
The piece of the surface T , around the branch point, cut o by this circle, is
divided on both sides of this diameter into n separate pieces of semicircular
form, provided R is suciently small. The surface portions on the side of the
diameter where y y is positive will be denoted by a1 , a2 , , an ; those on
the opposite side by a1 , a2 , , an . We suppose further that for negative values
of z z , a1 , a2 , , an are connected respectively to a1 , a2 , , an , while for
positive values they are connected respectively to a1 , a2 , , an . Then a point
O , encircling the branch point in the appropriate direction, runs in succession
over the surfaces a1 , a1 , a2 , a2 , , an , an and form an back to a1 ; this hypothesis
is obviously permissible. We introduce polar coordinates on both planes, writing
z z = ei , = ei , and choose as the image of the surface portion a1
the value (z z )1/n = 1/n ei/n . Here we suppose that 0 . Then

28
R1/n , 0 /n for all points of a1 ; its image in lies in a sector from
= 0 to = n of a circle around of radius R1/n . Indeed, to each point of a1
corresponds a point of this sector, varying continuously along with it, and vice
versa. Thus the image of a1 is a connected surface spread out simply over this
sector. Analogously, the images of a1 , a2 , , an are respectively sectors from
2
3
2n1
= n to = 2
n , from = n to = n , , from =
n to = 2.
Here is successively to run from to 2, from 2 to 3, , from (2n 1)
to 2n for the points of these surfaces, which is possible in exactly one way.
These sectors follow one another in the same manner as the surfaces a and
a , in such a way that coincident points correspond to coincident points. The
sectors can thus be joined together to give the connected image of a portion of
T surrounding O , and this image is obviously a surface spread out simply over
the plane .
A variable that has a denite value foe each point O likewise has a denite
value for each point O likewise has a denite value for each point and conversely, since each O corresponds to only one and each to one O. If the
variable is a function of z, then it is a function of , and conversely. We conclude
that the results of Section 12 and 13 are applicable to all functions w of z at
branch points O , if they are treated as functions of (z z )1/n . This yields the
following result:
If a function w of z becomes innite as O tends to a branch point O of order
n 1, then this innite quantity is necessarily of the same order as a power of
the distance from O to O whose exponent is a multiple of n1 . If this exponent
is m
n , then on attaching an expression of the form
a1
a2
am
+
+ +

1/n

2/n
(z z )
(z z )
(z z )m/n
where a1 , a2 , , am are arbitrary complex number, the function becomes continuous at O .
This theorem has the corollary that the function w is continuous at O if
(z z )1/n w becomes innitely small as O tends to O .
15.
Now consider a function of z with a denite value for each point of an
arbitrary surface T spread over A, not everywhere constant. If we represent the
value w = u+iv at O geometrically as a point Q of the plane B with rectangular
coordinates u, v, we have the following consequence.
I.the totality of points Q can be considered as forming a surface S, in which
each point corresponds to one point O in T that varies continuously with Q.
To prove this, obviously we only need to show that the position of the point
Q always changes along with that of O (and, generally speaking, changes continuously). This is obtained in the following results:
A function w = u + vi of z cannot be constant along a line, unless w is
everywhere constant.
Proof. If w takes the constant value a + bi along a line, then u a and

29
(ua)
p

= v
s vanish not only on this line but everywhere, along with
2 (u a) 2 (u a)
+
x2
y 2

By section 11, I,u a = 0. Since


v u
v
u
=
,
=
x
y y
x
we also have v b = 0 everywhere, contrary to hypothesis.
II.By the hypothesis made in I, two parts of S cannot be connected together
unless the corresponding parts of T are connected together. Conversely, whenever a connection occurs in T and w is continuous, a corresponding connection
holds in S.
Assuming this, the boundary of S corresponds in part to the boundary of
T and on part to the discontinuities. The interior of S, however excluding
isolated points, is spread over B smoothly. That is, the surface never splits into
superimposed surfaces or folds back on itself.
The rst could only happen, since T is connected in the corresponding fashion, if T splits in this way-contrary to hypothesis. The latter case is treated
next.
We show rstly, that a point Q where dw
dz is nite, cannot lie in a fold of
the surface S.
To see this, we surround the point O corresponding to Q , with a piece of
the surface T of any shape and unspecied size. According to Section 3, the size
can be taken so small that the shape of the corresponding piece on S diers as
little as w wish from that of the piece on T . Consequently its boundary cuts
out a portion of B that surrounds Q . However, this is impossible if Q lies on
a fold of the surface S.
By I, dw
dz , as a function of z, can only vanish at isolated points. Since w is
continuous at the points of T considered here, dw
dz can only become innite at
the branch points of the surface. This yields the desired result.
III.Consequently, S is a surface obeying the hypotheses that were imposed
on T in Section 5. In this surface, the variable quantity z has one denite value
for each point Q, which varies continuously with the position of Q in such a way
dz
that dw
is independent of the direction of variation. Accordingly z, in the sense
of the term specied earlier, is a continuous function of the complex quantity w
over the domain represented by S.
From this, we further obtain:
Let O and Q be corresponding interior points of the surface T and S, at

which z = z , w = w . If neither of them is a branch point, ww


zz tends to a

nite limit as O tends to O , and the image of an innitely small portion is


similar to that portion. However, if Q is a branch point of order n 1 and O
1/n
)
has a nite limit as O tends to
is a branch point of order m 1, then (ww
(zz )1/m
O . For the adjacent regions we have a form of mapping, readily obtained from

30
Section 14.
16.
Theorem 5 Let and be arbitrary functions of x, y for which the integral

2
2
[(

) +(
+
) ]dT
x
y
y
x
taken over the arbitrary surface T spread over A, is nite. If we vary by
functions vanishing on the boundary that are continuous, or only discontinuous
at isolated points, the integral takes a minimum value for one of these functions.
If we exclude discontinuities that can be removed by changing the function at
isolated points, the minimum is obtained for only one function.
Let be an unspecied function vanishing on the boundary, either continuous or possessing only discontinuities at isolated points, for which the integral

L = (( )2 + ( )2 )dT
x
y
taken over the whole surface, is nite. Denote by an unspecied member of
the set of functions + , and nally denote by the integral

2
2
[(

) +(
+
) ]dT
x
y
y
x
over the whole surface. The family of functions forms a closed connected
domain, in that any function passes continuously into any other, while no limit
of the functions is discontinuous along a line unless L simultaneously becomes
innite (Section 17). For each , writing = + , the integral has a nite
value, that tends to innity with L and varies continuously with the form of ,
but can never be less than 0. Consequently has a minimum value for at least
one form of the function .
To prove the second part of our theorem, denote by u one of the functions
for which attains its minimum value. Let h be an unspecied constant
function, so that u + h satises the required conditions for the function . For
= u + h, we write the value of as

u 2
u 2
u
u
[(

) +(
+
) ]dT + 2h [(

)
+(
+
) ]dT
x
y
y
x
x
y x
y
x y

+ h2 (( )2 + ( )2 )dT
x
y
2
= M + 2N h + Lh
say. This must be at least M for every (by denition of the minimum),
provided h is suciently small. Now it follows that N = 0 for each . Otherwise,
2N h + Lh2 = Lh2 (1 +

2N
)
Lh

31
would be negative if h is opposite in sign to N and of absolute value < 2N
L . The
value of for = u + , a form which obviously contains all possible value for
, is thus M + L. Since L is essentially positive, can attain for no form of
the function a smaller value than it attains for = n.
Now suppose that for another u the function yield a minimum value M
of . The same deductions give M M and M M , consequently M = M .
Write u in the form u + ; then we obtain for M the expression M + L , where
L denotes the value of L for = . The equation M = M yields L = 0. This
is only possible if, throughout the surface,

= 0,
=0
x
y
wherever is continuous, it is necessarily a constant. Since vanishes at the
boundary and is not discontinuous along a line, it can be nonzero at isolated
points. Thus two functions for which attains a minimum, dier only at
isolated points. If all discontinuities of u that can be removed by changing its
value at isolated points are discard, the function is completely determined.
17.
We now need to supply the proof of the fact that , if L is to remain nite,
cannot tend to a function discontinuous along a line. That is, if is supposed
to coincide with outside a portion of surface T that surrounds the line of
discontinuity, and T is taken suciently small, then L exceeds an arbitrary
given number C.
We assign s and p their usual meanings with respect to the line of discontinuity. For indeterminate s, denote by the curvature, with a convex curvature
on the side of positive p taken to be positive. Let p1 denote the value of p on the
boundary of T on the side where p < 0. Denote the corresponding values of
by 1 and 2 . Consider any portion of this line with continuous curvature. The
part of T between the normals at its endpoints, taken only up to the center of
curvature, contributes to L the quantity

p1

1
ds
dp(1 p)[( )2 + ( )2
]
p
s (1 p)2
p2
The smallest value of the expression
p1

( )2 (1 p)dp
s
p2
for xed boundary values 1 and 2 of , by known procedures, is
(1 2 )2
log(1 p2 ) log(1 p1 )
Consequently the above contribution is necessarily

(1 2 )2 ds
>
log(1 p2 ) log(1 p1 )

32
however we choose in the interior of T .
The function will be continuous for p = 0 if the greatest value of (1 2 )2 ,
for 1 > p1 > 0 and 2 < p2 < 0, becomes innitely small with 1 2 .
Consequently, for each value of s we can choose a nite quantity m such that,
however small 1 2 is taken, there are values of p1 and p2 with 1 > p1
0, 2 > p2 0 (the equality signs mutually exclusive) for which
(1 2 )2 > m
Consider an arbitrary form of T consistent with the earlier restrictions. Let
P1 and P2 be the values of p1 and p2 for this form, and denote by a the value
of the integral

mds
log(1 P2 ) log(1 P1 )
over the part of the discontinuity line in question. Obviously we can obtain

(1 2 )2 ds
>C
log(1 p2 ) log(1 p1 )
by choosing p1 and p2 for each value of s so that the inequalities
p1 <

1 (1 P1 )a/C
1 (1 P2 )a/C
, p2 >

and
(1 2 )2 > m
are satised. It follows that, however is dened in the interior of T , the contribution to L from the piece of T in question exceeds C. Thus L itself exceeds
C, as we wished to prove.
18.
By Section 16, for the function u xed there and for any function , we have
N = 0. Here

u
u

)
+(
+
) ]dT
N = [(
x
y x
y
x y
the integral extends over T . We now draw further consequences from this equation.
Cut from T a piece T surrounding the discontinuities of u, , . The contribution to N from the remaining part T of T is

2u 2u
u
( 2 + 2 )dT (
+
)ds
x
y
p
s

u
by Section 7 and 8 with X = ( u
x y ) and Y = ( y + x )
By the boundary condition imposed on , the contribution to

u
+
)ds
(
p
s

33
from the part of the boundary of T , common with that of T , is 0. So one can
regard N as the combination of

2u 2u
( 2 + 2 )dT
x
y
taken over T and

u
u
u

)
+(
+
) ]dT + (
+
)ds
[(
x
y x
y
x y
p
s
where the integrations are respectively taken over T and its boundary.
2
2
Clearly, if xu2 + yu2 diers from 0 in some part of T ,N would take a nonzero
2

value provided that , as is permissible, takes the value 0 in T , and ( xu2 + yu2 )
2

has the same sign throughout T . However if xu2 + yu2 = 0 throughout T , then
the contribution to N from T is 0 for every . Now the condition N = 0 yields
the result that the contribution from the discontinuities must be 0.

u
Now the functions X = u
x y , Y = y + x , do not merely, generally
speaking, satisfy
X
Y
+
=0
x
y
but also

x
y
+Y
)ds = 0
p
p

(X

the integral being taken over any part of the boundary of T , at least when this
integral has a denite value.
If T is multiply connected, we reduce T via transverse cuts (by V of Section
9) to a simply connected surface T , and we see that the integral

(
Oo

u
+
)ds
p
s

has the same value for every line from Oo to O in the interior of T . Taking
Oo xed, the integral is a function of x, y that is continuous throughout T and
has the same variation along either side of a transverse cut. Attaching to this
function yields a function v = + for which
v
u v
u
= ,
=
x
y y
x
We obtain the following
Theorem 6 Suppose that a complex function + i of x, y is given throughout
a connected surface T that reduces by transverse cuts to a simply connected
surface T , and

2
2

) +(
+
) ]dT
[(
x
y
y
x

34
taken over the whole surface, is nite. Then there is one and only one choice
of function + i of x, y with the following properties, that can be adjoined to
+ i to give a function of z:
1) vanishes on the boundary, or at least only diers from 0 in isolated
points; is given arbitrary at one point;
2)variations of in T and of in T are discontinuous only at isolated
points, and the discontinuities are restricted by the niteness of the integrals

[( )2 + ( )2 ]dT, [( )2 + ( )2 ]dT
x
y
x
y
over the surface. Further, has the same variations on either side of the transverse cuts.
These conditions suce to determine and . For , from which is determined up to an additive constant, always yields a minimum of the integral ;
since if u = + , then clearly N = 0 for every . By Section 16, this property
is possessed by only one function.
19.
The principles that underlie the result at the end of the previous section
open the way for the investigation of specic functions of a complex variable,
independently of expressions for the functions.
For orientation in this eld, we consider the extent of the conditions needed
to determine such a function in a given domain.
We restrict ourselves initially to a particular case. When the surface spread
over A that represents the domain is simply connected, the function w = u + iv
of z can be determined from the following conditions:
1.the values of u on the boundary are given, and vary, with innitely small
changes of position, by innitely small quantities of the same order. Elsewhere,
the variation of u is unrestricted5 ;
2.the value of is given arbitrary at one point;
3.the function is nite and continuous at all points.
The function is determined completely by these conditions. Indeed, this
follows from the theorem in the previous section; we dene + i, as we may, so
that takes the given value on the boundary and, throughout the surface, +i
varies, with innitely small changes of position, by innitely small quantities of
the same order.
Thus, generally speaking, u can be given as an arbitrary function of s on the
boundary, and thereby v is determined everywhere. Conversely v can be chosen
arbitrary at the boundary points, and the value of u follows. Consequently, the
eld of play for the choice of the value of w on the boundary encompasses a onedimensional manifold for each boundary point, and the complete determination
5 The variations of this value need only be restricted by not being discontinuous along
any part of the boundary. A further restriction is made here simply to avoid unnecessary
complications.

35
of these values requires one equation for each boundary point. All the equations need not necessarily give one specic term a value, at one specic boundary
point. We can also arrange our determining conditions so that at each point of
the boundary there is one equation containing both terms, varying in form continuously with the position of that point. Alternatively, dividing the boundary
into several parts, we associate to each point of one part, n 1 specic points,
one in each of the other parts, and set up n equations for these n points, varying in form continuously as the points vary. However, these conditions, whose
totality forms a continuous manifold, and which are expressed as equations between arbitrary functions, must be subject to a certain restriction in order to be
necessary and sucient to determine a single function continuous in the interior
of the domain. The restriction is given by particular conditions (equations containing arbitrary constants), since the exactitude of our determination clearly
does not extend as far as these.
For the case where the domain of the variable z is represented by a multiply
connected surface, our treatment needs no essential modication. Applying the
theorem of Section 18 gives a function with the above properties, up to the
variation across the transverse cuts. These variations can be made equal to 0,
provided that the boundary conditions contain as many available constants as
there are cuts.
The case where continuity is interrupted along a line can be reduced to the
foregoing by regarding this line as a cut of the surface.
Finally, suppose that we relax continuity at certain isolated points, that is,
by Section 12, allow points where the function becomes innite. Leaving unchanged the other hypotheses of the case initially studied, we can give arbitrary
a function z, whose subtraction makes the function we must determine a continuous function. The original function is completely determined by this. Take the
quantity + i equal to the given function in a circle, as small as we wish, with
center at the point of discontinuity, and elsewhere conforming to the previous
specication. The integral

2
2
((

) +(
+
) )dT
x
y
y
x
over the interior of the circle is 0, and is nite over the remainder of the surface.
One can apply the theorem of the previous section to obtain a function with
the required properties. We deduce with the help of the theorem of Section
13, that in general, if the function becomes innite of order n at an isolated
discontinuity, we have 2n constants at our disposal.
We represent geometrically (by Section 15) a function w of a complex quantity z varying over the interior of a given two-dimensional domain. To a given
surface T , covering A, corresponding an image S, covering B, which after excluding isolated points is similar to T in its smallest parts. The conditions
found previously, necessary and sucient to determine the function, are relative to the values on the contour and at the points of discontinuity. They thus
arise (according to Section 15) as conditions on the position of the boundary of
S, giving a single equation to be satised for each boundary point. If each of

36
these equations concerns just one boundary point the conditions are represented geometrically by a family of curves, one of which gives the location of each
boundary point. When two boundary points, one varying continuously when
the other, appear in a pair of these equations, it follows that between two parts
of the contour there is a dependence which xes one part once the position of the
other part is arbitrary assigned. In a similar way one obtains, for other forms
of these equations, a geometrical interpretation, which we shall not pursue here.
20.
The introduction of complex quantities into mathematics has its origin and
immediate purpose in the theory of simple6 laws of dependence between variables arising from algebraic operations. For if the eld of these laws of dependence is extended by permitting the variable quantities to have complex values,
a formerly hidden harmony and regularity emerges. The cases where this extension has been made are admittedly of limited scope, and reduce almost entirely
to the laws of dependence between two variable quantities of which one is either
an algebraic function of the other7 ; or, a function whose derivative is algebraic.
Nearly all writings on the subject not only gave simpler, more eective form to
results established without the help of complex quantities, but opened the way
to new results, as witnessed by the history of the study of algebraic functions,
circle or exponential functions, elliptic and Abelian functions.
We indicate briey the progress which present study yields for the theory of
the above functions.
Previous methods of treating these functions always based the denition of
the function on an expression that yields its value for each value of the argument.
Our study shows that, because of the general nature of a function of a complex
variable, a part of the determination through a denition of this kind yields
the rest. Indeed, we reduce this part of the determination to that which is
necessary for complete determination of the function. This essentially simplies
the discussion. For example, to show the equality of two expressions for the
same function, one formerly needed to transform one into the other: that is,
show that the expression coincide for every value of the variable. Now it suces
to prove the expressions coincide in a much more restricted domain.
A theory of these functions based on the principles introduced here would x
the form of the function (that is, its value for every value of the argument) independently of a means of determining the function via operations on quantities.
We determine the function by a appending to the general idea of a function of
a complex variable just those features that are necessary for the determination.
Only at this stage do we pass to the various representations that the function
permits. The common character of a class of functions formed in a similar way
by operations on quantities, is then represented in the form of boundary con6 Here we consider the elementary operations to be addition, subtraction, multiplication,
division, integration and dierentiation: and simpler dependence indicates that fewer elementary operations are required. In fact, all functions used up to now in analysis can be dened
via a nite number of these operations.
7 That is, an algebraic equation holds that connects the two quantities.

37
ditions and discontinuity conditions imposed on them. Suppose for example
that the domain of a complex variable extends over the plane in either simple
or multiple fashion, that the function has only isolated points of discontinuity
interior to the domain, and only has innities of nite order. (For innite z,
1
the quantity z itself; for nite z , the order of zz
, is an nite quantity of rst
order.) Then the function is necessarily algebraic, and conversely each algebraic
function fullls these conditions.
We leave aside for the present the development of this theory which, as
observed above, will shed light on the simple laws of dependence arising from
operations on quantities, since we exclude for the present the discussion of expressions for a function.
On the same grounds we do not consider here the utility of our results as the
basis of a general theory of the laws of dependence. For this, one would need to
show that the concept of a function of a complex variable, taken as fundamental
here, coincides fully with that of a dependence8 that can be expressed in terms
of operations on quantities.
21.
In order to clarity our general results, it is useful to give an example of their
application.
The application mentioned in the previous section, while its discussion attains the immediate objective, nevertheless is somewhat special. When dependence is regulated by a nite number of operations of elementary type, the
function contains only a nite number of parameters. As for the form of a system of mutually independent boundary conditions and discontinuity conditions,
sucient to determine the function, this implies that they cannot include arbitrary conditions at each point of a line. For our present objective it is more
appropriate to take an example not of this type, but rather an example where
the function of a complex variable depends on an arbitrary function.
For a clear and convenient framework, we present this example in the geometric form used at the end of Section 19. It then appears as a study of the
possibility of forming a connected image of a given surface, similar to the surface in the smallest parts, whose form is specied. That is, all points of the
boundary should be located on particular curve; moreover, recalling Section 5,
the sense of the boundary, along with the branch points, are given. We restrict
ourselves to the solution of this problem in the case where every point of one
surface corresponds to only one point of the other, and the surfaces are simply
connected. For this case, the solution is contained in the following theorem.
Two given simply connected plane surfaces can always be related in such a
way that each point of one surface corresponds to a point of the order, varying
continuously with that point, with the corresponding smallest parts similar. One
interior point, and one boundary point, can be assigned arbitrary corresponding
8 The dependence expressed here denotes dependence via a nite or innite number of the
four simplest operations, addition, subtraction, multiplication and division. The expression
operations on quantities (by contrast to operations on numbers) indicates operations in
which the rationality of the quantities does not play a role.

38
points; however, this determines the correspondence for all points.
When two surfaces T and R both correspond a third surface S in such
a way that similarity holds between corresponding smallest parts, this yields
a corresponding between T and R that clearly has the same property. The
problem of producing a correspondence between two arbitrary surfaces with
similarity between the smallest corresponding parts now reduces to mapping an
arbitrary surface onto a specic surface, with similarity in the smallest parts.
To prove our result, draw a circle K in the plane B with radius 1 and center
w = 0. We need only show that an arbitrary simply connected surface T
spread over A can be mapped in a connected way throughout the circle K,
with similarity of the smallest parts; and this in a unique way once the center
corresponds to an arbitrary given interior point Oo , while an arbitrary given
point of the circumference corresponds to an arbitrary given boundary point O
of the surface T .
We denote the particular values of z, Q for the points Oo , O by corresponding indices, and describe in T an arbitrary circle with midpoint Oo that does
not reach the boundary of T and contains no branch point. We introduce polar
coordinates: let z z0 = rei , then log(z z0 ) = log r + i. The real part
is continuous in the entire circle except at point Oo , where it becomes innite.
Among the possible values of , we take the smallest positive values, so that
along the radius where z z0 is real and positive, the imaginary part is 0 on
one side, and 2 on the other side, of this radius, while varying continuously
elsewhere. Clearly we can replace this radius by an arbitrary line from the
center to the boundary in such a way that log(z z0 ) has a jump of 2i where
O crosses this line from the negative side (that is, according to Section 8, the
side of negative p) to the positive side; otherwise, log(z z0 ) varies continuously
in the whole circle .
We now take the complex function + i of x, y equal to log(z z0 ) in
the circle, while, outside the circle, having extended in any fashion up to the
boundary T , the function is specied as follows:
1)on the circumference of , the function is log(z z0 ); on the boundary of
T , it is purely imaginary.
2)on crossing from the negative side of to the positive side the function
varies by 2i, while otherwise it varies with an innitely small change of
location by an innitely small quantity of the same order;
This is always possible. The integral

2
2
((

) +(
+
) )dT
x
y
y
x
taken over , has the value 0; and over the remainder of the surface, a nite
value. Consequently, + i can be transformed, by adjoining a continuous
function, dened up to an imaginary constant, that has imaginary values on the
boundary, into a function t = m + ni of z. The real part m of this function
will be 0 on the boundary, at Oo , and vary continuously throughout the
remainder of T . For each value of a of m between and 0, the surface T
is divided by a line, on which m = a, into a part where m < a, containing

39
Oo in its interior, and parts where m > a whose boundary is made up of the
boundary of T and of lines where m = a. Either the connectivity of T , which
is 1, is unchanged by this decomposition, or T is divided into two pieces with
connectivity 0 and 1, or into more than two pieces. However the last case
is impossible, since in at least one of these pieces m is nite, continuous, and
constant on all parts of the boundary. It is then either constant in a portion of
the surface; or has a maximum or minimum value at a single point or along a
line, which coordinates Section 11,III.
Thus the points where m takes a constant value form simple closed lines. One
of these lines bounds a region that surrounds Oo , and m must decrease towards
the interior: consequently, with a positive circuit of the boundary (where by
Section 8, s increase) the quantity n, while continuous, is increasing. Now the
function has a jump9 of 2 only with a crossing from the negative to the
positive side of the line , and so it must take each value between 0 and 2,
excluding multiples of 2, exactly once. Write et = w, then em and n are polar
coordinates of Q with respect to the center of the circle K. The totality of
points Q clearly forms a surface S spread out simply over the whole of K; the
point Qo falls at the center of the circle. However, the point Q can be placed
at an arbitrary given point of the circumference by means of the constant still
at our disposal in n. This completes the proof.
In the case where the point Oo is a branch point of order n 1, we reach
our goal by replacing log(z z0 ) by n1 log(z z0 ); the reasoning is analogous,
and the treatment is easily completed with the help of Section 14.
22.
The complete extension of the investigation in the previous section to the
general case, where a point of one surface corresponds to several points on the
other, and simple connectedness is not assumed for the surfaces, is left aside
here. Above all this is because, from the geometrical point of view, our entire
study would need to be put in a more general form. Our restriction to plane
surface, smooth except for isolated points, is not essential: rather, the problem of
mapping one arbitrary given surface onto another with similarity in the smallest
parts, can be treated in a wholly analogous way. We content ourselves with a
reference to two of Gausss works; that cited in Section 3, and Disquisitiones
generales circa supercies curvas, 13.

9 The line joins one point interior to this piece to another; and so, if it cuts the boundary
several times, it must pass from inside to outside one more time than from outside to inside.
Thus the sum of the jumps of n throughout a positive circuit is always 2.

40
On the number of primes less than a given magnitude.
(Monatsberichte der Berliner Akademie,November 1859.)

I think I can best express my thanks for the honor which the Academy
has conferred in admitting me as a correspondent, by making prompt use of
the privilege now aord me to report on an investigation into the frequency of
prime numbers. This is a subject to which Gauss and Dirichlet have devoted
much eort, and may therefore be considered to be not unworthy for such a
communication.
I take as my starting point in this investigation the remark made by Euler,
that
1
1
1 =
ns
1 ps
where p runs through all prime numbers, and n through all natural numbers. I
shall denote by (s) the function of the complex variable s, which is represented
by each of these two expressions when they converge. Both expressions converge
only if the real part of s exceeds 1; however, it is easy to nd an expression for
the function which is always valid.
The equation

(s 1)
enx xs1 dx =
ns
0
gives immediately

(s 1)(s) =
0

xs1
dx
ex 1

If we now consider the contour integral

(x)s1
dx
ex 1
taken from + to + in a positive sense over a closed path, which includes
in its interior the point 0 but no other point of discontinuity of the integrand,
then it is easily seen to be equal to
s1
x
(esi esi )
dx
ex 1
0
provided that, in the multi-valued function (x)s1 = e(s1)log(x) , the logarithm is determined in such a way that is real when x is a negative real number.
It follows that
s1
x
dx
2sins (s 1)(s) = i
x
e 1
when the integral is interpreted as above.
This equation now gives the value of (s) for every complex s and shows
that it is a single-valued function whose value is nite for every nite s with

41
the exception of 1. It also shows that (s) vanishes when s is an even negative
integer.
If the real part of s is negative, the integral can also be evaluated by being
taken over a path which, instead of surrounding in a positive direction the
domain described earlier, surrounds in a negative direction all other complex
numbers; because the integral is then innitely small for all s of innitely large
modulus. In the interior of the domain, the integrand is discontinuous only
when x is equal to an integral multiple of 2i, and the integral is therefore
equal to the sum of the integral taken in a negative sense around each of these
points. The value of the integral around the point n2i is (2ni)s1 (2i).
Hence

2sins(s 1)(s) = (2)s


ns1 ((i)s1 + (i)s1 )
which gives a relation between (s) and (1 s). This can also be expressed,
by known properties of the function ,as follows:
s
( 1) s/2 (s)
2
remains unchanged when s is replaced by 1 s.
This property of the function led me to introduce the integral (s/2
1 1)
instead of (s 1) as a multiplier of the general term of the series
ns and
thus to obtain a very convenient expression for the function (s). In fact,

s
1
2s
=
(s/2 1)
ennx x 2 1 dx
s
n
0
Therefore,if we set

ennx = (x)

n=1

then

s
s
( 1) 2 (s) =
2

(x)x 2 1 dx
s

Since

1
x(2(x) + 1) = 2( ) + 1
x
(Jacobi,Fund,p.18410 )we have

s3
s
s
1 1 s3
s/2
1
2
2
(s/2 1)
(s) =
(x)x
(1/x)x
(x 2 x 2 1 )
+
+
2
1
0
1
1
s/21
(s+1)/2
+
(x)(x
+x
)
=
s(s 1)
1
I now write s =

1
2

+ ti and
(s/2 1) s/2 (s) = (s)

10 Jacobis

collected works vol.I,p.235. W.

42
so that
(t) =

1
1
(tt + )
2
4

and also

(t) = 4
1

3
1
(x)x 4 cos( t log x)dx
2

d(x 2 (x)) 1
1
x 4 cos( t log x)dx
dx
2
3

This function is nite for all nite values of t and can be expanded
in a very

rapidly convergent series in powers of t2 . Since log (s) = log (1 ps )


remains nite for any value of s whose real part exceeds 1, and since the same s
true of the logarithms of the other factors of (t), it is clear that (t) can vanish
only if the imaginary part of t lies between 21 it and 21 it. The number of roots
of the equation (t) = 0, whose real parts lie between 0 and T , is about
T
T
T
log

2
2 2

because the integral dlog(t) taken in a positive sense around a contour which
includes in its interior all the values of t whose imaginary part lie between 12 it
and 12 it whose real parts lie between 0 and T (ignoring a small fractional term
T
whose order of magnitude is T1 ) has the value (T log 2
T )i; this integral is
however equal to 2i times the number of roots of the equation (t) = 0 in this
region. Now we nd in fact that there are about this number of real roots in this
domain, and it is very likely that all the roots are real. A rigorous proof of this
would certainly be desirable; however after a few brief and fruitless attempts to
nd one, I have put this on one side for the time being, because it did not seem
to be essential to the immediate object of my investigation.
If we denote by an arbitrary root of the equation () = 0, then log (t)
can be expressed by

tt
log (1
) + log (0)

t
Since the density of the roots increase with t only as fast as log ( 2
), this
expression converges and is of order tlog(t) as t tends to innity. The expression
thus diers from log(t) by an amount which is a continuous function of t2 and
which remains nite and continuous for all nite t and tends to zero for innite
t after division by t2 . The dierence is therefore a constant whose value can be
found by setting t = 0.
Let F (x) denote this number when x is not a prime number, but this number
plus 12 when x is a prime number,so that whenever F (x) has a jump in value,

F (x) =

F (x+ ) + F (x )
2

In the series
log (s) =

log (1 ps ) =

ps +

1 2s 1 3s
p
+
p
+
2
3

43
we now replace

xs1 dx, s

xs1 dx,

p2

We obtain
log (s)
=
s

f (x)xs1

where f (x) denotes


1
1
1
1
f (x) = F (x) + F (x 2 ) + F (x 3 ) +
2
3

This equation is valid for every complex value s = a + bi of s, when a > 1.


However,if the equation

g(s) =
h(x)xs dlog x
0

holds in this domain, then the function h can be expressed in term of the function
g using Fouriers theorem. If h(x) is real and
g(a + bi) = g1 (b) + ig2 (b)
the equation splits into the following equations:

g1 (b) =
h(x)xa cos(blog x)dlog x
0

g2 (b) = i

h(x)xa sin(blog x)dlog x

If we now multiply these two equations by


(cos(blog y) + isin(blog y))db
and integrate from to +, the right-hand side of each equation becomes
h(y)y a , by Fouriers theorem. Accordingly, after adding the two equations
and multiplying by iy a ,we obtain

a+i

g(s)y s ds

2ih(y) =
ai

where the integration is carried out so that the real part of s remains constant.
The integral represents,for every value of y at which the function h(y) jumps
in value, the mean of the values on either side of the jump. The function f (x)
dened as above possess this same property, and therefore the equation
f (y) =

1
2i

a+i
ai

log (s) s
y ds
s

44
holds in full generality.
The expression found earlier for log,

(s 12 )2
1
1
log log (s 1) log ( ) +
log (1 +
) + log (0)
2
2

could now be substituted in this equation. However, the integrals of the individual terms of the resulting expression do not converge when the limits of the
integration are nite. It is therefore expedient to begin by transforming the
equation by partial integration into

1
1
f (x) =
2i log x
Since

a+i

ai

d log s(s) s
x ds
ds

s
s
s
log ( ) = lim (
log (1 +
) log n)
m
2
2n
2
1

and therefore

d 1 log (1 +
d 1 log 2s
s
s
=
ds
ds
1

s
2n )

all the terms of the expression for f (x) with the exception of the term
a+i
1
1
1
log (0)xs ds = log (0)
2i log x ai ss
assume the form
1
1

2i log x
Now

a+i

ai

d( 1s log(1 1s )) s
x ds
ds

d( 1s log(1 s ))
d

1
( s)

and,when the real part of s exceeds that of ,


d( 1s log(1 s ))
d

or

1
( s)

x1 dx

=
0

depending on whether the real part of is negative or positive. Accordingly,


a+i d( 1 log(1 s ))
a+i
1
1
1
1
s
1
s

s
x ds =
log (1 )xs ds
2i log x ai
ds
2i log x ai s

x 1
x
dx + const.(in the f irst case)
=
log x

45
and

x1
dx = const.(in the second case)
log x

The constant of integration can be determined in the rst case by allowing


the real part of to become negative innite. In the second case, the value of
the integral from 0 to x takes on two values depending on whether the path of
integration lies above or below the real axes, the values diering by 2i. In the
former case, the integral will become innitely small when the coecient of i in
is innite and positive. In the latter case, the integral is innitely small when
this coecient is innite and negative. This shows how the expression log(1 s )
on the left side is to be determined so that the constant of integration disappears.
By inserting these values in the expression for f (x), we obtain
f (x) = Li(x)

Li(x 2 + ) + Li(x 2 ) +
1

x2

1
dx
+ log (0)
1 xlog x

where the summation


is over all positive roots (or more precisely all complex
roots having a positive real part) of the equation () = 0, arranged in order
of increasing modulus. It can be shown, by a more detailed discussion of the
function , that with this ordering the series
(Li(x 2 + ) + Li(x 2 ))log x
1

converges to be the limit of the integral


1
1
2i log x

a+i

d 1s

ai

log(1 +
ds

(s 12 )2
) s

x ds

when b tends to innity. If the order were to be changed, however, the series
could converge to an arbitrary real value.
The function F (x) can be found from f (x) by inverting the relation
f (x) =
which yields
F (x) =

1
1
F (x n )
n

1
1
(1) f (x m )
m

in which m runs through all the natural numbers not divisible by any square
other than 1, and
denotes the number of prime factors of m.

If, in the sum


, we restrict the summation to a nite number of terms,
then the derivative of the expression for f (x) (neglecting a term which decreases
very rapidly which increasing x) becomes
cos(log x)x 2
1
2
log x
log x

46
which gives an approximate expression for the density of the primes of magnitude
x, plus half the density of squares of primes, plus one-third of the density of
cubes of primes, and so on.
The well-known approximation F (x) = Li(x) is therefore correct only to
1
within an order or magnitude x 2 and gives rather too large a value. For the
non-periodic terms in the expression for F (x) are,excluding those which remain
bounded as x increases without limit,
1
3
1
1
1
1
1
1
1
1
Li(x) Li(x 2 ) + Li(x 2 ) Li(x 5 ) + Li(x 6 ) Li(x 7 ) +
2
3
5
6
7

In fact the comparison between the number of primes less than x and Li(x),
undertaken by Gauss andGoldschmidt and taken up to x = 3, 000, 000, have
revealed that the number of primes is already less than Li(x) after the rst
hundred thousand and that the dierence, with many uctuations, gradually
increases with x. The increases and decreases of density in the prime numbers
due to the periodic terms had already been observed in the counts, but it had
escaped notice that it is regulated by a certain law. If a future count is taken, it
would be interesting to follow up the inuence of the individual periodic terms
in the expression for the density of the primes. The function f (x) should exhibit
a more regular behavior than F (x), and indeed already substantially coincides
on average with Li(x) + log(0) in the rst hundred.

47
On the Hypotheses which lie at the Bases of Geometry.
(Translated by William Kingdon Cliord )
Plan of the Investigation.
It is known that geometry assumes, as things given, both the notion of space
and the rst principles of constructions in space. She gives denitions of them
which are merely nominal, while the true determinations appear in the form of
axioms. The relation of these assumptions remains consequently in darkness;we
neither perceive whether and how far their connection is necessary, nor a priori,
whether it is possible.
From Euclid to Legendre (to name the most famous of modern reforming
geometers) this darkness was cleared up neither by mathematicians nor by such
philosophers as concerned themselves with it. The reason of this is doubtless that the general notion of multiply extended magnitudes (in which spacemagnitudes are included) remained entirely unworked. I have in the rst place,
therefore, set myself the task of constructing the notion of a multiply extended
magnitude out of general notions of magnitude.It will follow from this that a
multiply extended magnitude is capable of dierent measure-relations, and consequently that space is only a particular case of a triply extended magnitude.But
hence ows as a necessary consequence that the propositions of geometry cannot be derived from general notions of magnitude,but that the properties which
distinguish space from other conceivable triply extended magnitudes are only to
be deduced from experience. Thus arises the problem, to discover the simplest
matters of fact from which the measure-relations of space may be determined;
a problem which from the nature of the case is not completely determinate,
since there may be several systems of matters of fact which suce to determine
the measure-relations of spacethe most important system for our present purpose being that which Euclid has laid down as a foundation. These matters of
fact arelike all matters of factnot necessary, but only of empirical certainty; they are hypotheses. We may therefore investigate their probability, which
within the limits of observation is of course very great, and inquire about the
justice of their extension beyond the limits of observation, on the side both of
the innitely great and of the innitely small.
I. Notion of an n-ply extended magnitude.
In proceeding to attempt the solution of the rst of these problems, the development of the notion of a multiply extended magnitude, I think I may the
more claim indulgent criticism in that I am not practised in such undertakings
of a philosophical nature where the diculty lies more in the notions themselves than in the construction; and that besides some very short hints on the
matter given by Privy Councillor Gauss in his second memoir on Biquadratic
Residues, in the G
ottingen Gelehrte Anzeige, and in his Jubilee-book, and some
philosophical researches of Herbart, I could make use of no previous labours.

48
1. Magnitude-notions are only possible where there is an antecedent general
notion which admits of dierent specialisations. According as there exists among
these specialisations a continuous path from one to another or not, they form a
continuous or discrete manifoldness; the individual specialisations are called in
the rst case points, in the second case elements, of the manifoldness. Notions
whose specialisations form a discrete manifoldness are so common that at least
in the cultivated languages any things being given it is always possible to nd a
notion in which they are included. (Hence mathematicians might unhesitatingly
found the theory of discrete magnitudes upon the postulate that certain given
things are to be regarded as equivalent.) On the other hand, so few and far
between are the occasions for forming notions whose specialisations make up
a continuous manifoldness, that the only simple notions whose specialisations
form a multiply extended manifoldness are the positions of perceived objects
and colours. More frequent occasions for the creation and development of these
notions occur rst in the higher mathematic.
Denite portions of a manifoldness, distinguished by a mark or by a boundary, are called Quanta. Their comparison with regard to quantity is accomplished in the case of discrete magnitudes by counting, in the case of continuous
magnitudes by measuring. Measure consists in the superposition of the magnitudes to be compared; it therefore requires a means of using one magnitude as
the standard for another. In the absence of this, two magnitudes can only be
compared when one is a part of the other; in which case also we can only determine the more or less and not the how much. The researches which can in this
case be instituted about them form a general division of the science of magnitude in which magnitudes are regarded not as existing independently of position
and not as expressible in terms of a unit, but as regions in a manifoldness. Such
researches have become a necessity for many parts of mathematics, e.g., for
the treatment of many-valued analytical functions; and the want of them is no
doubt a chief cause why the celebrated theorem of Abel and the achievements
of Lagrange, Pfa, Jacobi for the general theory of dierential equations, have
so long remained unfruitful. Out of this general part of the science of extended
magnitude in which nothing is assumed but what is contained in the notion of
it, it will suce for the present purpose to bring into prominence two points; the
rst of which relates to the construction of the notion of a multiply extended
manifoldness, the second relates to the reduction of determinations of place in a
given manifoldness to determinations of quantity, and will make clear the true
character of an n-fold extent.
2. If in the case of a notion whose specialisations form a continuous manifoldness, one passes from a certain specialisation in a denite way to another,
the specialisations passed over form a simply extended manifoldness, whose true
character is that in it a continuous progress from a point is possible only on two
sides, forwards or backwards. If one now supposes that this manifoldness in
its turn passes over into another entirely dierent, and again in a denite way,
namely so that each point passes over into a denite point of the other, then
all the specialisations so obtained form a doubly extended manifoldness. In a

49
similar manner one obtains a triply extended manifoldness, if one imagines a
doubly extended one passing over in a denite way to another entirely dierent;
and it is easy to see how this construction may be continued. If one regards
the variable object instead of the determinable notion of it, this construction
may be described as a composition of a variability of n + 1 dimensions out of a
variability of n dimensions and a variability of one dimension.
3. I shall show how conversely one may resolve a variability whose region
is given into a variability of one dimension and a variability of fewer dimensions.
To this end let us suppose a variable piece of a manifoldness of one dimension
reckoned from a xed origin, that the values of it may be comparable with one
anotherwhich has for every point of the given manifoldness a denite value,
varying continuously with the point; or, in other words, let us take a continuous function of position within the given manifoldness, which, moreover, is
not constant throughout any part of that manifoldness. Every system of points
where the function has a constant value, forms then a continuous manifoldness
of fewer dimensions than the given one. These manifoldnesses pass over continuously into one another as the function changes; we may therefore assume
that out of one of them the others proceed, and speaking generally this may
occur in such a way that each point passes over into a denite point of the
other; the cases of exception (the study of which is important) may here be left
unconsidered. Hereby the determination of position in the given manifoldness
is reduced to a determination of quantity and to a determination of position in
a manifoldness of less dimensions. It is now easy to show that this manifoldness
has n 1 dimensions when the given manifold is n-ply extended. By repeating
then this operation n times, the determination of position in an n-ply extended manifoldness is reduced to n determinations of quantity, and therefore the
determination of position in a given manifoldness is reduced to a nite number
of determinations of quantity when this is possible. There are manifoldnesses in
which the determination of position requires not a nite number, but either an
endless series or a continuous manifoldness of determinations of quantity. Such
manifoldnesses are, for example, the possible determinations of a function for a
given region, the possible shapes of a solid gure, &c.
II. Measure-relations of which a manifoldness of n dimensions is capable on
the assumption that lines have a length independent of position, and
consequently that every line may be measured by every other.
Having constructed the notion of a manifoldness of n dimensions, and found
that its true character consists in the property that the determination of position in it may be reduced to n determinations of magnitude, we come to the
second of the problems proposed above, viz. the study of the measure-relations
of which such a manifoldness is capable, and of the conditions which suce to
determine them. These measure-relations can only be studied in abstract notions of quantity, and their dependence on one another can only be represented
by formul. On certain assumptions, however, they are decomposable into relations which, taken separately, are capable of geometric representation; and thus

50
it becomes possible to express geometrically the calculated results. In this way,
to come to solid ground, we cannot, it is true, avoid abstract considerations in
our formul, but at least the results of calculation may subsequently be presented in a geometric form. The foundations of these two parts of the question
are established in the celebrated memoir of Gauss, Disqusitiones generales circa
supercies curvas.
1. Measure-determinations require that quantity should be independent of
position, which may happen in various ways. The hypothesis which rst presents
itself, and which I shall here develop, is that according to which the length of
lines is independent of their position, and consequently every line is measurable
by means of every other. Position-xing being reduced to quantity-xings, and
the position of a point in the n-dimensioned manifoldness being consequently
expressed by means of n variables x1 , x2 , x3 , . . . , xn , the determination of a line
comes to the giving of these quantities as functions of one variable. The problem
consists then in establishing a mathematical expression for the length of a line,
and to this end we must consider the quantities x as expressible in terms of certain units. I shall treat this problem only under certain restrictions, and I shall
conne myself in the rst place to lines in which the ratios of the increments dx
of the respective variables vary continuously. We may then conceive these lines
broken up into elements, within which the ratios of the quantities dx may be
regarded as constant; and the problem is then reduced to establishing for each
point a general expression for the linear element ds starting from that point,
an expression which will thus contain the quantities x and the quantities dx. I
shall suppose, secondly, that the length of the linear element, to the rst order,
is unaltered when all the points of this element undergo the same innitesimal
displacement, which implies at the same time that if all the quantities dx are
increased in the same ratio, the linear element will vary also in the same ratio.
On these suppositions, the linear element may be any homogeneous function of
the rst degree of the quantities dx, which is unchanged when we change the
signs of all the dx, and in which the arbitrary constants are continuous functions
of the quantities x. To nd the simplest cases, I shall seek rst an expression for
manifoldnesses of n 1 dimensions which are everywhere equidistant from the
origin of the linear element; that is, I shall seek a continuous function of position
whose values distinguish them from one another. In going outwards from the
origin, this must either increase in all directions or decrease in all directions; I
assume that it increases in all directions, and therefore has a minimum at that
point. If, then, the rst and second dierential coecients of this function are
nite, its rst dierential must vanish, and the second dierential cannot become negative; I assume that it is always positive. This dierential expression,
of the second order remains constant when ds remains constant, and increases
in the duplicate ratio when the dx, and therefore also ds, increase in the same
ratio; it must therefore be ds2 multiplied by a constant, and consequently ds
is the square root of an always positive integral homogeneous function of the
second order of the quantities dx, in which the coecients are continuous functions of the quantities x. For Space, when the position of points is expressed by

51

rectilinear co-ordinates, ds =
(dx)2 ; Space is therefore included in this simplest case. The next case in simplicity includes those manifoldnesses in which
the line-element may be expressed as the fourth root of a quartic dierential
expression. The investigation of this more general kind would require no really dierent principles, but would take considerable time and throw little new
light on the theory of space, especially as the results cannot be geometrically expressed; I restrict myself, therefore, to those manifoldnesses in which the
line element is expressed as the square root of a quadric dierential expression.
Such an expression we can transform into another similar one if we substitute
for the n independent variables functions of n new independent variables. In
this way, however, we cannot transform any expression into any other; since the
expression contains 12 n(n + 1) coecients which are arbitrary functions of the
independent variables; now by the introduction of new variables we can only
satisfy n conditions, and therefore make no more than n of the coecients equal to given quantities. The remaining 12 n(n 1) are then entirely determined
by the nature of the continuum to be represented, and consequently 21 n(n 1)
functions of positions are required for the determination of its measure-relations.
Manifoldnesses in which,
as in the Plane and in Space, the line-element may be
reduced to the form
dx2 , are therefore only a particular case of the manifoldnesses to be here investigated; they require a special name, and therefore
these manifoldnesses in which the square of the line-element may be expressed
as the sum of the squares of complete dierentials I will call at. In order now
to review the true varieties of all the continua which may be represented in the
assumed form, it is necessary to get rid of diculties arising from the mode of
representation, which is accomplished by choosing the variables in accordance
with a certain principle.
2. For this purpose let us imagine that from any given point the system of
shortest limes going out from it is constructed; the position of an arbitrary point
may then be determined by the initial direction of the geodesic in which it lies,
and by its distance measured along that line from the origin. It can therefore
be expressed in terms of the ratios dx0 of the quantities dx in this geodesic,
and of the length s of this line. Let us introduce now instead of the dx0 linear
functions dx of them, such that the initial value of the square of the line-element
shall equal the sum of the squares of these expressions, so that the independent
varaibles are now the length s and the ratios of the quantities dx. Lastly, take
instead of the dx quantities x1 , x2 , x3 , . . . , xn proportional to them, but such
2
that the sum of their squares
= 2s . When we introduce these quantities, the
square of the line-element is dx for innitesimal values of the x, but the term
of next order in it is equal to a homogeneous function of the second order of
the 21 n(n 1) quantities (x1 dx2 x2 dx1 ), (x1 dx3 x3 dx1 ) . . . an innitesimal,
therefore, of the fourth order; so that we obtain a nite quantity on dividing
this by the square of the innitesimal triangle, whose vertices are (0, 0, 0, . . .),
(x1 , x2 , x3 , . . .), (dx1 , dx2 , dx3 , . . .). This quantity retains the same value so long
as the x and the dx are included in the same binary linear form, or so long as the
two geodesics from 0 to x and from 0 to dx remain in the same surface-element;

52
it depends therefore only on place and direction. It is obviously zero when the
manifold
2 represented is at, i.e., when the squared line-element is reducible
to
dx , and may therefore be regarded as the measure of the deviation of
the manifoldness from atness at the given point in the given surface-direction.
Multiplied by 34 it becomes equal to the quantity which Privy Councillor Gauss
has called the total curvature of a surface. For the determination of the measurerelations of a manifoldness capable of representation in the assumed form we
found that 12 n(n 1) place-functions were necessary; if, therefore, the curvature
at each point in 12 n(n 1) surface-directions is given, the measure-relations of
the continuum may be determined from themprovided there be no identical
relations among these values, which in fact, to speak generally, is not the case.
In this way the measure-relations of a manifoldness in which the line-element is
the square root of a quadric dierential may be expressed in a manner wholly
independent of the choice of independent variables. A method entirely similar
may for this purpose be applied also to the manifoldness in which the lineelement has a less simple expression, e.g., the fourth root of a quartic dierential.
In this case the line-element, generally speaking, is no longer reducible to the
form of the square root of a sum of squares, and therefore the deviation from
atness in the squared line-element is an innitesimal of the second order, while
in those manifoldnesses it was of the fourth order. This property of the lastnamed continua may thus be called atness of the smallest parts. The most
important property of these continua for our present purpose, for whose sake
alone they are here investigated, is that the relations of the twofold ones may be
geometrically represented by surfaces, and of the morefold ones may be reduced
to those of the surfaces included in them; which now requires a short further
discussion.
3. In the idea of surfaces, together with the intrinsic measure-relations
in which only the length of lines on the surfaces is considered, there is always
mixed up the position of points lying out of the surface. We may, however,
abstract from external relations if we consider such deformations as leave unaltered the length of linesi.e., if we regard the surface as bent in any way
without stretching, and treat all surfaces so related to each other as equivalent.
Thus, for example, any cylindrical or conical surface counts as equivalent to a
plane, since it may be made out of one by mere bending, in which the intrinsic
measure-relations remain, and all theorems about a planetherefore the whole
of planimetryretain their validity. On the other hand they count as essentially
dierent from the sphere, which cannot be changed into a plane without stretching. According to our previous investigation the intrinsic measure-relations of a
twofold extent in which the line-element may be expressed as the square root of
a quadric dierential, which is the case with surfaces, are characterised by the
total curvature. Now this quantity in the case of surfaces is capable of a visible
interpretation, viz., it is the product of the two curvatures of the surface, or
multiplied by the area of a small geodesic triangle, it is equal to the spherical
excess of the same. The rst denition assumes the proposition that the product of the two radii of curvature is unaltered by mere bending; the second, that

53
in the same place the area of a small triangle is proportional to its spherical
excess. To give an intelligible meaning to the curvature of an n-fold extent at
a given point and in a given surface-direction through it, we must start from
the fact that a geodesic proceeding from a point is entirely determined when its
initial direction is given. According to this we obtain a determinate surface if
we prolong all the geodesics proceeding from the given point and lying initially
in the given surface-direction; this surface has at the given point a denite curvature, which is also the curvature of the n-fold continuum at the given point
in the given surface-direction.
4. Before we make the application to space, some considerations about at
manifoldness in general are necessary; i.e., about those in which the square of
the line-element is expressible as a sum of squares of complete dierentials.
In a at n-fold extent the total curvature is zero at all points in every direction; it is sucient, however (according to the preceding investigation), for the
determination of measure-relations, to know that at each point the curvature is
zero in 12 n(n 1) independent surface directions. Manifoldnesses whose curvature is constantly zero may be treated as a special case of those whose curvature
is constant. The common character of those continua whose curvature is constant may be also expressed thus, that gures may be viewed in them without
stretching. For clearly gures could not be arbitrarily shifted and turned round
in them if the curvature at each point were not the same in all directions. On
the other hand, however, the measure-relations of the manifoldness are entirely
determined by the curvature; they are therefore exactly the same in all directions at one point as at another, and consequently the same constructions can
be made from it: whence it follows that in aggregates with constant curvature
gures may have any arbitrary position given them. The measure-relations of
these manifoldnesses depend only on the value of the curvature, and in relation
to the analytic expression it may be remarked that if this value is denoted by
, the expression for the line-element may be written

1
2
dx2 .
1
1 + 4 x
5. The theory of surfaces of constant curvature will serve for a geometric
illustration. It is easy to see that surface whose curvature is positive may always
be rolled on a sphere whose radius is unity divided by the square root of the
curvature; but to review the entire manifoldness of these surfaces, let one of
them have the form of a sphere and the rest the form of surfaces of revolution
touching it at the equator. The surfaces with greater curvature than this sphere
will then touch the sphere internally, and take a form like the outer portion
(from the axis) of the surface of a ring; they may be rolled upon zones of
spheres having new radii, but will go round more than once. The surfaces
with less positive curvature are obtained from spheres of larger radii, by cutting
out the lune bounded by two great half-circles and bringing the section-lines
together. The surface with curvature zero will be a cylinder standing on the

54
equator; the surfaces with negative curvature will touch the cylinder externally
and be formed like the inner portion (towards the axis) of the surface of a ring.
If we regard these surfaces as locus in quo for surface-regions moving in them,
as Space is locus in quo for bodies, the surface-regions can be moved in all these
surfaces without stretching. The surfaces with positive curvature can always
be so formed that surface-regions may also be moved arbitrarily about upon
them without bending, namely (they may be formed) into sphere-surfaces; but
not those with negative-curvature. Besides this independence of surface-regions
from position there is in surfaces of zero curvature also an independence of
direction from position, which in the former surfaces does not exist.
III. Application to Space.
1. By means of these inquiries into the determination of the measurerelations of an n-fold extent the conditions may be declared which are necessary
and sucient to determine the metric properties of space, if we assume the
independence of line-length from position and expressibility of the line-element
as the square root of a quadric dierential, that is to say, atness in the smallest
parts.
First, they may be expressed thus: that the curvature at each point is zero in three surface-directions; and thence the metric properties of space are
determined if the sum of the angles of a triangle is always equal to two right
angles.
Secondly, if we assume with Euclid not merely an existence of lines independent of position, but of bodies also, it follows that the curvature is everywhere
constant; and then the sum of the angles is determined in all triangles when it
is known in one.
Thirdly, one might, instead of taking the length of lines to be independent of
position and direction, assume also an independence of their length and direction
from position. According to this conception changes or dierences of position
are complex magnitudes expressible in three independent units.
2. In the course of our previous inquiries, we rst distinguished between
the relations of extension or partition and the relations of measure, and found
that with the same extensive properties, dierent measure-relations were conceivable; we then investigated the system of simple size-xings by which the
measure-relations of space are completely determined, and of which all propositions about them are a necessary consequence; it remains to discuss the question
how, in what degree, and to what extent these assumptions are borne out by
experience. In this respect there is a real distinction between mere extensive
relations, and measure-relations; in so far as in the former, where the possible
cases form a discrete manifoldness, the declarations of experience are indeed
not quite certain, but still not inaccurate; while in the latter, where the possible cases form a continuous manifoldness, every determination from experience
remains always inaccurate: be the probability ever so great that it is nearly
exact. This consideration becomes important in the extensions of these empirical determinations beyond the limits of observation to the innitely great and

55
innitely small; since the latter may clearly become more inaccurate beyond the
limits of observation, but not the former.
In the extension of space-construction to the innitely great, we must distinguish between unboundedness and innite extent, the former belongs to the
extent relations, the latter to the measure-relations. That space is an unbounded
three-fold manifoldness, is an assumption which is developed by every conception of the outer world; according to which every instant the region of real
perception is completed and the possible positions of a sought object are constructed, and which by these applications is for ever conrming itself. The
unboundedness of space possesses in this way a greater empirical certainty than
any external experience. But its innite extent by no means follows from this;
on the other hand if we assume independence of bodies from position, and therefore ascribe to space constant curvature, it must necessarily be nite provided
this curvature has ever so small a positive value. If we prolong all the geodesics
starting in a given surface-element, we should obtain an unbounded surface of
constant curvature, i.e., a surface which in a at manifoldness of three dimensions would take the form of a sphere, and consequently be nite.
3. The questions about the innitely great are for the interpretation of
nature useless questions. But this is not the case with the questions about the
innitely small. It is upon the exactness with which we follow phenomena into
the innitely small that our knowledge of their causal relations essentially depends. The progress of recent centuries in the knowledge of mechanics depends
almost entirely on the exactness of the construction which has become possible through the invention of the innitesimal calculus, and through the simple
principles discovered by Archimedes, Galileo, and Newton, and used by modern
physic. But in the natural sciences which are still in want of simple principles
for such constructions, we seek to discover the causal relations by following the
phenomena into great minuteness, so far as the microscope permits. Questions
about the measure-relations of space in the innitely small are not therefore
superuous questions.
If we suppose that bodies exist independently of position, the curvature is
everywhere constant, and it then results from astronomical measurements that
it cannot be dierent from zero; or at any rate its reciprocal must be an area
in comparison with which the range of our telescopes may be neglected. But
if this independence of bodies from position does not exist, we cannot draw
conclusions from metric relations of the great, to those of the innitely small;
in that case the curvature at each point may have an arbitrary value in three
directions, provided that the total curvature of every measurable portion of
space does not dier sensibly from zero. Still more complicated relations may
exist if we no longer suppose the linear element expressible as the square root
of a quadric dierential. Now it seems that the empirical notions on which the
metrical determinations of space are founded, the notion of a solid body and of
a ray of light, cease to be valid for the innitely small. We are therefore quite
at liberty to suppose that the metric relations of space in the innitely small do
not conform to the hypotheses of geometry; and we ought in fact to suppose it,
if we can thereby obtain a simpler explanation of phenomena.

56
The question of the validity of the hypotheses of geometry in the innitely
small is bound up with the question of the ground of the metric relations of
space. In this last question, which we may still regard as belonging to the
doctrine of space, is found the application of the remark made above; that
in a discrete manifoldness, the ground of its metric relations is given in the
notion of it, while in a continuous manifoldness, this ground must come from
outside. Either therefore the reality which underlies space must form a discrete
manifoldness, or we must seek the gound of its metric relations outside it, in
binding forces which act upon it.
The answer to these questions can only be got by starting from the conception of phenomena which has hitherto been justied by experience, and which
Newton assumed as a foundation, and by making in this conception the successive changes required by facts which it cannot explain. Researches starting from
general notions, like the investigation we have just made, can only be useful in
preventing this work from being hampered by too narrow views, and progress
in knowledge of the interdependence of things from being checked by traditional
prejudices.
This leads us into the domain of another science, of physic, into which the
object of this work does not allow us to go to-day.