The origin of Cauchy's conceptions

of a definite integral
and of the continuity of a function.

Many words which are often used have at least two very different
meanings. Thus, the word ( mathematics) is sometimes used to
denote the processes which have been and are now used to discover
entities or truths of a certain kind, and sometimes to denote the
entities or truths discovered by means of these processes. From this
verbal confusion has grown a very harmful confusion which has
shown itself repeatedly in discussions on the principles of mathe-
matics, when some who rightly perceived the great importance of
discovery mistakenly imagined that logic was at fault in ignoring the
processes of discovery - ((intuition ). We speak of (the history of
mathematics), and this, in the fullest sense, would seem to mean the
history of those processes which have been used to attain to knowledge
of those truths which mathematicians,qta mathematicians, seek to
find out. At least, this would presumably be the ideal history of
mathematics: as it is, we can only even approachthis ideal in a very
few cases. Mathematiciansusually consider that the description of
their processes of thought is of little importance, and that the chief
interest of their work lies in the concepts and the truths and proofs
of them that they reach, or think that they reach, or try to reach.
Consequently, the history of mathematicsis usually made up of only
a part of the above description of processes, and the description and
comparison of the variousnotions that men have had of mathematical
truths, mathematicalentities, and mathematicalmethods.
Again, we speak of ((the principles of mathematics), when we are
not speaking of the principles that may have guided mathematicians
in their search for certain truths, but of the logical analysis of these
truths.
Muchthe same is the case with the word ((concept > o'r (concep-
662 PHIIUP E. B. JOURDAIN.

tion ,. In the logical meaning of the word, what we call a (concept )

is the entity itself that may be partially or wholly discovered, but has
itself no origin or development. It did not come into being at a
certain time and plaee and proceed to maturity or death, any more
than a proposition may be true at one time and false at another. A
particular human being may have been the first to recognize its
character,and he would have recognized it at a certain time, and he
or others may have gradually known more about it or forgotten it as
time went on; but this is a totally different question. Because some
reflectivesavage of past ages may have been the first to form an idea
of the number 2 or recognizethe truth of the proposition 2 + 2 = 4 at
some early period of the world's history, we obviously cannot con-
clude that the concept in question or the truth in question first had its
being then. A solipsist may think that his friend only exists in his
perception, and he cannot be decisively refuted as long as he confines
his attention to sense-perceptions; but, if he attempts to extend his
solipsism to the domain of logic and mathematics,he can be refuted.
It may be urged that a proposition may be true now and false, for
example, last year. Thus, (the present year is 1913t may seem to be
the expression of such a proposition. Now, while even such logicians
as BOOLE and MACCOLL spoke of c propositions which are sometimes
true and sometimes false)), modern logicians carefully distinguish
between the constant (propositions), and the ((propositional func-
tions)) which depend on one or more (variables) - representatives
of any member or members of classes of determinedarguments. The
above is the expression of a propositional function; the a variable) is
the time, and is indicated by the words c the present year)). These
words may refer to any year, but, when a phrase marking a certain
date, - like c the year in which I wrote these lines)) or (cthe year in
which the battle of Waterloo took place , - is substituted for the
above words, we get an expression of a proposition, of something true
or false eternally.
In this essay, the words ((process of conception)) will be used to
denote a psychological process whose object it is to discover a logical
entity called a acconcept). Now, as a slight acquaintance with the
history of science teaches us, there is a third thing which it is impor-
tant to consider: it is a partial glimpse of that which is here called the
(<concept)),which has often been mistaken for a complete view of the
concept. Examples of vague and instinctive perceptions of concepts
are affordedby the work of EULER treatedin the present essay. These
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 663

vague perceptions were thought by him and others completely to

denote a concept: thus, EULER'S( continuous function)) was thought to
denote the precise concept which we now call <analytic function)).
Such a partial glimpse, and not that part of the concept which is
glimpsed, will here be called a ((conception)). Thus, conceptions
and the process of conception are of a psychological nature and form
part of the subject-matter of what is called ((the history of mathe-
matics)); while concepts are logical entities and form part of the sub-
ject-matter of what is called ((the principles of mathematics)). We
shall now be able both to use correctlythe current phrases about the
origin and development of conceptions and to discard, to a great
extent, the language of pedantry. If, led astrayby the ambiguities of
ordinary language, we confuse what is here called a cprocess of con-
ception)) with a <(concept)), our confusion is analogous to that
between the tools used for excavationand what is excavated(1); if we
do not distinguish between a (concept)) and a ((conception)), we are
confusing an object with the blurred images of it in the mirrors of
our minds. A coneeption, being psychological, is something belong-
ing to one man, though it is, perhaps, a fairly good copy of that
belonging to another, and so we may speak of < EULER'S conception,
or ( CAUCHY'S conception)). But a is
concept nobody's property. A
conception is often given to the world, and becomes everybody'spro-
a
perty: concept is not even everybody'sproperty.
Conceptsare dealt with by logic, in which all psychological ques-
tions are irrelevant; but deduction is as much at a standstill if con-
ceptions are given it to work with as geometry would be if it had to
work with strings instead of conceptual lines, and knots in them
instead of points. To deal with conceptions seems to me the chief
present function of history. Some are of opinion that history is an
end in itself, and some think that the only goofd in history is its
heuristic value. It seems clear to me that history provides an
enormously valuable - perhaps the only - means of attaining a just
idea of our knowledge by stimulus to criticize. Then, it gives a
stimulus to original discovery; and then, again to criticism.
Since CAUCHY'S time, we have had some definite knowledge of the

(t) Cf. my little book on The nature of mathematics,Edinburghand London,

1913, p. 8-9, 83-84. I there proposedto distinguishbetween Mathematics.,
a class of concepts,and - mathematics,, a psychologicalprccess.
664 PHILIP E. B. JOURDAIN.

concepts denoted by the words ( function ), (? integral ,, and (( con-

tinuous function ). CAUCHY, indeed, only defined the (Cdefinite
integral of a certain class of functions, - a restriction partly
removed by RIEMANN and others, - but it was defined with con-
sciousness of what functions were to be included and what excluded.
The case was different with EULER.He only had a vague feeling of
what ought appropriatelyto be classed as a ( function ), and conse-
quently had only glimpses of certain concepts which seemed to him to
be fitly named ( integral , and ( continuous function ). He was in
the same kind of position as those people who hold that X was a poet
without being exactly sure how ((poet)) should be most appropriately
defined.
The process of search for the complete concept of ( definite inte-
gral o, of which parts only had been seen by the minds' eyes of mathe-
maticians, has been, broadly speaking, a process of sucessive genera-
lization. The conception of the <(continuity ) of a function has had
a different history. The original meaning was even more vague, and
less of it was preservedas time went on. Indeed, it almost seems as
if the name was the only permanentthing about it.
It is well-known (1) that, though LEIBNIZ,in continuity with his-
torical development, conceived the integral calculus as a calculus
summatorius(2), yet, in the formal development of the process of
integration by JOHANN
BERNOULLI
and EULER, that aspect of the integral
as the inverse of a differential became more prominent, and EULER,
indeed, defined the integral.calculus, at the beginning of his Institu-
liones calculiintegralis of 4768, as the method of finding, from a given
relation of differentials, the relation of the quantities themselves.
Indeed, EULERonly used the sum-conception for the approximate
evaluation of integrals.
The sum-conception was reinstated as the fundamental notion of
the integral calculus by CAUCHY. In his Resumedes lefons donndesa
I'ecolepolytechnique sur le calcul infinitesimalof 1823, the existence of
a limit - an arithmeticaltranslation of the geometrical ( area)) - of
certain sums formed with the aid of a continuous function I(x) was
provedto exist (3) and called by the old name of < definite integral .

(') Cf. A. Voss, Differential und Integralrechnungn, Encykl. der math.

Wiss., vol. II, A 2, p. 88-89, 95.
(e) With NEWTON,integration was simply uthe inverse method of fluxions .
(s) See section XVIII, below.
 THK ORIGIN OF CAUCHY'S CONCEPTIONS. 665

It was a function of the limits of the integral and the form of f(x).
CAUCHY'S manner of founding the integral calculus showed, by con-
struction, the existence of the class of functions such as F(x) whicli
admit for derivative a given continuous (1) function f(x). Before
CAUCHY, such integrals were found, and thus shown to exist, for many
f(x)'s; but CAUCHY proved the general proposition.
This sum-conception was already mentioned in CAUCHY'S memoir on
definite integrals of 1814, but no indication was given either there or,
so far as I am aware, elsewhere of the reasons which led CAUCHY to the
abandonment of EULER'Sconception of the integral. This point is a
very important one in the history of the theory of functions, because
of the fact that the sum-definition underlies both the conception of
CAUCHY and GAUSSof an integral between complex limits, and those of
CAUCHY, RIEMANN, DU BOIS-REYMOND, SMITH,DARBOUX, JORDAN, and more
modern authors (2) of an integral of a real non-analytic function
between real limits. This gap between the conceptions of EULERand
those of CAUCHYremained unfilled in my paper on (( The Theory of
Functions with CAUCHY and GAUSS>) (3), although, in the historical
investigations into the theory of functions which I made as preparatory
to a connected account of the development of the theory of transfinite
numbers (4), and which led to the writing of the before mentioned
paper, I have briefly indicated (5) how, in my opinion, the gap must
be filled. In short, it seems to me that FOUlIER'S discovery of the tri-
gonometrical representation of certain discontinuous functions led to
the perception that, by the sum-conception of a definite integral, we
can easily define the definite integrals of such functions, while these
integrals are not differentiable at all points.
In the following paper, I have also attempted a thorough investiga-

(i) This word is to be taken in the modern sense, which was first given by
CAUCHY. See sections XI, XIII, and XIV below.
(I) Cf. H. I EBESGUS;,LeSons sur l'integration et la recherche des fonctions
primitives, Paris, 1904; W. H. YOUNG," On the General Theory of Integration n,
Phil. Trans., A, vol. CCIV, 1905, p. 221-252, and SCHOENILIES, Die Entwich-
elung der Lehre von den Punhtmannigfaltigheiten, part II, Leipzig, 1908,
p. 318-325.
(3) Bibl. Math. (3), vol. VI, 1905, p. 190-207; cf. especially the note on
p. 193.
(4) Cf. Archiv der Math. und Physih (3), vol. X, 1906, p. 254-281, and fol-
lowing volumes.
(6) Ibid., note on p. 261.
X
666 PHILIP E. B. JOURDAIN.

tion of the change in meaning of the term ( continuous function ),.

This change is frequently neglected and has given rise to many mis-
interpretations of expressions of views held by older analysts. This
change was also brought about by the work, especially of FOURIEa, on
trigonometrical developments; but FOURIER himself always used the
term c continuous function ) in the sense which it had before CAUCHY.
The point of view from which the two chief subjects of this paper -
the reinstatement of the sum-conception of an integral and the deve-
lopment of the notion of continuity)) - can best be considered is that
of the profound modification of the conceptions of pure mathematics
which was a result of FOUIlER'Swork. It is well known that FoURIER
had precise notions on the convergence of series and demonstrated
the convergence of the series and integrals named after him in a
manner which, at bottom, is exact, and is, in fact, the way which
was later put into the form of a model of mathematical deduction by
DIRICHLET. In my (<Note on FOURIER'S Influence on the Conceptions
of Mathematics ) (i), I have shown that FOURIERcame very close to
the discovery of non-uniform convergence. FOURIER, as we should
expect from what we know of his views on the relation of mathematics
to physics, never carried out the purely mathematical developmentof
his conceptions. This was done by CAUCHY, and the theories of func-
tions of real and of complex variables and the various theories which
grew out of them and have now become almost autonomous, show
how transcendantly important these conceptions were.
I will now give a more detailed account of the contents of the fol-
lowing essay.
Aftera short account (? I) of the early history of the conception and
word <(function ,), and a notice (? 11) of the early instances of the
integration of partial differential equations, which is of some impor-
tance in connection with EULER'S notion of ? continuity ) of a function
and is not discussed in MORITZ CANTOR'S Geschichte,I proceed (? III) to
give as complete a set of references that I can to the treatment of and
controversies on the problem of vibrating cords with D'ALEMBERT,
EULER, DANIEL
BERNOULLI,and LAGRANGE,
and a rather more detailed
discussion of the distinction that arose chiefly therefrom between

(') Proc. Fifth internat. Congress of Mathematicians (1912), Cambridge,

1913, vol. II, p. 526-527. The subject of the present paper was shortly touched
upon in ? I of this communication.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 667

(( continuous ) and ( discontinuous) functions, as shown in the

works of EULER, LAGRANGE, ARBOGAST (? IV), LACROIx (? V), and Fou-
RIER (5VI). A short summary of the history of the determination of
the coefficientsin trigonometrical series before the time of FOURIER is
in in
given ? VII; ? VIII, FOURIER'S first example of the determination
of coefficients is considered, for it clearly shows the psychological
necessity of extending, by the sum-conception, the conception of a
definite integral even to such functions whose integral has not at
every point a determinatedifferentialquotient; and, in ? IX, FOURIER'S
analytical representation, by integrals, of what ARBOGAST would call
? contiguous but discontinuous functions )) is given. In ?? X, XI,
and XII, CAUCHY'S early memoirs of 1815, 1814, and 1821 are con-
sidered from the point of view of a search for early indications of the
new conception of o continuity )) and the sum-conception of a definite
integral. After a summary (? XIII) of the development of the con-
ception of function from EULER to CAUCHY, with some reflections and
further details, ? XIV deals with the formulation of the conception of
the Ocontinuity )) of a function which was given by BOLZANO in 1817
and is almost identical with that of CAUCHY; and ? XV is on the curious
combination of the old and new conceptions of ( continuity ) pre-
sented in DE MORGAN'S text-book of 4842. CAUCHY'S simple example
constructed for the same purpose as those of FOURIER (? IX) is given
in ? XVI; CAUCHY'S ( singular integrals)) are discussed in ? XVII; and
CAUCHY'S ((existence theorems , are discussed in ? XVIII.

One of the chief characteristicsof modern mathematics is the use

that is made of the notion of variability, which finds an expression in
what is pictured to be the motion of a point along a line, on a plane,
or in space. It appears that the difficulties in the idea of motion
shown to exist by ZENO caused the ancient Greeks to look with disfa-
your on any attempt to introduce the idea of motion into their system
of geometry - at any rate in the rigorous form in which they strove
finally to present it. Modern mathematics is characterized at the
beginning and up to about the middle of the nineteenth century by
almost a disdain of logic, and, to a logician, the rapid growth of the
mathematicalsciences on foundations which were quite unexamined,
but turned out to be, to a great extent, solid, seems almost to point to
668 PHILIP E. B. JOURDAIN.

some very fairly trustworthy instinct for the truth in mathematicians

which gave rise to a faith that has often been justified.
The notion of a ((variable ), to which corresponds a point moving
on an ordinate and generating a curve, was familiar to DESCARTES and
FERMAT, was expressed explicitly by ROBERVAL, and was put to great use
in mechanics by GALILEO and NEWTON.This correspondence of posi-
tions of two points, one moving on the abscissa and the other on the
moving ordinate drawn through the first point, first had to have a
special name when the discovery of the infinitesimal calculus showed
that it was necessary to consider such correspondences more gene-
rally than had hitherto been the case. Thus (1) LEIBNIZ, in 1692, used
the word (function)) to denote lengths- like ordinate,tangent, radius
of curvature, etc., - which stand in a definite relation to the variable
points of a curve. In the more modern sense, the word was used by
JOHANN BERNOULLI from 1698 onwards, and taken over and exhibited
as fundamental in analysis by EULER.At the beginning of EULER'S
lntroductio of 1748, a ((function)) of a variable quantity was defined
as (an analytical expression composed in any way of that variable
quantity and numbers or constant quantities ), and, some years before
this, CLAIRAUT and EULER had used the modern notation of a function,
which consists of a letter placed before the letter representing the
variable.
It must be remembered that the word or other notation used to
denote a conception is, in itself, of little importance. What is
important is the first perception - even though it be not quite clear
- of a concept and of its great importance for our mathematical ana-
lysis. The name and notation are only outward and visible signs of
this; they were invented because it was foreseen that the notion would
play a greatpart in that generality in the conception and expression of
certain truths for which mathematiciansstrive.
It is not difficult, especially at the present time, to show that the
classification which EULER gave of ( algebraic)) and ((transcendental))
functions is open to criticism. But we shall now have to examine the
circumstances which led to the extension of the conception and word
( function )) beyond the < analytical expressions ) of BERNOULLI and
EULER. Thus, EULER was soon led to consider ( arbitraryfunctions)
by the integration of partial differential equations.

was dealt
(i) The history of the conceptionof function before and after EULBR
with by A. PRINGOSEIM and J. MOLK (cf. ? XIII).
 THK ORIGIN OF CAUCHY'S CONCEPTIONS. 669

II

EULER (1) first investigated partial differential equations. In a

memoir of 1734, he succeeded in integrating completely differential
equations with three variablesby making o arbitraryfunctions ) enter
into their integrals. Buthe seems to haveforgottenthis calculusof par-
tial differentials until d'AIL.MBERTused it in mechanics and was the first
to show its importance. D'ALEMBERT, as we shall see, maintained that
these arbitraryfunctions must each be expressed, for its whole course,
by one and the same - algebraicor transcendental- equation. This
property was, at that time, expressed by the phrase: <(the function is
subject to the law of continuity ). On the other hand, EULERmain-
tained that the curves which the arbitrary functions represent need

(i) Additamentumad dissertationemde infinitiscurvis ejusdemgeneris",

Commen,t. Acad. Petrop., vol. VII, 1734 and 1735 (published 1740), p. 184-
200 (this memoirimmediatelyfollowsone - De infinitis curvis ejusdem generis,
seu methodus inveniendi aequationespro infinitis curvis ejusdemgeneris" on
p. 174-183, where p. 180 nearthe end is a mistake for p. 190, and thus the fol-
lowing pages are numberedwronglyby 10). See, among other places, p. 192,
? 19. This fact was mentioned by JACQUES ANTONIEJOSEPHCOUSIN (Astronomie
physique, 1787). Cf. ARBOGAST, Mdmoire of 1791, quoted below. p. 3. This
fact is not noticed by M. CANTOR in his account (Vorlesungen uiber Geschichte der
Mathematik, vol. III, p. 881-882) of the memoir of EULER'Sin question.
That part of the second edition of J. F. MONTUCLA'S, Histoire des Mathdma-
tiques(vol. III, Paris, 1802, p. 342 352) wich dealt with the integrationof partial
differentialequations,was only printedafter Montucla'sdeath, and J*ROMEDE
who then completed the work, had it revised by LACROIX,who added
LALANDE,
a note (p. 344) which is all the more valuableas the early history of partial dif-
ferential equations is not completely dealt with in M. CANTOR'SGeschichte. Cf.
also S. F. LACROIX, Traitd du calcul diffdrentiel et du calcul intdgral, vol. I,
p. xII, 225-249.
We may also mention here that LACRO1X (MONTUCLA,op. cit., p. 344) remar-
ked that NICLAUS BERNOULLI (Acta Eruditorum, 1720; JOHANNBERNOULLI'S
Opera,vol. II, p. 443) investigatedthe relationsbetweenthe partialdifferentials
in an exact equation involving two variables,andhis work implicitlycontains
FONTAINE'S results. This was not mentioned by CANTOR(op. cit., vol. III,
p. 473). On FONTAINE'S and CLAIRIUTSresearches on this subject, cf MON-
TUCLA, op. cit. p. 167-171, 190-193; CANTOR,op. cit., p. 882-889. IIowever,
this is not immediatelyconnectedwith our presentsubject,since the integration
of exact equationsonly dependson that of functionsof one variableand all the
functions which appear in the integrals are determined, arbitraryconstants
alone entering into the final form of the integrals.
 670 PHILIP E. B. JOURDAIN.

not be subject to any law, but may be (( irregular and discontinuous ),

that is to say, either formed by the assemblage of many portions of
different curves, or traced by the arbitrary motion of a hand moving
without a law.
It was first in the discussion of the equation of vibrating cords that
these reasonings appeared.

III

It was D'ALEMBERT with whom began (1), in 1747, a systematic inves-

tigation of the problem of vibrating cords and of the partial difteren-
tial equation to which it leads. D'ALEMBERT was only concerned, at
first, with proving that the problem has an infinity of solutions besides
the one which Brook TAYLORfound in 1713. From EULER'S(1748)
development of, and commentary on, D'ALEMBERT'Smemoir, it would
appear at first sight as though their solutions only differed in points
of secondary importance. But, as BURKHARDT(2) says, D'ALEMBERT and
EULERused, indeed, the same words, but connected different ideas
with the words. The controversy (3) which then arose between

(') On the history of the problem and the controversies to which it gave rise,
cf. M. CANTOR, op. cit., p. 900-906; MONTUCLA, op. cit., p. 659-667; the refe-
rences (to which this section is a supplement) in my paper in the Archiv der
Math. und Phys. (3), vol. X, 1906, p. 255-256; H. LRBESGUE, Lecons sur les
sdries trigonomdtriques, Paris, 1906, p. 19-36; E.W. HoBSON,The theory of func-
tions of a real variable, and the theory of Fourier's series, Cambridge, 1907, p. 635-
641; and G. A GIBSON. On the History of the FOURIERseries n, Proc. Edinb.
Math. Soc. vol. XI, 1892-1893, p. 137-166. But the most detailed history of
this problem and the controversies is given on p. 10-14 of H. BUBKHARDT'S
bulky report: , Entwickelungen nach oscillirenden Functionen und Integration
der Differentialgleichungen der mathematischen Physik., Jahresber. der Deutsch.
Math. Ver., vol. X, Part II, Leipzig, 1901-1908. In this report is brought out
(in the section on p. 47-342) the fact that the theory of trigonometrical series did
not wholly rise from the problem of vibrating cords, but also in part from the
need shown by theoretical astronomy of developments of analytic functions in
such series. Secondly, on p. 1-10 is an account of the treatment, by the BaR-
NOULLIS, EULER,and others, of vibrating systems, before the date of d'ALEM-
BERT'Smemoir of 1747. Thirdly, there are some important corrections and
remarks on the well-known historical sketch in RIRMANN'S Habilitations
schrift ; for them see p. 12, 41.
(2) Op. cit., p. 14.
(3)Ibid., p. 14-18.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 671

and EuLER served the useful purpose of compelling

D'ALEMBEiT the
disputants to state their conceptions precisely: both understood by
the word equationan equation between two analytical expressions,
and neither of them had the least doubt that two such expressions
which coincide for a definite interval of the variable must also
coincide outside this interval. They differed in the use of the
word function; D'ALEMBERT
always imagined a function as an analy-
tical expression, while EULERimagined it as representingan arbitrarily
- graphically - given curve (1). EtlLER believed that it was admis-
sible to apply certain of the operations of the infinitesimal calculus to
arbitrarycurves, but the legitimacy of his reasoning was generally
contested (2). (3) maintained that irregular curves, not
D'ALEMBERT
being expressible by one definite function through their whole course,

(1) Histoire de rAcad. de Berlin, 1748 (published in 1750), p. 80: - Ayant

donc decrit une semblable courbe anguiforme, soit reguli6re, contenue dans une
certaine equation, soit irreguli6re ou m6chanique, son appliquie [ordinate] quel
conque fournira les fonctions dont nous avons besoin pour la solution du pro-
bleme. , EULER(Ibid., p. 84) also remarked: , Ayant ainsi donn6 la solution
g6n6rale, comprenons y encore quelques cas, auxquels la courbe anguiforme est
une courbe connue, dont les parties soient liies en vertu de la loi de continuite,
de maniere que sa nature puisse ltre comprise par une equation. n Then he
gave the solution in a series of sines, referred to below; and he did not say expli-
citly whether the series was finite or not.
We may remark here that, in the first chapter of the second volume of EULEK'S
Introductio in Analysin Infinitorum (Lausanne, 1748), curves are divided into
"continuous curvesn and "discontinuous curves .. Curves are opposed to
-functions *, and, in the first chapter of the first volume, a function is (p. 4)
defined as an analytical expression and no mention is made of a discontinuous
functions.. The division referred to in the second volume is (p. 6):' Ex hac
linearum curvarum idea statim sequitur earum divisio in continuas, et disconti-
nuas seu mixtas. Linea scilicet curva continua ita est comparata, ut ejus na-
tura par unam ipsius x Functionem [that is, analytical expression] deftnitam
exprimatur. Quod si autem linea curva ita sit comparata, ut variae ejus por-
tiones BM, MD, DM, etc., per varias ipsius x Functiones exprimantur: ita ut,
postquam ex una Functione portio BM fuerit definita, tur ex alia Functione
portio MD describatur; hujusmodi lineas curvas discontinuzas seu mixtas et irre-
gutlres appellamus; propterea quod non secundum unam legem constantem for-
mantur, atque ex portionibus variarum curvarum continuarum componuntur".
'() LEBEBGUE,Sdries trigonometriques, p. 20 n.
(3) Histoire de lAcad. de Berlin, 1750, p. 358; d'ALEMBERT'S - Addition au
memoire sur la courbe que forme une corde tendue, mise en vibration " is on
p. 355-360 of this volume. Cf. Opuscules, vol. I, 1761, p. 7; LAGRANGR,
(Euvres, vol. I, p. 68.
672 PHILIP E. B. JOURDAIN.

cannot form the subject of analysis. *cIt seems to me, ) says D'ALEM-
BERT, (( that we cannot express y analyticallyin a more general man-
ner than by supposing it to be a function of t and of s. But, with
this supposition, we only find the solution of the problem for the
cases in which the different figuresof the vibratingcord can be inclu-
ded in one and the same equation. In all the other cases, it seems to
me impossible to give a general form to y ).
We will here pass over many of the details of the researchesand
discussions of DANIEL BERNOULLI, EULER, and D'ALEMBERT ('), remarking
(1753) stated, on very inadequate grounds, that wholly
that BERNOULLI
arbitraryfunctions could be representedby trigonometricalseries. It
is worth while to notice that MACH'S
(2) statement that BERNOULLI
only
used finite periodic series and FounRIER was the first to use infinite
series of this nature is contradictedby RIEMANN'S (3) account. Here
RIEMANN stated that BERNOULLI appealed to the fact that there are an
infinity of constants in the solution by trigonometrical series:
nrx 2.xTr 3rCX
y a sin - + [ sin -- 4 sin + ...

which he gave, and these coefficients can be so determined as to

make y = f(x) any assigned curve. The historical circumstancesare
as follows (i).
had given (5) the above equation as a
In his memoir of 1748, EULER
particular solution; and BERNOULLI (6) observed this, but held, on
in
physical grounds, that the solution was perfectly general. EULER,
memoir, did not admit this gene-
his criticism (1753) of BERNOULLI'S
rality, for that would be equivalentto admitting that everycurvecould
be represented by a trigonometrical series, and this proposition he
considered to be certainly false, seeing that a curve given by a trigo-
nometrical series is periodic - a property not possessed by all
curves. It may be remarkedthat the objection was frequently urged
that an algebraic function could not be representedby a trigonome-

(i) Cf. BURKHARDT,op. Cit., p. 18-24.

(s) Die Principen der Wdrmelehre, 2nd ed., Leipzig, 1901, p. 104, 111.
(3) Partielle Differentialgleichungen, ed. HATTENDORFF,
Braunschweig, 1869,
p. 43-44.
(A)Cf. GIBSON, loc. cit., p. 142.
(s) Histoire de l'Acad. de Berlin, 1748, p. 84-85.
(') Ibid., 1753, p. 147.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 673

trical series, because the series gives a periodic curve while the func-
tion does not. It is possible that there was a difficulty, not only in
believing that, when a function is defined for a given range of values
of the argument, its course outside that range is not determined, but
also that a function whose course within a certain range is not deter-
mined by its course within another interval within that range may be
determined for certain other ranges. It was FOURIER who first recog-
nized and stated both of these apparently paradoxicalfacts. Retur-
ning, now, to EULER; when seeking to establish his position, he
remarked(L)that it might be argued that, since there is an infinite
number of constants, a, P, T,..., at our disposal, it must be possible
to make the proposed curve coincide with any given curve; but he
stated explicitly that BERNOULLI
himself had not used this argument (2).
BERNOULLI, indeed, did not seem, in his memoir of 1753, to have quite
grasped the mathematical consequences of his solution; his results
seemed so satisfactoryin their explanation of the facts of observation
that he was prepared to maintain the generality of his solution on
that ground alone. In a letter addressed to CLAIRAUT
and published
in the Journal des Scavans for March, 1759 (3), he stated very clearly
thesubstance of his memoirs of 1753 and the line of reasoning that had
led him to his treatment of the problem. In criticizing EULER'S
views of his memoirs he (4) explicitly acceptedthe argument from the
infinite number of disposable constants.
The weakness of the argument from the infinity of constants does
not seem, in spite of RIEMANN'S assertion to the contrary in his lec-
tures, to have been brought forward by EULER.
DIRICULET (5), in his second memoir (1837) on the representation of
arbitraryfunctions by trigonometrical series, pointed out in the follo-
wing way that the infinity of coefficients of a power-seriescannot be
arbitrarilydetermined. The coefficientsof a rational whole function
of x of the nth degree can be determined so that the series becomes
equal to an arbitraryfunction f(x) for n - 1 values of x. Further,
the coefficientof the mth term of the series approachesa certain limit

(i) Histoire de l'Acad. de Berlin, 1753, p. 200.

.
(2) .... n'a pas fait cette objection. ,
(3) P. 59-80; cf. BURKHARDT, op. cit., p. 21.
(4) Loc. cit., p. 77.
(b) Repertorium der Physik, vol. I, p. 161; Ostwald's Klassiker, n? 11 ,
p. 16-17.
674 PHILIP E. B. JOURDAIN.

as n increases without limit; but we cannot conclude the false result

that a wholly arbitrary function can be represented by a power-
series,
We now come to the investigations of LAGRANGE
(1) and the discus-
sions on these investigations between D'ALEMBERT,LAGRANGE, and
DANIEL BERNOULLI(2).

(1759) resolved the problem of vibrating cords for a cord

LAGRANGE
loaded with a finite number of masses and then passed to the case of
an infinity of masses; and concluded that, as in the first case the law
of continuity need not subsist, it need not do sQ in the second.
However, some time afterwards,having solved the same problem by a
different analysis, he approached in part to D'ALEMBERT'S
opinion, and,
without admitting that arbitrary functions are subject to the law of
continuity, limited their discontinuity by requiring that the diffe-
b'~y
rential quotients should make a leap nowhere in the curve
which represents the initial figure of the cord (3). And yet LAGRANGE
in 1759 came very near to the discovery of (FouRIER's)formula for the
development of an arbitraryfunction in a trigonometrical series (4).
And it must be noticed that D'ALEMBEIIT, EULER,and LAGRANGE all
agreed in holding that BERNOULLI'Ssolution was not general; but,
while D'ALEMBERT (5), in order to be able to hold that BERNOULLI'S
solution was less general than his own, had to assert that even an
analyticallygiven periodic function cannot always be represented by
a trigonometrical series, LAGRANGE (6) believed that he could prove
this possibility (7). But FoURIER's assertion (1807) that an arbitrary
function could be expressed by a trigonometrical series appearedto
LAGRANGEso impossible that he combated it in the most decided way.
RIEMANN (8) stated that, according to an oral communication from

(i) BURKHARDT, op. cit., p. 27-37 (p. 25-27 containan accountof EuL R's inves-
tigationsof 1746-1748 on the propagationof sound in air).
() Ibid., p. 37-43.
(3) On the views of d'ALEMBERT, EULER, and LAGRANGE, see also DE MORGAN,
The differential and integral calculus, London, 1842, p. 727-729.
(4) BURKHARDT, op. cit., p. 32-33 ; GIBSON, Ioc. cit., p. 143-144.
(5) Opusc. math. vol. I, p. 42 art. XXIV.
(G)Misc. Taur., vol. III., 1766, p. 221, art. XXV.
(7) Cf. RIBMANN, Ges. math. Werke, Leipzig, 1876, p. 218; 1892 edition,
p. 232.
(8) Ges. math. Werke, 1876, p. 219; 1892, p. 233.
 THE ORIGIN OF CAUCHY'8 CONCEPTIONS. 675

DIRICHLET,there is a manuscript on this point in the archives of the

Paris Academyof Sciences.

IV

Out of these discussions grew the investigations as to what discon-

tinuities in the integrals of partial differentialequations were permis-
sible, of CONDORCET. (I), DELORGNA,
ARBOGAST
LAPLACe,DI CALUSO, and
others (2). We must consider the important distinction formulated
by ARBOGAST.
memoir (3), which gained the
ARBOGAST'S
LOUIs-FRANtOIS-ANTOINE
prize offered by the Academy of St. Petersburg in 1787 for the best
answer to the problem: ocSi les fonctions arbitraires,auxquelleson
parvientpar l'int6grationdes equationsa trois ou plusieurs variables,
representent des courbes ou surfaces quelconques, soit algebriques
ou transcendantes, soit mechaniques, discontinues, ou produites par
un mouvementvolontaire de la main; ou si ces fonctions renferment
seulement des courbes continues repr6sentees par une Equation
algebrique ou transcendante (4), )) led to results in conformity with
EULER'S views; but the most important feature in it from our present
point view is the distinction expressed there (5) for the first time
of
between discontinuity and discontiguity. s The law of continuity
consists ), says ARBOGAST, ( in that a quantity cannot pass from one
state to another without passing through all the intermediate states
which are subject to the same law. Algebraic functions are regarded
as continuous because the different values of these functions depend
in the same manner on those of the variable; and, supposing that
the variable increases continually, the function will receive corres-
ponding variations; but it will not pass from one value to another
without also passing through all the intermediate values. Thus the

(1) BURKHAIDT, Op. cit., p. 43-45.

(2) Cf. LACROIX,op. cit., vol. II, p. XIIi., 228-238.
(3) Mdmoire sur la nature des fonctions arbitraires qui entrent dans le inte-
grates des dquations aux diffdrentielles partielles, St Petersburg, 1791. Cf.
M. CANTOR,Op.cit., vol. IV, Leipzig, 1908, p. 880 (article by C. R. WALLNER).
Cf. Ibid., p. 878-882 on the natureof the arbitraryfunction in the integrals
of partialdifferentialequations. Cf. also p. 552-553, 790-791.
(4) Cf. a note on p. 44 of BURKHARDT,
op. cit.
(S) ARBOGAST, op. cit., p. 9-11.
 676 PHILIP E. B. JOURDAIN.

ordinate y of an algebraic curve, when the abscissa x varies, cannot

pass brusquely from one value to another; there cannot be a saltus
(saut) from one ordinate to another which differs from it by an
assignable quantity; but all the successive values of y must be linked
together by one and the same law which makes the extremities of
these ordinates make up a regular and continuous curve.
( This continuity may be destroyed in two manners:
? (1) The function may change its form, that is to say, the law by
which the function depends on the variable may change all at once.
A curve formed by the assemblage of many portions of differentcurves
is of this kind... It is not even necessary that the function y should
be expressed by an equation for a certain interval of the variable; it
may continually change its form, and the line representing it, instead
of being an assemblageof regular curves, may be such that at each of
its points it becomes a different curve; that is to say, it may be
entirely irregular and not follow any law for any interval however
small.
(( Such would be a curve traced at hazardby the free movement of
the hand. These kinds of curves can neither be representedby one nor
by many algebraicor transcendentalequations. ) ARBOGAST called all
such curves ( discontinuous curves ), and similarly for ( disconti-
nuous surfaces)) and functions. Then:
( (2) The law of continuity is again broken when the differentparts
of a curve do not join on to one another (ne tiennentpas les unes auz
autres)... We will call curves of this kind ccdiscontiguous curves,,
because all their parls are not contiguous ), and similarly for ( conti-
guous functions )).
ARBOGAST then decided that the arbitraryfunctions which enter the
solution of partial differential equations of the first (1) and second (2)
orders need neither be continuous nor contiguous, and maintained(3)
against d'ALEMBERT that, in the equation of vibratingcords, a saltus in
the values of b2ylbx2 is permissible - the equation only requires that
in that case by/bt2 should make the same saltus at the same place.
It is of some interest to notice that MONTUCLA, who just mentions (4)
the work of AHBOGAST in this connexion, shows what is apparently the

op. cit., p. 12-53.

(') ARBOGAST,
(2) Ibid., p. 54-94.
() Ibid., p. 77.
(4) Op. ct., vol. III, p. 351.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 677

nomenclature in a passage shortly before the

influence of ARBOGAST'S
one in which he mentions ARBOGAST'Smemoir. OEULER), says MON-
TUCLA(1), pr6tendoit que cela n'etoit pas n6cessaire, et que ces fonc-
tions pouvoient meme ,tre discontinues, au point d'dtrerepresent6es
par les ordonn6es d'une courbe quelconque sans equation, telle que
seroit une courbe trac6e i la main et liberoducto, et meme sans conti-
guit6 dans ses parties, comme une suite de points placEsad libitum. ?

Although the second edition of S. F. LACROIX'S Traitede calcul diffe-

rentiel et de calcul integral (2) is subsequent to the communicated
(3), it is SOmuch under the influence of the ideas of
results of FOURIER
EULER and LAGRANGE that we will consider it shortly before dealing
with FOURIER'S work. In the first place, we have the following defini-
tion of a function. ((Toute quantite dont la valeur depend d'une ou
de plusieurs autres quantites, est dite fonction de ces dernieres, soit
qu'on sache ou qu'on ignore par quelles operations il faut passer
pour remonter de celles-ci a la premiere. (4) This definition has
been noticed by PRINGSHEIM as
(5) being, in its wording, similar to the
famous one due much later by DIRICHLET, which was the principal
sign of the fact that, owing principally to the discoveries of FOURIER,
the notion of functionality was to be thenceforwarddivorced from the
idea of analytically expressible relation. However, as PRINGSHEIM
remarked, the examples show that LACROIX not really anticipate
did
DIRICHLET.
With regard to (continuity ), LACROIX
only once, so far as I can dis-
used the name of ( dis-
cover, although he mentioned (6) ARBOGAST,
contiguity (7). I havecollected in a note the referencesto statements,
which do not seem to contain anything new, about the c law of conti-

(t) Op.cit., vol. III, p. 350.

(') Paris, vol. I, 1810; vol. II, 1814; vol. III, 1819.
(3)These results were not, however, printedfully until 1822; a short account
of them was publishedby POISSON in 1808.
4) Op. cit., vol. I, p. 1-4.
(s) Encykl. der math. Wius., vol. I, A. 1 p. 7.
(6) Op. cit., vol. II, p. 686, xx; vol. m, p. a', 310-311.
(7) Ibid., vol. IH, p. 249.
678 PHILIP E. B. JOURDAIN.

nuity ) in geoinetry and analysis (1). Continuity presupposes, accor-

ding to him, an analytical expression, but arbitraryfunctions in the
integrals need not be subject to the law of continuity.
Another example of LACROIX'S standpoint being that of LAGRANGEin
his earlier works, is afforded by his treatment of convergence (2).
The general idea of convergence is assumed as something already
known and the word a convergence) is applied to the getting smaller
of the terms of a series. We know that FOURIER laid stress on this
important point (3).

VI

We now come to the treatment of ( continuity ) of functions with

FoURIER(4). We shall refer to the English translation, by A. FREE-
MAN(5), of the Theorie analytique de la chaleurof 1822, to the reprint
of the Theorie in the first volume of DARBoux'sedition of the OEuvres
de Fourier, and to the first part of the prize essay of 1811 -- which
is, in essentials the same as much of the Thdorie- printed in 1824
in the fourth volume (for 1819-1820) of the Memoiresde 'lnstitul.
The first mention of what FOURIER understood by a (( discontinous O
function in the Theorieis in article 14: ( ...We may develop in con-
vergent series, or express by definite integrals, functions which are
not subject to a constant law, and which represent the ordinates of
irregular or discontinuous lines. This property throws a new light
on the theory of partial differential equations, and extends the
employment of arbitrary functions by submitting them to the ordi-
nary processes of analysis., (6)
In article 219, FOURIER proves that the result that a function is deve-

(I) Op. cit., vol. I, p. 454; vol. II, p. 161, 673, 685-686, vol. III, p. 249, 307-
311.
(2) Ibid., vol. I, p. 4-13.
(3) We may notice, by the way, that MONTUCLA (op. cit., vol. III, p. 207-209,
221n; cf. p. 258-259) used correct expressions about the theory of convergence,
but did not do so on p. 209-210 (ibid.).
(4) On FOURIER,Sworks and those allied ones of his contemporaries, of.
BURKHARDT,op. cit., p. 409-526.
(5) The analytical theory of heat, Cambridge, 1878.
(6) (Euvres, vol. I., p. 10; FREEMAN'Stranslation, p. 22. This was stated in
the same words in 1811 (Memoires, p. 191-192).
 THE ORIGIN OF CAUCHY'S CONCEPTIONS 679

lopable in a series of sines of multiple arcs, which had been obtained

under the supposition that the function can be developed in a series
of powers, can be extended ((to any functions, even to those which
are discontinous and entirely arbitrary,, (1); and, in the next article,
he remarks that, ( whatever the given curve may be which corres-
ponds to f (x), whether we can assign to it an analytical equation, or
whether it depends on no regular law, it is evident that it always
serves to reduce in any manner whateverthe trigonometric curve; so
that the area of the reducedcurve has, in all possible cases, a definite
value, which is the value of the coefficient of sin x in the development
of the function ))().
In article 228, FOURIER says: ((With regard to those series in which
only the sines and cosines of multiple arcs enter, it is equally easy to
prove that they are convergent,although they represent the ordinates
of discontinuous lines. ) (s)
( It is remarkable,, says FOURIER in article 230, that we can
express by convergent series, and, as we shall see inthe sequel, by
definite integrals, the ordinates of lines and surfaces which are not
subject to a continuous law. We see by this that we must admit into
analysis functions which have equal values whenever the variable
receives any values whateverincluded between two given limits, even
though, on substituting in these two functions, instead of the
variable, a number included in another interval, the results of the
two substitutions are not the same. The functions which enjoy
this property are represented by different lines, which coincide in a
definite portion only of their course, and offer a singular species of
finite osculation. )) (4)

p. 184. This passageoccurs on p. 299

(') Oeuvres,vol. I, p. 207; FREEMAN,
of the Mdrtoires.
(2, Eurres,vol. I, p. 210; FREEMAN, p. 186; Mdm6ires,p. 301-302. Onthe
formationof the coefficientsby integration,DARBOUX ((Envres de Fourier, vol. I,
p. 208-209)remarkedin a note that the determinationof the coefficientsof the
series by definiteintegralswhich keep a meaningeven when the functionis dis-
continuousis due whollyto FOURIER n. It must be rememberedthat DaRBoux,
like most writers after the publication(1821) of CAUCHY's Analyse algdbrique,
usedthe wor4 , discontinousn in a differentsense to that in which FouRImused
it.
(3) (Eures, vol. I, p. 221; FREEMAN,p. 196. In the Memoires,p. 314-315,
there is the same passage.except that - certainn replaces equally easy .
(4) Cuvres, vol. I, p. 224; FREEMAN,p, 199; Mdmoires,p. 318.
680 PHILIP E. B. JOURDAIN.

From this passage we see that FOURIER considered it as true - per-

haps self-evidently so - that two functions which are continuous
in form coincide throughout if they coincide for any connected
portion, however small, of the domain of the variable.
In article 416, FOURIER, speaking of the curve represented by the
arbitrary function f (a), when (( the number p is infinitely great o,
that, ( for one of these points situated at a certain distance from the
point c, the value of f (a) varies infinitely little when we increase the
distance by a quantity less than 2rr/p); and: (( In this [infinitely
small] interval the function f(a) does not vary. . (1) In article 417,
FOURIER emphasizes that f (x) ( is entirely arbitraryand not subject to
a continous law)) (2), and: ((We do not suppose these ordinates to
be subject to a common law; they succeed each other in any manner
whatever, and each of them is given as if it were a single quantity. It
may follow from the very nature of the problem, and from the ana-
lysis which is applicable to it, that the passage from one ordinate to
the following one is effected in a continuous manner. But special
conditions are then concerned, and the general equation, considered
by itself, is independent of these conditions. It is rigorously appli-
cable to discontinuous functions. )) (3)

VII

We will now sum up what there is to say about the determination

of the coefficients of trigonometrical series. EULER(4), in 1754 and
1755, had expressed certain rational functions by infinite series
which proceed according to sines and cosine of integer multiples of
arcs; but it was only later that he (5) actually determined the coeffi-
cients in the same way as that followed by FOURIER.The processes
of EULER,LAGRANGE, for the summation of trigonometric
and FOURIER
series were dealt with by LEBESGUE (6), who has also made various
interesting historical remarks.

(i) (Euvres, vol. I, p. 498-499; FREEMAN, p. 429 (similarly in article 423;

Oeuvres,vol. I, p. 510, FREEMAN, p. 439). This partis not in the Mdmoires.
(2) (Euvres, vol. I, p. 499-500; FREEMAN, p. 430.
(3) (Euvres, vol. I, p. 500; FREEMAN, p. 430.
(4) R. REIFF, Geschichte der unendlichen Reihen, Tibingen, 1889, p 127-
129, 131-132; cf. p. 138.
(5) Ibid, p. 138.140; GIBSON,loc. cit., p. 145.
(6) Sries trigonomdtriques, p. 23-36.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 681

VIII

We will now consider the first example treated by FOURIER

(i) of
determination of coefficients in a trigonometrical series. FOURIER'S
problem was to determine the coefficients in the infinite cosine-series
which is to represent the constant 1, so that
1 =- b cos x + b c 3 b5 cos
cs + ...
He found, by means which are not rigorous but are interesting and
suggestive,
4 1 1
1 ==(cos x---cos 3x+ -cos x- ...) (1),

an equation valid for all values of x between + t /2 and- i /2, and,

by the periodic nature of the terms, a function f (x) is represented by
the series on the right of (1) for all values of x, as follows:

f (x) ==+ 1 for - 7r/2 < x < + rr/2and for all points
x ? n Tr(n 1, 2, ...), where x is one of the above points;
f(x)= forx =(/2)? n n = 0, i, 2, ...);
f (x) = - 1 for n/2 < x < 3rt/2 and for all the (congruent)points;
x + n r(n = 1, 2,...).

Now, in the determination of the coefficients, we have to integrate

such functions as this f (x); and we easily find:

f (x) dx x+ T
T/ (- 712 xc /2);
r

f(x) dx-- x + w/2 (/2 < 3 /2);

and so on. Now it is natural if we start from the sum-conception -

and this is the essential point- to give f (x) dx such an extension
of meaning that points like i/2 come into the range of integration. We

(1) Thdorie, chap. IIi, section II, art. 171-178; (Euvres,vol. I, p. 149-158;
p. 137-144.
FREEMAN, This investigation was contained in FoURIrR's work of
1811; cf. Mdmoires,p. 261-270.
682 PHIIIP E. B. JOURDAIN.

will return to this point. We put, say, expressing the equation

symbolically for the sake of emphasizing the essential point,
S> ftI2 Er !

Jo Jo Jr/2

When speakingof the definite integrals of the functions 4)(x)sin nx

in the coefficients of the trigonometrical development, FOURIER (i)
said: (( Whatever be the function 0(x), or the form of the curve
which it represents, the integral has a definite value which may be
introduced into the formula. The values of these definite integrals
are analogous to that of the whole area IP(x)dx included between
the curve and the axis in a given interval, or to the values of mecha-
nical quantities, such as the ordinates of the centre of gravity of this
area or of any solid whatever. It is evident that all these quantities
have assignable values, whether the figures of the bodies be regular,
or whether we give to them an entirely arbitraryform. )
We now see that a function which is (cdiscontinuous))in the Eule-
rian sense may be contiguous and have no definite differential
quotient at certain points. In the above example, there is no diffe-
rential quotient at the point Tr/2. And yet it seems proper that the
conception of a definite integral should be extended so as to embrace
certain discontiguous functions as integrands. Such an integral is
not, then, the inverse of a differential: the integral, though conti-
guous, cannot be differentiatedat all points.
This also makes clear why CAUCHY, when he saw thatthe old concep-
tion of c continuous function)) was useless, and gave its name to the
conception ((discontiguous function)), wa so carefulnever to say that
continuity was a sufficient condition for differentiability.
Thus, when the sum-conception of an integral was adopted, geome-
trical considerations suggested the extension of the conception to
certain discontiguous functions. For a second example, it seems
indubitable that there is an area included between a curve and its
asymptote; but, in calculating this area, we have to integrate over a
point of discontiguity - a point in the neighbourhoodof which the
function increases beyond all finite limits (2).

(1) Theorie, art. 229; FREEMAN, p. 198; Oeuvtes, vol. I, p. 223; Mdmoires,
p. 316-317.
(2) Cf. LEB,ESGU,Integration, p. 7-8.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 683

IX

The integrals of discontiguous functions representableby FOURIER'S

series may thus represent contiguous functions composed of different
analytical functions in differentintervals; in the above example, f (x)
was
+ x in (- w/2 ... Tr/2)and - x + 7 in (Tr/2...3 n/2)

(1) obtained the

In the ninth chapter of the Theorie, FOURIER
expression 00
2 f .
_-dq sin q. cos qx

for a certain initial state, and remarkedthat it is equal to unity if we

give to x any value included between - 1 and 1, and to zero for all
other values of x. ((We see, by this,), he said, ((that discontinuous
functions also may be expressed by definite integrals. )
Rather farther on (2), FOURIER
obtained for another initial state the
expression
2 dq cosqx

which is equal to e" when x is positive and e-- when x is negative,

so that this function does not change its value when x becomes nega-
tive. (The heat communicatedby the source before the initial state
was formed is propagated equally to the right and the left of the
point 0, which directly receives it: it follows that the line whose
equation is
dq cos qx
Yf
Jo t+ q

is composed of two symmetrical branches which are formed by

repeating to right and left of the axis of y the part of the logarithmic
curve which is on the right of the axis of y, and whose equation is
y = e-x We see here a second example of a discontinuous function
expressed by a definite integral.))

(I) CEuures,vol. I. p. 393; FREBMAN,

p. 339; MJmoires, p. 490-491.
(2) (Euvres, vol. I, p. 396; FREEMAN. p. 340 (FREEMAN refers to RIEMANN,
Part. Diff. Gleich., ? 16, p. 34); MAmoires. p. 492.
684 PHILIP E. B. JOURDAIN.

On this second example DARBOUX (I) gave the following note: ((It is
not here a question of a function which is really discontinuous, but
rather of a function which is expressed by two different laws accor-
ding as the variable is positive or negative.>

The exposition which RIEMANNgave of the development of the theory

of trigonometrical series has become classical and has been rather
uncritically accepted by many writers on history. (2) has
As BURKHARDT
remarked, this is due to the fact that RIEMANN'S
historical exposition
is based on oral communications of This fact results,
DIRICHLET.
according to BURKHARDT, from an indication in a letter of RIEMANN'S;
but there are also such indications in RIEMANN'S printed work. The
must, then, have been those he gained on his visit
views of DIRICHLET
to Paris. FOURIER,but not PoissoN and CAUCHY,were personally
accessible to young foreigners who then visited Paris; and this
explains the fact that POISSON are in the background as
and CAUCHY
comparedwith FOURIER.To this circumstance is due the fact that the
works of POISSON on the physical problems which led to
and CAUCHY
the representationof arbitraryfunctions by means of definite integrals
have been very little considered in comparison with the researches of
FOURIER on the propagationof heat. It seems that justice was done
for the first time to the work of POISSON
and CAUCHYby BURKHARDT.
Such problems as that of vibrating cords led to partial differential
equations which involved the time and one coordinateas independent
variables, and thus their solution only required the development of
functions of one variable in series of a prescribed form. However,
EULER began, in 179 (3), to investigate the equation of the vibrations
of a membrane, which involves three independent variables.
EULER(4) often treated the integration of ordinary differential equa-

(i) Oeuvres de Fourier, vol. I, p. 396.

(2) Op. cit., p. in, note.
(3) On this and on other works of EULER,LAGRANGE, and JAKOBII. BERNOULLI
on such quesfions in the eighteenth century, see BURKHARDT,op. cit., p. 363-
366.
(4) Cf. the summary in his Calc. Int., vol. II, sect. I, cat. X, XI. An account
of the work of EULER, LAPLACE, PRONY, PARSEVAL, AMP*RE, PLANA, POISSON,
FOURIER, and LEGENDREis given by IACROIX,op. cit., vol. III, p. xvII-xix, 529-
574.
 THE ORIGIN OF CAUCHY S CONCEPTIONS. 685

tions by means of definite integrals, and the corresponding integra-

tion of partial differential equations occurs in isolated cases with
d'ALEMBERT(l) (1747) and EULER(2) (1762). But it was LAPLACE(3)
who first obtained general formulae ot this kind in 1779, and his
work was continued by BRIsSON (1804), POISSON (1808) and others (4).
LAPLACE continued his investigations in 1809 (5).
FOURIER'S researches (6) were first communicated to the Paris Aca-
demy in 1807 and 1811, but were not published, with the exception
of a short note on them by POISSON in 1808 (7), until 1822. By the
side of the method of integration of partial differential equations by
definite integrals, developed, at the beginning of the nineteenth cen-
tury, another method in v hich the arbitrary functions were repres-
ented by definite integrals that depend on oscillating functions.
FOURIER,POISSON and CAUCHY took part in this latter development; but
it cannot now be ascertained what is due to each and in what way the
method may have developed, say, out of LAPLACE'S work just spoken
of (8). POISSON and CAUCHY applied their methods to almost all the
domains of physics which were then susceptible of mathematical
treatment, while FOURIER had only opportunity, owing to the circum-
stances of his life, to treat thoroughly the theory of the conduction of
heat (9). BURKHARDT, after discussing the integrals treated in the last
section of FOURIER'sprize memoir of 1811 (10), dealt with CAUCIIY'S
memoir of 1815 on waves (it), and POIssoN's memoir of 1816 on
waves (12).

(1) BURKHARDT, op. cit., p. 12, equation (10).

(2) Ibid., p. 351, equation (22).
(3) Ibid., p. 398-401; cf. LACROIx, op. cit., art. 1129.
(4) BURKHARDT, op. cit., p. 401-407.
(5) Ibid.. p. 407-408.
(6, Ibid., p. 409-423.
(7) Bull. de la Soc. philomath. de Paris, 1808, p. 112-116; (Euvres de
Fourier, vol. II, p. 215-221.
(8) BURKHARDT, op. cit., p. 423-424. On the differences between their phy-
sical views, see ibid., p. 424-426.
() Ibid,, p. 426.
(40) Ibid., p. 426-428. The final sections of FOURIER'sThdorie of 1822 are
dealt with on p. 463-469.
(1i) Ibid., p. 429-438.
(12) Ibid., p. 439-447, SoPHIEGERMAIN'S, POISSON'Sand NAVIER'S researches
on the theory of elastic surfaces, which date from 1811 and later, are dealt with
on ibid., p. 447-454; and the discussion between FOUvrER,POISSON and CAUCHT
on waves and on vibration of plates on ibid., p. 454-463.
686 PHILIP E. B. JOURDAIN.

The problem of the propagationof the waves at the surface of a

fluid was treated by LAPLACE (i) in 1778, and LAGRANGE (2)treated the
propagation of waves at the surface of water a
of very small depth in
1781. The question of the propagationof waves on waterunder more
general conditions was made by the Paris Academy the object of a
prize-questionfor 1816, and CAUCHY () carriedoff the prize. He began
with the establishment of the general hydrodynamicalequations, and
arrived at the equation (4)
ba qo __ b2 qo
O
bas b b2

which he integrated in the form :

ro bm - bm
qo-==I cos(am) .e f(m) .dmEj- cos[am).e .f(m).dm.

but did not give the slightest indication of how he arrivedat this form
of solution. In proof of the generality of the solution, he showed
that the development according to powers of b transforms it into the
form that is directly given by the method of integration by a power
series.
We must remember that the account which POIssoN gave in 1808 of
FOURIER's work was the only information about this work which
CAUCHY had. FOURIER'S prize memoir of 1811 was unknown to CAUCHY,
as we shall shortly see, till about 1818. Further, both in this memoir
and CAUCHY'S memoir on definite integrals of 1814, it must be noticed
- and this is why it is important for the historian to study CAUCHY'S
memoir in the Memoires,rather than in the OEuvres- that, neither
in the memoir on waves, LEGENDRE'S report on the memoir of 1814,
nor in what CAUCHY wrote in 1814, is use made either of round b's to
denote partial differential quotients or of FOURIER'S notation for
definite integrals. FOURIER'Snotation was first used in the memoir of
1814, in a note on p 623, added about 1825 to his memoir, with the

(i) UEuvres,vol. IX, p. 301.

(2) (Euvres, vol. IV, p. 746.
(3)
" Th6orie de la propagation des ondes a la surface d'un fluide pesant d'une
profondeur ind6finie ,, Mdmoires prdsentes par divers savants a I'Acad. roy. des
Sc. de l'Inst. de France (Sc. math. etphys.), vol. I, 1827, p. 3-312; CEuvres(1),
vol, I, p. 5-318.
(4) Section 3 and note IX.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 687

prefatory words: ( Si l'on d6signe avec M. FOURIBR.... . In the

memoir on waves, FOURIER'S notation was first used in the sixteenth
note (1)which was added shortly before publication.
In the nineteenth note, which CAUCHY also subsequently added to
this memoir, he noticed the (( reciprocal) properties of, for example,
the functions f(x) and cp(x) in the formulae

f (x) ) (p() cos p x. d

and
2
vp (x)= (-) f (A) cos x .d

and continued () : ( The remarkable properties of these functions

and the advantagesofferedby them in the solution of a great number
of problems furnished me with the subject of a note in the Bulletin de
la Societdphilomathiqueof August, 1817. It is essential to remark
that, when I drew up this note, I did not yet know any other memoirs
where the formulae from which the reciprocity is deduced or used
than those of M. POISSON and myself on the theory of waves. Since
this time, M. FOURIER communicatedto me his researches on heat pre-
sented to the Institute in 1807 and 1811 and only published in 1819.
1 there saw the same formulae and I hastenedto render to him in this
respect the justice due to him in a second note printed under the date
of December, 1818. )
We will now turn to CAUCHY'S early memoir of 1814 on definite
integrals.
XI

We will here examine CAUCHY'S memoir of 1814 on definite inte-

grals (3) for the conception which ARBOGAST called (? discontiguity
and which we now, following CAWCHY, call (( discontinuity ),

(') Mem. de l'Inst. vol. I, 1827, p. 194.

(9) M4noires, vol. I. p. 293-294: (Euvres (1), vol. I, p. 300-301.
(3) M6moire sur les integrales definies ,. This memoir was read to the Ins-
titute of France on August 22"d, 1814, and printed, together with some foot-
notes added afterwards and LEGENDRE'S report of November, 1814, in the
Memoires prdsent.s par divers savants a I'Academie royale des sciences de 'In-
stitut de France (sciences mathdmathiques et physiques), vol. I, 1827, p. 611-799
(LEGENDRE'S report, p. 601-610): (Euvres (1), vol. I, p. 329-506 (LKENDRB'S
report, p. 321-327).
688 PHILIP E. B. JOURDAIN.

First of all, it must be noticed that LEGENDRE(1), in his report,

remarked that some of the integrals which were first evaluated by
CAUCHY o present cases where the law of continuity is violated. One
such formula,
x cos aw dx
J sin bx 'i-
1 .

increases or diminishes suddenly by Tr/2when the ratio a/b, which is

at first supposed equal to a whole number, diminishes or increases by
an infinitely small quantity ).
In the introduction to CAUCHY'S
memoir, it is pointed out (2) that the
theorem that, if an indefinite integral is expressed by a certain function
of the variable augmented by an arbitraryconstant, the same integral,
taken between two limits a and b, will be expressed by the differences
of the values of the function for these limits, (cis only true in the case
of the function found increasing or decreasingin a continuous manner
between the limits in question. If, when the variable increases by
insensible degrees, the function found passes suddenly from one value
to another, the.variablebeing always comprised between the limits of
integration, the differences between each of the abrupt leaps (sauts
brusques)that the function may make will necessitate a correction of
the same nature. We easily obtain this rule by considering the pro-
posed integral as the sum of the elements which correspond to the
various values of the variable and by dividing the total sum into as
many partial sums as there are abrupt leaps in the function found,
plus one. ))
In the body of the memoir, the third section (3)of the second part
concerns the problem of finding the value of

J dI (z) dz when q' (z) dz is given as qp(z) + C.

( If), says CAUCHY(4), ( the function qp(z) increases or decreases in a

continuous manner between the limits z = a and z = b, the value of
the integral will be, as usual, represented by p (b)- q (a). But if,

(i) (Eures de Cauchy(1), vol. I, p. 326.

(2) Mefmoires, p. 614-615; (Euvres, p. 332.
(3) a Sur la conversion desintegrales ind6finies en int6grales d6finiesn, ibid.,
p. 402-406 (Mgmoires, p. 686 sqq.)
(4) Ibid., p. 402-404 (Mdmoires, p. 687-689).
 THE ORIGIN OF CAUCHYS CONCEPTIONS. 689

for a certain value of z represented by Z. and comprised between the

limits of integration, the function qp (z) passes suddenly from one
determinate value to another value sensibly differentfrom the first, so
that, if Zdenotes a very small quantity, we have
( (Z + Z) - (Z- t) -A,

then the usual value qp(b) - qp(a) of the definite integral must be
diminished by the quantity A. In fact, we may divide the definite
integral $ q' (z) dz into two others of the same form, of which one
is taken between the limits z = a, z == Z - and the other between
the limits z = Z + Z, z = b, provided that, in the sum of these last
two integrals, we suppose that L = 0. Evidently these integrals are
equal to cp (Z - t)- p (a) and q( (b) - qp(Z + Z), respectively.
Their sum is, then qp(b)- p (a) - A. If the function (p(z) changes
its value suddenly many times between the limits a and b, then, if we
denote by A, A', A", ... the sudden variations in question, we would
have qp (b) - q (a) -A - A' - A" - ... for the value of the
definite integral sought. )
(i) considered the cases:
As examples, CAUCHY
4
dz/z = log (4) - log ( ) - A,
,2

where A = - log (- 1) (2); and

3sr 3ir
4 sin z 4
Si sinz ==| arc tan ( ) A,
o + (cos z)2 -Cos Z It
when the sign of substitution was not, of course, used by Cauchy.
Here arc tan (cy) is the smallest of the positive or negative arcs
whose tangent is 1/cos z; and it is Tr/4 for z = 0, increases
(( continuously)) from z = 0 to z = Tr/2- Z, where Z is a very small
quantity, suddenly passes from nT/2,which is its value for z -= i/2 - Z,

(1) CAUCHY, Ibid. p. 404-406 (Mdmoires, p. 689, 691).

(C) CAUCHY(ibid., p. 402-403; Mdmoires, p. 686-687) added a note to the effect
that the values here considered are what he later called - principal values " and
that in the more general case, the value of A in this example is log (h), k being
an arbitrary constant.
690 PHILIP E. B. JOURDAIN.

to - n/2, which is its value for z ==7/i + , and thence it is negative

and decreasingup to z = 3n/4, where it is - are tang /EL In this case
Ai = p (lt/2 ? t) - p (1T/2- t) - tt/2 - 7r/2,

and the integral is 3n/4 - arc tan /2, which is positive, whereas
the wrong value given by the ordinaryintegrationis negativealthough
a bum of positive elements.
The hypothesis of a function being o continuous ) in the modern
sense is made in theorems in the later part of this memoir (I).

XII

In October, 1821, CAUCHY presented to the Academy of sciences a

memoir on the integration of linear partial differentialequations with
constant coefficients; it was very much developed, some additionswere
indicated in the Bulletinde la Societt philomathiquefor 1821 and the
((Analyse des travauxde l'Acad6mie des sciences) during 1821, was
presented to the Academyof sciences in September,1822, and printed
in the Journalde I'Ecolepolytechniquein 1823 (2). This memoir con-
tained a simplified solution ((deduced from a formula which was first
used by M. FOURIER in the memoir on heat and afterwards applied to
other problems - in particularby M.POISSON and myself to the theory
of waves ,, (3).
The <(GeneralObservations and Additions)) (4) begin: (( In the
above memoir, I (5) consider each definite integral as being just the
sum of the indefinitely small values of the differentialexpression pla-
ced under the 5, which correspondto the various values of the variable
included between the limits in question. When we adopt this way of
regarding definite integrals, we easily prove that such an integral-has

(1) CAUCHY, Ibid., p. 428, 441 (Me6noires, p. 714, 729).

(2) Cah. XIX, vol. XII, p. 511; Oeuvres (2), vol. I, p. 275-357. This memoir
was reviewed by BURKHARDT (op. cit., p. 671-680); cf. also the following pages
for other memoirs of CAUCHY in the section on the influence exerted by the theory
of functions of a complex argument on this subject.
(3) a(Eures, (2), vol. I, p. 276.
(4' (Euvres (2), vol. I, p. 333-324 (Journ. de l'Ec. polyt., cah. XIX, p. 571-
590).
(5), Nous is here always translatedby - I ,.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 691

a unique and finite value whenever, the two limits of the variable
being finite, the integrand remains finite and continuous between
these limits. ) Then (1), after the fundamental properties of definite
integrals are shortly given, there is an account of the ( singular inte-
grals)) describedin the memoir of 1814 and in the Bulletin de laSociete
philomathiqueof 1822. Then (2) the application of these considera-
tions to the evaluation of definite integrals, to the resolution of alge-
braic and transcendental equations, and to the integration of ordi-
nary differential equations are given.
In a postscript (3) to this memoir CAUCiHY added: ( We are natu-
rally led by the theory of quadraturesto consider each definite inte-
gral which is taken between two real limits as being the sum of the
infinitely small values of the differential expression under the sign i
which correspond to the various real values of the variable which are
included between the limits in question. Now, it seems to me that
this manner of regarding a definite integral ought to be adopted in
preference, as I have just done, because it is equally suitable to all
cases, even to those in which we cannot pass generally from the func-
tion under the sign S to the primitive function. Besides, it has the
advantageof giving always real values for the integrals which corres-
pond to real functions. Finally, it allows us easily to separateeach
imaginary equation into two real equations. All that would no lon-
ger be so if we considered a definite integral taken between two real
limits as necessarily equivalentto the differenceof the extreme values
of a discontinuous primitive function, or if we made the variable
pass from one limit to another by a series of imaginaryvalues. ) In
the two last cases we often obtain imaginary values like that which
POISSON (4) has given. Then also it may happen that one and the same
integral correspond many primitive functions, some of which lead
to real values of the integral and others to imaginaryvalues (5).

(i) Ibid., p. 334-336.

(') Ibid., p. 336-354.
(3) (Euvres (2), vol. I, p. 354-357 (Journ. de l'Ec. polyt., cah. XIX, p. 590).
Cf. Bull. Soc.phil., 1882, p. 171.
(4)Journal de I'Ec.polyt., cah. XVIII, p. 329.
op. cit., p. 355-357.
(5) CAUCHY,
692 PHILIP E. B. JOURDAIN.

XIII

We will now sketch broadly the development of the conception of

a function from the time of EULER(1). EULER,at the beginning of
the first volume of his Introductioof 1748 (2), defined a ((function))
as ((an analytical expression composed of a variable and constants))
- a definition that he took over from JOHANNBERNOULLI.FOURIER'S
work brought out the tacit suppositions underlying this definition,
which made it seem possible to assimilate functions in general to the
easily managed algebraic functions in virtue of a ((principle of conti-
nuity ); which referred to the analytical expression of a function.
Since, now, some - if not all - quite arbitrary((functions))--notice
that this is an extension, due to EULER,of the use of the word - could
be represented by a single analytical expression, there appeared to be
no longer any reason for allying the definition of a function to the
pre-existence of a unitary analytical expression, for such an expres-
sion appeared always to be discoverablea posteriori. Thus, under
worked with the idea that y is a one-
FOURIER'Sinfluence, DIRICHLET
valued ((function))of x in the interval a _ x< b if a definite value
y belongs to each such x, irrespective of how the correspondence is
established. But then, too, besides this, a narrower conception of
(analytic) function was advisable, because very many important func-
tions have the properties which EULER vaguely felt as a continuity )).
But into this we cannot enter here.
conception of a function, it is interest-
On the subject of DIRICHLET'S
(3) remarked, in his well-known
ing to note that HANKEL ( Unter-
suchungen fiber die unendlich oft oscillirenden und unstetigen Func-
tionen)) of 1870, that the definition: a y is said to be a function of the
variablex if to every value of x within a certain intervala definite value
of y corresponds; whether or no y depends on x accordingto the same
law in the whole interval, and whether or no the dependence can be

(i) Cf. A. PRINGSHEIM, Encykl. der math. Wiss., vol. II, A. 1, p. 3-8, and
the much fuller exposition by J. MOLKof this article in Encycl. des Sci. math.,
vol., II, 1, p. 1-16.
(2) The above-cited (? III) passage on - discontinuous curves ? from the second
volume of the Introductio was only referred to by PRINGSHEIM (loc. cit., p. 6;
French ed., p. 11) in a note.
(3)See Ostwald's Klassiker, 11?153, p. 49.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 693

expressed by means of mathematical operations ), lies at the founda-

tion of DIRICHLET'S memoirs on FOURIER'S series. This assertion seems
to have been construed by many into an assertion that DIRICHLET,at
the place (1)quoted by HANKEL, actually defined (a function )) in this
way. However, DIRICHLE1 only explicitly defined a continuous func-
tion of x, and the very general conception of a function that he had is
to be sought rather among the ideas which were expressed in his first
memoir of 1829 on trigonometrical series (2).
in connexion with the general
(3) did not mention DIRICHLET
LEBESGUE
conception of function, and attributedto RIEMANN the definition usually
known by DIRICHLET'S name. Presumablyhe meant the general defini-
tion of a function at the beginning of RIEMANN'S dissertation of 185 :
((Grundlagen fur eine allgemeine Theorie der Functionen einer
veranderlichen complexen Grosse,, and it is possible that this was
the first explicit statement of what was not explicitly stated, though
obviously implied, in DIRICHLET'S work.
Attempts to distinguish between algebraic and transcendent func-
tions by charactersof their analytical expression fail (4) and the signi-
ficance of the rejection by mathematicians at the beginning of the
nineteenth century of the conception of ? continuity of form )) may be
stated as being a consequence of the perception which forced itself
upon them that a function was not, in essentials, connected with its
expression. From this point of view, there seems to be some analogy
between this rejection and RIEMANN'S method of defining certain func-
tions by conditions with respect to their boundaries.
Thus, the existence of a single analytical expression which repre-
sented a function did not guarantee the possesion by this function of
common propertiesthroughout its course. This existence of common
propertieswas really that which was thought to be possessed by what
d'ALEMBERT called c<functions)) and EULERcalled ( continuous func-
tions ). work showed that analytical expressions
FOURIER'S were
capable of representing dependences on the variable x which seemed

(1)Repertoriumder Physik, vol. I, 1837, p. 152; Werke, vol. I, p. 135-136;

Ostwald's Klassiker, n? 116, p. 3-4.
(2) , Sur la convergencedes series trigonometriquesqui serventA representer
une fonction arbitraireentre des limites donn6es,n Journ. fur Math., vol. IV,
1829, p. 157-169; Werke, vol. I, p. 119-132.
22n.
(3) Intdgration, p. 4; Series trigonometriques, p.
(4) Encykl., vol. II, A. 1, p. 4-5; Encycl., vol. II, 1, p. 7-10.
694 PHILIP E. B. JOURDAIN..

wholly arbitrary. That they cannot be strictly described as wholly

arbitraryis simply a consequence of the fact that, since FOURIER lived,
mathematicians have become more and more logical, and have, to a
very great extent, become careful not to let what is called (( geome-
trical intuition ) get the upper hand of logical exactness when they
are stating general propositions.
The old distinction between o continuous ) and ( discontinuous)
functions could, then, no longer be maintained in the old form. That
there was a valuable distinction struggling for expression and failing
in this struggle, in the writings of D'ALEMBERT,
EULER,and LAGRANGE,
we shall see at the end of the next section but one. Here we must
rememberthat FOURIER, though he showed almost conclusively that the
old conception of (( continuity ), was, strictly speaking, valueless,
yet made use of this conception in his writings.
This seems to be due, at least in part, to his lack of interest in
pure mathematics; and so the enormously important purely mathe-
matical developments which lay implicit in FOURIER'S work were first
carried out by CAUCHY.
As BRILL (1) says: ( But, since people became conscious of the
fact that every line - even if it is a broken line - formed by the
ends of the ordinates to the axis of the independent variable can be
represented by means of a trigonometrical series, they could no
longer speak of laws of dependence which differed in form although
belonging to the same function in EULER'S sense. For this reason
CAUCHY replaced EULER'S definition by a new one. )) In fact, Eulerian
(ccontinuity )) was merely concerned with the verbal or symbolic
description of certain functions, and not with their nature. It was
FOURIER'Sgreat service to have shown this clearly, not, perhaps, to
himself, but at least to CAUCHY.
In CAUCHY'S long memoir of 1844 on definite integrals, which
contains the germs of so many of CAUCHY'S later discoveries, we
find (t) the new conception of the ( continuity )) - ARBOGAST'S (( con-
tiguity ) - of a function formulated, though not with precision.
The precise formulation was given in the Cours d'analyse of

(t) In A. BRILLand M. NORTHER,

Die Entwicklungder Theorie der alge-
braischenFunktiouenin iltererund neuererZeitn, Jahresber.der Deutsch.
Math.-Ver.,vol. II, 1894,p. 162.
(2) (Eutes (1) vol. I, p. 406.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 695

1821 (1). In this book CAUCHY did not mention FOURIER'S name.
Indeed, the apparently studied omission of FOURIER'S name throug-
hout CAUCHY'S work is rather remarkable.
We must remember that what is important in the revolution in
ideas brought about by CAUCHY is not the new definition, which did
appear, but the concept defined, which is not, like the phrase or sign
which translates it, a matter of arbitrariness. In CAUCHY'S Analyse
algebrique, which was published at Paris in 1821 as the first part of a
Coursd'analyse,we read: (The function f (x), which is one-valued
at every point between certain limits, remains continuous with
respect to x between given limits if, between these limits, an infinitely
small increment of the variable always produces an infinitely small
increment of the function itself. ) Analogous definitions were
given of the (( continuity ) of functions of many real and of an
imaginary variable, and in Note III. was given a purely analytical
proof of the theorem which, as we shall see in the next section,
occupied BOLZANO and led him to practically the same conception of
(( continuity ) and the same method of proof, in a paper published
four years before the Analyse algebriqueappeared.

XIV

A conception of the o continuity ) of a function, which is almost

identical with that of CAUCHYwhen real functions only are considered,
was given, before CAUCHY, by BOLZANO (2) in 1817. BOLANO was
concerned with the purely analytical proof of the theorem that, if a
rational and whole function takes, for two values of the (real) argu-
ment, values of different sign, there is at least one real root between
these two values of the argument. For this purpose, the functionf(x)
must be of a certain nature; and BOLZANO dealt with the question in
rather a peculiar way. First of all, he remarked that the proof that
used to be given and that was based on the ( continuity ) of a
function, with an intermixtureof the conceptions of time and motion,

(1) Cours d'analyse de l'lcole royale polytechnique; ire partie: Analyse alge-
brique, Paris, chap. II, ? 2; (Euvres(2), vol. III, p. 43-51.
(2) , Rein analytischerBeweis des Lehrsatzes,dass zwischenje zwei Werthen,
die ein entgegengesetzesResultat gewlhren, wenigstens eine reelle Wurzel der
Gleichungliege ", Abhandlungender Kgl. Ges. der Wiss. zu Prag., 1817; fac-
simile reprint, Berlin, 1894; annotatededition in no 153 of Ostwald'sKtassiker.
696 PHILIP E. B. JOURDAIN.

is to be disregarded. We are there supposed to imagine a variable x

moving from a position in which f (x) is less than qp(x) to one for
which f (x) is greater than <p(x), where f (x) and qp(x) ( vary accor-
ding to the law of continuity . Hence,because ( both functions must,
because of their continuity, go through all the intermediate values
before they can attain to a higher one, )) there must be a moment at
which f(x) is equal to qp(x). The illustration of two moving bodies
is only an example which does not prove the theorem but must be
proved by the theorem itself (1). With regard to the rest of the
reasoning, BOiZANO (2) remarked that c in it an incorrect conception
of continuity is taken as basis. According to a correct explanation
of the conception of continuity, we understand by the phrase: ( a
function f (x) varies according to the law of continuity for all values
of x which lie inside or outside certain,limits , simply that, if x
is any such value, the difference f (x + w) = f (x) can be made
smaller than any given magnitude, if w can be taken as small as
wished ,. In a note, BOLZANO remarked: c There are functions
which are continuously variablefor all values of their root (argument),
a x + P x for example.
?( But there are others which vary according to the law of conti-
nuity only within or without certain limiting values. Thus,
x + / [(1 -) (2 - x)]
varies continuously only for all values of x which are less than + 1
or greater than - 2, and not for the values which lie between + 1
and + 2 (3)). Later on; BOLZANO (4) said: The function

x+ \/ [(x-2) (x + )]
has, indeed, a positive value for x =+2 and a negative value for
x = -1, but, because it does not vary according to the law of conti-
nuity within these limits, there is there no value of x for which it be-
comes zero or infinite)). The first function becomes imaginary for
x' s lying between 2 and 1; and so we must conclude that BOLZANO only
called functions (of a real variable)( continuous )) when they are real

(A) Ostwald's Klassiker, N? 153, p. 6-7.

(2) Ibid., p. 7-8.
(3) Ibid., p. 7.
(4) Ibid., p. 10.
 THE ORIOIN OF CAUCHY'S CONCEPTIONS. 697

and continuous. The second function show that the tleorem men-
tioned in BOLZANO'S heading does not hold for continuous but com-
plex functions. But, as CAUCHY showed, this BOLZANO-CAUCHY con-
ception of ( continuity ) can be extended to complex functions of a
complex variableby the addition, which is necessary even when only
real values are considered, that

f (x + w)- f (x)
can be made, in absolutevalue, less than any given positive magni-
tude. With this addition, BOLZANO'S above two functions become
( continuous ,, (1).

XV

DE MORGAN, in his treatise on The Differential and Integral Cal-

culus (2), gives three postulates relating to the character of func-
tions and the frequencyof occurrence of their singular points (3).
What is particularly interesting is that DE MORGAN describes both
CAUCHY'S conception of ((continuity)) and the older conception. He
distinguishes between them in words by referring to the two con-
ceptions as of acontinuity of value)) and cccontinuityof form)) re-
spectively. That algebraical functions are continuous in value is,
says DE MORGAN (4), (( a part of our experience of ) them; and ex-
ceptions to the claw)) of continuity of form could not, at that stage
in DEMORCAN'S book, be algebraically formed. This (law)) was the
subject of the third postulate (5).
c If any function follow one law for every value of x between
x - a and x = a + h, howerer small h may be, it follows the same law
throughout: that is, the curves of no two algebraical functions can
entirely coincide with each other, for any arc, however small. If $ x

(1) Cf. Ostwald'sKlassiker, n? 153, p. 40.

(2) London, i842, p. 44-46.
(3) The first postulateis (p. 44-45): - If (a be an ordinaryvalue of 0a, then
h can always be taken so small that no singular value shall lie between Oa and
0(a+h), that is, no singular value shall correspondto any value of x between
= a andx = a - h. n De Morganthen remarks: "The truth of this postu-
late is matter of observation. We always findsingularvalues separatedby an
infinite numberof ordinaryvalues. *
(4) bid., p. 45.
(s) Ibid., p. 46.
y
698 PHILIP E. B. JOURDAIN.

be x2 for every value of x between a and a + h, however small h

may be, it is x2 for every other value of x. This we may call the
law of continuity of form , or permanence of form ).
(( The continuity of law of a function ), says DE MORGAN
(i), (( is
not to be presumed from the simple continuity of its values. To re-
turn to the geometrical illustration: two different curves may join in
such a way that the value of y increases continuously in passing from
one to the other through the point of junction. If they have a com-
mon tangent at the junction, dy/dx may also vary continuously in
value; if they have there a common radius of curvature, d2y
may do the same. And two curves may be distinct, though the
value of y and of any finite number of differential coefficients
increase or decrease continuously in passing through the point of
junction. But if all the differential coefficients increase or decrease
continuously, then the second curve is only the continuation of the
first. )
(2) says : (...By
Later on, DE MORGAN a discontinuous function
is meant... one.. composed of branches of different curves,joining
or not. ))
It may be added.that, according to DE MORGAN (3), a function is an
((expression )): and in DE MORGAN'SElementary Illustrations(4), conti-
nuity of a point's motion is explained as ((not suddenly increased or
dec reased)).
If an algebraic equation has more roots than there are units in
its degree, it is an identity; and consequently an algebraic function
which satisfies an algebraic condition for an infinity of values of the
independent variable satisfies it identically. In other words, if the
values of an algebraic function are given for an infinity of values of
the independent variable, this function is wholly determined. A
transcendentalequation may, on the other hand, have an infinity of

(i) Ibid., p. 232.

(2) Ibid., p. 616. DE MORGAN proceedsto employ "symbols of discontinuity
which may be either made conventionallyor obtainedfrom the limiting formsof
algebraicalexpressions.
(8)Ibid., p. 35.
(4) Elementary illustrations of the differential and integral calculus, p. 25.
This little treatise is usually bound at the end of DE MORGAN'S large work on
the calculus. A convenient separate edition of it was publishedat Chicagoin
1909, and the passagereferredto is on p. 53 of this edition.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 699

roots; but the roots of equations formed with exponential, logarith-

mic, and circular functions never form a continuum; so that these
equations cannot be satisfied for all the values of the unknown is a
certain interval without being an identity. (It was,)) says HADAMARD(1)
(( to functions possessing this kind of solidarity that EULER'S
termino-
logy of ((continuous functions)) was applied; and the rigorous ma-
thematical translation of EULER'Snotion of ((continuous func-
tion is the nation of analytic function , (2).
Weierstrasswas the first to give the necessary and sufficient con-
ditions that two one-valued and analytic functions should be identi-
cal throughout their domains of existence (3). If two powerseries
have the same sum at some point z - a about which they both con-
verge, and at any infinite aggregateof points within any common
circle of convergenceround z = a, which condenses at a and at a only,
then the sums of the power-series are the same for all the points
within that circle, and consequently the functions defined by these
power-series and their continuations are identical.
We will now return to the consideration of CAUCHY'S
conception of
the (( continuity)) of a function.

XVI

returned to the question

CAUCHY of Eulerian ((continuity)) and his
own, in a note printed in the Comptesrendus for 4844 ('). There he
gave a simpler example than the above ones (? IX.) of FOURIER of a
single analytical expression which represented what EULERwould

(1) La srie de Taylor et son prolongement analytique, Paris, 1901, p. 2.

(2) Ibid., p. 5. Cf. LEBESGUE, IntEgration, p. 4, and Sdries trigonomdtriques
p. 21n.
(3) Cf. JOURPAIN,Journ. fur Math., vol. CXXVIII, 1905, p. 95. BOREL
(Lefons sur la thdorie des fonctions, Paris, 1898, p. 81-82) remarks upon the
importance of the property in question, but, by giving merely a sufficient con-
dition for identity (coincidence along a curve), he gets, it seems, into unneces-
sary difficulties about the nature of the curve. That the property is not sufficient
to characterize an analytic function in the way maintained by LEBESGUE (Sdries
trigonomdtriques, p. 21n), and consequently that such a function is not com-
pletely characterized by what JOURDAIN (loc. cit., p. 170, 196) has called its "agg-
regate of definitions, seems to result from BOREL'Stheorem (op. cit., p. 94,
100-101).
(4) uM6moire surles fonctions continues ,, (Euvres (1), vol. VIII, p. 145-160;
BRILL,op. Cit., p. 162-163.
 700 PHILIP E. B. JOURDAIN.

call a ((discontinuous) function and CAUCHY

would call a <<conti-
nuous)) one. The integral
00
Ir f x2dt
2 Jt +
coincides with + x for positive x's and - x for negative x's.
This was apparently the example to which LEBESGUE (1) referred
when he said : (( CAUCHY remarkedthat the difficulties which result
from FOURIER'S researchespresent themselves in very simple examples,
that is to say, according to the process used to give a function, it
appearsas acontinuous)), in EULER'S sense, or not. CAUCHY'S example
is the function equal to + x when x is positive and to - x when x
is negative. Thisfunction is not ((continuous)); it is formed of parts
of the two ((continuous)) functions + x and - x. But in appears as
((continuous)) when we express it as + /x2)).

XVII

Henceforth, by (( discontinuity )), we shall, following CAUCHY under

stand what ARBOGAST called ( discontiguity )).
If a function f(x) is continuous within the interval (a...b) except at
the point x = c, we have seen that CAUCHY defined the ( integral
P
f(x) dx, as the sum, which is expressed symbolically as
r+c rb
a Jc

This sum is more precisely written as

lim bf^, + ^
r 0 +C

and this is the conception of an integral taken over a point of discon-

tinuity that we meet in CAUCHY'S memoir of 1814. But we can still
further generalize the definition. The number

a Jc + E2

may or may not tend to a definite limit as e, and ?2 diminish to zero

independently of one another. Here we have what CAUCHY called, in

(1)Intdgration, p. 3-4.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 701

his Resume of 1823, (( singular integrals ). When cE = E2, we get

called the ( principal value)) of the singular integrals (i).
what CAUCHY

XVIII

The question of the existence of integrals did not arise, says

C. R. WALLNER (2), before GAUCHY.It seems probable that CAUCHY, on
developing a rigid exposition of the integral calculus on the basis of
the conception of the integral as the limit of a certain sum, which he
had found to have merits - among others, those described in the
thirteenth section above -, he perceivedthat the conception had the
further merit of giving a general demonstration of a theorem that,
if f(x) is a continuous function, we can construct, out of given quan-
tities alone, a function y of x such that
dyldx = f(x).
Such functions had, of course, been constructed for many particular
values of f(x), but people seem to have been unconscious of the advisi-
bility or possibility of any general theorem of this nature (3). It is
very often the case in science that a proposition is only discoveredto
be an answer to a hitherto unknown question after the proposition
itself has been discovered.
In LACROIX's book, the question of the existence of the roots of an
equation was, as was generally the case until the work of GAUss,mis-
understood and merged into the question as to the form of ima-
ginaries (4).
(5) gave a (<proof ) of the existence
But on the other hand, LACROIX
of differential quotients. The first attempt of this kind was due to

(i) We may remark, by the way, that LEBESGUO'S

(Integration, p. 9n) state-
- principalvalues- is an error. Herehe says that, if
ment aboutCAUCHY'S
imurn c-h _
h-O j(. a jc +h]

has a value without the separateintegrals having definite limits, CAUCHY

calls
the secondmemberthe - principalvalue wof the integral .
(2) CANTOR op. Cit., vol. IV, 1908, p. 877.
(3) Cf. for example, LACROIx,op. cit., vol. II, p. 1-2.
(4) Ibid., vol. I., p. 131-138.
(5) Ibid., p. 241-242; cf. vol. I1I, p. 712.
702 PHII,P E. B. JOURDAIN.

series by a (( proof )
AMPERE. AMPiRE(1) began a memoir on TAYLOR'S
that the function of x and i
[f (x i) - f(x)]li
can become, when we put i = 0, neither zero nor infinite for all
values of x. It necessarily results that this ratio then becomes a
function of x. , In the first place ), said AMPRRE (), (( notice that,
in order that a real function of x and i should become zero or infinite
when i = 0, this function must diminish or increase according as i
diminishes, so as, for a small enough value of i, in the first case, to
become less than every given magnitude, and, in the second case, to
surpass every given magnitude. ) To the word 0 real ,, AMPERE
added the note: (( A function might pass from a finite value to an
infinite or zero value without increasing or diminishing indefinitely,
if it became imaginary in the interval; but that cannot happen to the
ratio considered, which can only become imaginary if fx) were also
to cease to be real, and if, consequently, there would not be any occa-
sion for the researches in this memoir. ,
A point which seems to be of very great interest is the implied sup-
position that f(x) is continuous. Manyauthors, like DINI(3), REIFF (4)
and PRINGSHEIM (5) just say that this proof is to apply to ( continuous ))
functions, and do not add that the word ((continuous )) had a different
meaning in 1806, and was not used by AMPsRE.
The (( proof )) of the existence of the limit of the ratio
f (x') - f (x)
(x' - x)

given by LACROIX,was due to (( M. BINETain6, professeur de mathema-

tiques transcendantesau Lyc6ede Rennes , (6).

(i) Recherches sur quelques points de la theorie des fonctionsderiv6esqui

conduisenth une nouvelle demonstrationde la s6rie de Taylor, et A l'expression
finie des termes qu'on neglige lorsqu'onarrete cette serie A un terme quel-
conquen, Journ. de l'tc. polyt., vol. VI (cah. 13), 1806, p. 148-181.
(2)Ibid., p. 150.
(8) Fondamenti per la teoricadelle funzioni di variabili reali, Pisa, 1878,
?? 69, 169; Germantranslationby J. LUROTHand A. SCHEPP under the title:
Grundlagen fir eine Theorie der Functionen einer ver4nderlichen reellen
Gr6sse, Leipzig, 1892, p. 88-90, 298.
(4) Op. cit., p. 156.
(5)Bibl. Math. (3), vol. I, p. 449.
(6) PHILIPPE MARIE BINET who
Presumably this would be the father of JACQUES
was born at Rennesin 1786 and died at Paris in 1856 (Poggendorff's biog.-lit.
 THE ORIGIN OF CAUCHY'S CONCEPTIONS. 703

XIX

Let us now return to CAUCHY'S proof of the existence of a definite

integral of a continuous function.
(1) as marking
This existence-proof has been noticed by PRINGSHEIM
the transition from the geometrical conception of a limit to the arith-
metical one. It seems to be true that, with the exception of GAUSS,
CAUCHYwas the first to realize the necessity of existence-proofs, and
was quite the first to see that the nature of a very important class of
them consisted in the setting up of certain convergentsequences which
would define a limit. In this way, solutions of certain classes of
differential equations were first proved to exist by CAUCHY, and these
proofs were rightly emphasized by him. It is in
usually the cases
most interesting to mathematicians that the existence of an entity
can only be proved by showing that it is the limit of a convergent
sequence.
But the existence of a limit of a convergent sequence itself was
never proved by CAUCHY. This question is, of course, bound up with
the theory of irrational numbers, - a much more modern creation.
And if CAUCHY'S remark (2) that irrational numbers are the limits of
rational sequences is taken - what does not seem to me to have been
CAUCHY'S intention - as a definitionof irrationals as limits, there is a
logical error which was first avoided by Weierstrass.

PHILIPE. B. JOURDAIN.

Girton, Cambridge (England).

Handw6rterbuch,vol. I, p. 194. For other early . proofs of the existence of

derivatives,see GALOIS'(1830-1831) (E,vres, Paris, 1897, p. 9; DE MORGAN,
Calculus,p. 47-48.
()) Encykl. der math. Wiss., vol. I, A 3, p. 65.
(2) Cf. Bibl. Math. (3) vol. VI, 1906, p. 206.