The Origin of Cauchy's Conceptions of A Definite Integral and of The Continuity of A Function
The Origin of Cauchy's Conceptions of A Definite Integral and of The Continuity of A Function
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.
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.
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
III
(') 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
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 + ...
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) 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
IV
VI
(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
VII
VIII
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,...).
f (x) dx x+ T
T/ (- 712 xc /2);
r
(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.
Jo Jo Jr/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
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.>
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
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
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
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).
XIII
(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
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
(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.
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
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
XVI
XVII
a Jc + E2
(1)Intdgration, p. 3-4.
THE ORIGIN OF CAUCHY'S CONCEPTIONS. 701
XVIII
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)
XIX
PHILIPE. B. JOURDAIN.