Annals of Mathematics

Cauchy's Paper of 1814 on Definite Integrals Author(s): H. J. Ettlinger Source: Annals of Mathematics, Second Series, Vol. 23, No. 3 (Mar., 1922), pp. 255-270 Published by: Annals of Mathematics Stable URL: http://www.jstor.org/stable/1967922 . Accessed: 24/03/2013 18:24
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Annals of Mathematics is collaborating with JSTOR to digitize, preserve and extend access to Annals of Mathematics.

http://www.jstor.org

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY'S

PAPER

OF 1814 ON DEFINITE
BY H. J. ETTLINGER.

INTEGRALS.*

Introduction.In 1814 Augustin Louis Cauchy presented before the Academie des Sciences a "memoir on definite integrals,"in whichappears forthe firsttime the essence of his discoverieson residues. The memoir was firstprintedin 1825t with additional notes and again in 18821 with no change save in the matter of notation. Althoughthe kernelof the idea, that the integralof an analytic function of a complex variable taken along a closed path depends entirely upon the behavior of the functionat points of discontinuitywithin the path, is contained in the paper, yet there are several reasons why the reader might notice nothingat all like this theorem. In the firstplace, is now an essential featureof every although geometricalrepresentation presentationof the theoryof functions,Cauchy used neitherfiguresnor geometrical language. In the second place, the fundamental theorem, and, indeed, all the applications in this paper, concern simple integrals; but the author states the central problem as the determinationof the in the value of an iterated integral according to the order of difference withrespectto the two variables. By the use ofthis difference integration he obtains the residue, therebyobscuringthe relation of the latter to a line integral. Thirdly, he refrainsfrom using complex quantities, invariably separating an equation into its real and imaginaryparts. This necessitates longer equations, more of them, clumsier notation, and a much more obscure treatmentthan would be the case had he used complex quantities. Cauchy himself came to appreciate this fact, for his footnotes of 1825 are devoted to the simpler complex equations from which his real ones can be readily deduced. Finally, all editions abound in misprints. For these reasons the discoveries contained in this memoirwere not appreciated even by the great mathematiciansof his time. Poisson? saw in the paper merelya means of evaluating integralsand remarkedthat, at least so far as the firstpart was concerned,no new formulae were anAs nounced. for the evaluation of iterated integrals by the so-called
surles integrates Savants Etrangers, t " M6moire definies," 1, p. 509, Academiedes Sciences de lInstitutde France. I serie,1, p. 319 ff. $ CEuvres Completes, ? Bulletin de la SocietePhilomatique (3), 1, 1814,p. 185.
255

* Presented to the Amer.Math. Soc., Sept. 2, 1919.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

256

H. J. ETTLINGER.

"singular" integrals(which are equal to the difference between the value first with respect to x and then with respect to y obtained by integrating and the value obtained by integrating in the reverseorder) he said that, though the new method was worthy of consideration,it ought not to replace the old ones! Lacroix and Legendre, in the officialreport on the paper, stated as the valuable results obtained by Cauchy: (1) the of a seriesof generalformulae construction fortransforming and evaluating definiteintegrals, (2) the pointing out of the fact that the value of an iterated integral may depend on the order of integration,(3) the disin value, (4) the derivacoveryof the cause and amount of this difference tion of new formulae which, to be sure, might have been otherwiseobtained. It seems, then, likely that the foremostmathematiciansof that time failed to recognize the contributionsof main importance in this paper. To appreciate thoroughlythe memoir, the followingfacts must be noted in addition: (1) imaginaries had no secure arithmeticalbasis in 1814, (2) this was the first deduction by rigorousmethodsof the formulae, hitherto obtained by purely formal processes, for evaluating definite integrals,(3) while the formhad not yet been cast in the e-mould,which itself originated with Cauchy, neverthelessthe proofs are so conceived that they correspondin substance to the standards of rigorof the present day.
PART I.

Continuous integrand. In discussing the memoir we shall frequently combine two separate real equations into one complex equation, as Cauchy did in his notes of 1825 and, very likely, in his original work. We shall also adopt the language of modern analysis for the sake of clearnessand accuracy. The first that if a function theoremproved in the memoiris, in effect, of a complexvariable is analytic throughout a regionof a certaintype and continuous in and on the boundary, the integral of the functiontaken along the boundary of the region is zero.* The regions considered are mapped in a one-to-one manner and continuously,but not in general on a rectangle in the real (x, y) plane. The mapping on conformally, the complex M + Ni plane is performed by taking M and N as real and continuousfunctions of x and y with derivativesof all orderswith respect to x and y, continuousin x and y regardedas independentvariables.
* For modern treatment ofthistheorem see Osgood,Lehrbuch der Funktionentheorie, erster Band, zweiteAuflage, pp. 284-285; Pierpont,Functions of a ComplexVariable, pp. 211-214; Goursat, Coursd'AnalyseMathematique, tomeII, pp. 82-92. These threetext-books will hereafter be referred to as O., P., G., respectively.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY S PAPER

OF

1814

ON DEFINITE

INTEGRALS.

257

Let* (1) f(M + Ni) = P + Qi be an analytic functionof M + Ni in a certainregion,S, of the M + Ni plane, and let and N =t(x, y) q(x, y) M = be single-valued functions, continuous in x and y, in a rectangle, R (0 ! x _ a, 0 c y c b), and on the boundary, r, and possessingcontinuous partial derivatives of all orderswith respect to x and y in R and on r. lett Furthermore, S + Ti? f(M + Ni) O(M+ Ni) (2) (3) U + Vi =f(M O(M + Ni) + Ni) d

(2) and (3) withregardto y and x respectively: Differentiate as+i Oy Oy z~

=f (M+Ni)

+f= au
Ox

f(f ~~~~~~~~ f'(M

d(M + Ni) a(M + Ni) Oy Ox Ni)'(M +

\i

2(M + Ni)

iOVf/(M Oi x

+Ni) O(M?+Ni)
Oy

O(M + Ni)

ax~

+f(M But under the conditionsimposed O2(M + Ni) -2(M or Hence + Ni) $

+ Ni)O(M+Ni).

Oxdy

O yax

OS iOT aU iOV Oy Oy Ox Ox
OS a
@---y

(4)

OU

and

aT = OV

Oy Ox

Multiplying the equations (4) by dydx and integratingfrom x = 0 to that, since the integrand x = a and y = 0 to y = b, and noting further is continuous,the order of integrationcan be reversed,we have:
*

y = x + zi to z = x + yi.

P' + P"i to P + Qi and The notationof the original paper has been changedherefrom

t Hereafter a(M + Ni) _ i V(M+ Nvi)willbe designated byd(M + Ni) ax ay d(x + yi) vol. I, p. 13. Mathematical Analysis, tSee Goursat-Hedrick,

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

258
(5)
ddx

H. J. ETTLINGER.

dy f

TX dx.

tion (5) becomes

(6)

Let S(x, b) = S, S(x, 0) = s, U(a, y) = U, U(O, y) = u; then equafSSdx-

ra

r~a sdx
ra

= f

Udy

rb
fb

udy.

In a similarmanner,lettingT(x, b) = T, T(x, 0) = t andV(a, y) =V, V(O, y) = v, we obtain (7)

fa Tdx -

tdx =

fb

Vdy-

vdy.

zplane

FIG. 1.

Multiplying(7) by -i f(s
+ ti)dx +Jb

and (6) by
-

1 and adding we have

Ti) dx
-

(U + Vi) dy

(S +

fb(u

+ vi)dy

0,

or

d(M + Ni) ff(M + Ni) d(x + yi) = 0, JL ~~d(x+ yi)

which heregivenin thez plane,and hence meansthataround therectangle in the Al + Ni planeabout the corresponding curve,* L1,
(8) ff(M + Ni)d(M + Ni) = 0.
be an analytic function

Hence

of M + Ni in a certainregion,S, of theM + Ni plane, continuousin S, and on the boundary, L, and letM = q(x, y) and N = 41(x,y) be singleR (O _ x _ a, 0 _ y valuedcontinuous functions of x and y in a rectangle, ! b), and on the boundary,r, possessingcontinuouspartial derivatives of
* Cf. equation (3) and equations(A), footnote, I serie,1, p. 338. This CEuvres Completes, memoir willbe referred to hereafter as O.C.

FUNDAMENTAL THEOREM I: Let f(M + Ni)

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY' S PAPER

OF

1814

ON DEFINITE

INTEGRALS.

259

all orderswithrespectto x and y and mapping theclosed region,S, on the closed rectangle,R, in a one-to-one manner and continuously;then the integraloff(M + Ni)d(M + Ni) takenaround L in the positivedirection vanishes, or ff(M + Ni)d(M + Ni) = O.

In the applications Cauchy uses the real equations in S and U, T and V respectively and does not combinethemas is here done. The functions used in this paper forM and N and the corresponding maps of the rectangle,R, on the M + Ni plane are given in figures 2, 3, 4, and 5..
10.

M=x,

N=y.

(o~~~o)( co0)~~~~
FIG. 2.

N = xy; 2?. M =ax, a > O.

,Ao)
FIG. 3.

In several of the applications Cauchy allows a, the upper limit of the x-interval,to become infinite. He considered that his conclusions could be extendedto this case if the function f(M + Ni) approaches a limitfor each value of y (O < y - b) when a becomes infiniteand the improper integrals thus introduced converge. These conditions are, of course,

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

260
30

H. J. ETTLINGER.

M = xcosy,
la x e

N = x siny.

1-t- N4'

(acosi

a4

X~&
4?. M =ax,

1
FIG. 4.

a > 0.

N =xy;

(aAeb)

(oo)
FIG. 5.

f(M + Ni) considered by Cauchy but sinceall the functions insufficient, in (0 - y _ b) and the integrals conapproachtheirlimitsuniformly correct. as established results may be verge, the and of the is an exampleof the methodof application The following results ofPart I. Region10. (See Fig. 2.) M =x, ,N = y. Let f(x + yi) = P(x, y) + Q(x, y)i, such that Q(x, 0) = 0, and S + Ti - P+Qi, U+ Vi = - Q+Pi. Equations (6) and (7) yield
(6') (7')
(7')

rba

P(x, b)dx Q0 b{hdx x

JQ(

tw~a

P(x, O)dx + -I
yd+b

rb b Q(a, y)dy rb

rb Q(0, y)dy = 0,

bx-JP

+f P(a, y~dy,

P(n-

= n. y,4dy

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY S PAPER

OF

1814 ON

DEFINITE

INTEGRALS.

261

Applytheseequationsto
z = x + yi.* where e-x2ey2 sin 2xy, cos 2xy, Q(x, y) - e-xey2 P(x, y) P(x, 0) = e-2, Q(0,y) = 0, P(0, y) = ey2,
=

f(z) = e-z2

so thatin thiscase equations(6') and (7') becomerespectively (9)

(10)
fex2eb2

cos 2bxdx -

e-a2

eY2

sin 2aydy = ,fe2 cos 2aydy = -

dx.

e-f'eb2sin 2bxdx -

ea2f

ey2

ey2dy.

Now let a increasewithoutlimit. The second integralin each of for theequations(9) and (10) vanishes,
fbe-a2ee2 sin 2aydy

cf

sin 2ay ea2ey2dy

_ ea2f
b

ey2dy _ beb2-a2

b > 0.

But hence Similarly

limbeb2-a2
a

0,

lim fbe-X2ei2
rb

rb

sin 2aydy= O.

cos 2aydy = 0. lim Jbe-a2ey2

we are Since it can be readilyshownthat the otherintegrals converge, in writing Cauchy'sequations: justified

f
ifwe assume

cos 2bxdx = e-2

e'2

rX

r~~~~~~~~~~b = e-b2 e-2 sin 2bxdx ey2dy,

-

e 2dx

e_ 2 4LW

~~~~~~~~2

e-X2dx

/2. PART II.

withpoles. In the second part of the memoirCauchy Integrands at isolated of functions whichare discontinuous deals withthe integrals are simplepoles.t thesesingularities points. In all the applications
t O.C., p. 413.
* Cf. O., p. 293, Beispiel, 4. G., p. 121, 30.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

262

H. J. ETTLINGER.

Here Cauchy obtains for the firsttime a formulawhich contains the essence of the theorem on residues. The true significanceof Cauchy's method at this point is very obscure. The result is apparentlystated in of an iterated integral ordters terms of the evaluation in two different whose integrand has a singularityat a single point.* As a matter of fact,the iteratedintegralplays an unessentialrole in Part II, since all the theorems and applications are concerned with simple integrals only. iterated integrals. Moreover no useful facts are developed concerning The expositionand criticismof Cauchy's method we lay aside forthe momentand proceed to set fortha method by which the results of Part II are very simply obtained from the fundamentaltheorem of Part I. This method is not so very unlike Cauchy's, as will be pointed out later, and would probably be used by him were he writingin the notation of the present-dayanalyst. It is the method used in many modern text books on the Theory of Functions.t FUNDAMENTAL THEOREM II: Let f(M + Ni) be an analyticfunction of M + Ni in a certainregion,S, of theM + Ni plane, exceptfor a single L, ofS, in and on theboundary, pole at m + ni, inside of S, and continuous funcy) and N = {(x, y) be continuous exceptat thispole. Let M = +k(x, R (O _ x _ a, tions of x and y in and on the boundary,r, ofa rectangle, mapping of all orders, partial derivatives 0 c y _ b), possessingcontinuous and manner in a one-to-one theclosedregion,S, on theclosedrectangle, R, = = {2(X, Y). Let R' (a' and such that m +k(X, Y) and'n continuously, _ x _ a", b' c y c b") be any rectangleinteriortto R and containing to F'. in S corresponding its boundaryF', and letL' be thecurve (X, Y) within L around in the taken positive Then theintegraloff(M + Ni) d(M + Ni) off(M + Ni) -d(M + Ni) takenin thepositive sense is equal to theintegral sense around L',
.

ff(M + Ni)d(M + Ni) = f(M

+ Ni)d(M + Ni).

4plan

C.

4b

-3~_FIG . a
FIG. 6.
* O.C., p. 388 ff.

4j5

t 0., p. 331 ff.;P., p. 206 ff.;G., p. 114ff. t I.e., 0 < a' < a" < a, 0 < b' < b < b.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY' S PAPER

OF

1814

ON DEFINITE

INTEGRALS.

263

The proof follows immediatelyfrom the fundamentaltheorem I by application successively to the rectangles marked 1 ... 8 in Fig. 6 and addition of the resultingequations. The equivalent of this equation in statementof his discoveryon Residues.* Cauchy's paper gives the first We proceed to derive the formula necessaryto make the first applica-

tion by evaluating f(M + Ni)d(M + Ni)t explicitlyfor the case

M
=

x, N = y, or z = M + Ni.

Let: A + Bi Case 1.

4'

f(z)dz,and suppose: Z A + Bi
where Z= X + Yi.

f(z) = Then Let Then or

LZ -

Cdz
Z

Z = re~i,

dz = iresido + e~idr.

A + Bi=

Cid4+

I dr ~~r
Lt

A + Bi = 2 riC,?

since the initial and finalvalues of r are equal. Case 2.

f(z) =(Z)

where+(z) is analytic in R and continuousin R and on the boundary,L, and whereZ is withinR,
A + Bi =f(z) dz + Z dz

or

since
*

4'

A+Bi=27riC,
If f(z) has a pole on the boundary,L, of the rectangle,R, but

I. theorem .(z)dz = 0 by thefundamental

Case 3.

t L' is identical withr' in thiscase.

O.C., p. 381, equation (4).

?Cf. 0. C., footnote, p. 412, equation (C).

has herebeen changedfrom A' + A"i to A + Bi. l The notation

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

264

H.

J. ETTLINGER.

not at one of the vertices,we constructR" as in figure7 and denote by the boundariesof the rectanglesR and R", omitting L and L" respectively in each case the segmentAB. We apply now the fundamentaltheoremI to the rectangles1, 2, 3 and sum the results. In this way we find

ff(z)dz+

ff(z)dz

O.

Z plane

11.~~
FI G.

4R
7.

A + Bi Z'), we wirite in particular,we suppose f(z) =Cl(z If nowa, A~~~~ = J~(z) dz =Cl(z Z')dz, and z Z' = refit then dz = ireside + eidr. Hence
33

A + Bi =,>Cidq5

+ jwCow

7ri

since L" may be chosen in such a mannerthat the initial and finalvalues of r are equal. Case 4. Let fp(z) = C/(z(--Z') + ,(z) where+(z) is analytic throughRout

and Z' is a point of L, not and vertex of R.

Now

q(z)dz = d

theoremI. the of ar fundamental by eqal.f+ Case 5.

Hence here also we obtain W

A +d Bi = riC.

In generallet us considera rational function

F(z) z
-Z')

- Z

Ra and on the boundary,L, butnotat whereF(z) has only simple roots in a vertex. Then A + Bi must be computed for each pole and the results summed. Hence

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY S PAPER

OF

1814

ON DEFINITE

INTEGRALS.

265

(11)

A + Bi =

2wrifCk

7iECk

pointof R, and of 1/(z-Zk), Zk an interior whereCk is the coefficient of 1/(z -Zk"), Zk" a pointon the boundary whereCk" is the coefficient notat a vertex. Then Let Ck = Xk - iLk and Ck" = Xk" - ik". A = 2drEuk + drEuk" (12) and
(13) B = 27riXk +
faXk"'

theorem II and equations(12) and On the basis of the fundamental (13) we may workout one of the examplesgivenby Cauchyin Part II. boundedby y = 0, y = b to the rectangle Cauchyappliestheseformula without limit > 0, x = - a, x = a, and thenallowsa and b to increase (Fig. 8).
f(M + Ni) = f(z) = f(x + yi) = P + Qi,

where Q(x, 0)

O. Ib

A + Bi = ff(z)dz
-

pa

[P(O y) + iQ(O, y)]idy. S n thisequationintoreal and imaginary parts,we have Separating

iQ(x, b)]dx -

sea

-a -a (X b) +
O)dx

P(x, O)dx +

[P(a, y) + iQ(a, y)]idy

pb

(14)
(15)

A
B

fP(x,
ob

pa

fo ~~b

Q(a,

y)dy-

Pa

P(x, b)dx+

rb

Q(O,.y)dy,

P(O, y)dy. ~ ~~~~~a all integrals from the formulae exceptthose To be able to eliminate to take f(z) to be a along the axis of reals, Cauchy thinksit sufficient
=
f

P(a, y)dy-

Q(x, b)dx -

ob

conditions are called forto forstronger not at all sufficient, however, in question. It is sufficient, however, oftheintegrals thevanishing insure manner suchas e.g. in a prescribed indefinitely ifa and b increase b lim -= k = O aco a and that lim a2 + b2maxIf(x+ yi) I= 0,
* Cf. O.C., p. 422, footnote, equation (G).
Gace0

functionsuch that P and Q vanish when x =

so, y = o.

This is

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

266

H. J. ETTLINGER.

0c y beingtakenforall pointsi a + yi in the interval the maximum c b and all pointsx + bi in theinterval-a c x c a, i.e., forall points determined by (- a, 0), (- a, b), (a, b), on threesides of the rectangle and (a, 0). That is, we shall assume that, givena positivenumber, {, X, suchthat number, a positive arbitrarily small,we can find
'a2 +
b2

max lf(x + yi)

Ke,

whena > X forall pointsi a + yi in the interval0 c y c b and all pointsx + bi in theinterval- a c x a. Then

Q(-- a, y)dy
fb

f(

a + yi) dy c j: 4
e Jo~a2
+z2

f~~~~~~~~~~

dy < a

+b
a2+
b2

< {',

whena > X. in thisprescribed manner, indefinitely Hence,as a and b increase

a, b-. cO

lim

Q(dI a, y)dy=0.
narb

it maybe provedthat Similarly,

a, ba -J*.0a

lim

ra

P(x, b)dx = 0, lim IQ(x, b)dx = 0, lim OD

a. baaJ0
baa-P

P(O, y)dy= 0

--O be

lim fQ(o0

rb

y)dy = 0,

a, b -ox=

lim JP(a,

rb

y)dy = 0.

Moreover,if in addition 1ima, af(a) = 0, then faP(x, O)dx converges forif e is positiveand arbitrarily small,there as a increases indefinitely, X suchthat If(x)I < e/afora > x > X, and number existsa positive
fP(x, O)dx |f f(x) dx

xa

_ (a-X)<e fora > X. theabove conditions, a function Formula(15) tellsus that,for fulfilling B = 0, and theformula (14) reducesto
fpdx
=

A,

A is takenforall the poles off(z) where where y _ 0. If f(z) is an even function, (16) becomes

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY' S PAPER

OF

1814 ON
=

DEFINITE

INTEGRALS.

267

on 2fpdx.

of the two positive Let f(z) = zlm/(1 + Z2n), wheren is the greater m and n. Then Q(x, 0) - 0 and P(x, 0) = x2m/(1 + X2n), an integers, evenfunction. Hence
(17) A =2 2f2 co x2m -+ 1dx.

a path fora + bi such thatlim b = k 4 0 and We mustnow specify lim ala2+ b2maxIf(x+ yi) I = 0 forall points -4 a + yi in theinterval
a--*oo 0 c y a-

b and all pointsx + bi in theinterval-a

x c a.

FIG. 8.

We willtake a = b (see Fig. 8).

Then

la2 + b2max f(i a + yi) _ '2a (1 a + ai)2| 2| ( t 1 + i)2m|

a2(n-m)-1(a-2n+ C 22m+J

1)

1
1 a2(n-m)-l

a-2n

indefinitely. zeroas a increases approaches The last expression Similarly Wa2+ b2max f(x + bi) l -'12a (a+ aiY?m 1 + (ai)2|
a2n
+ i2n 2(n-m)-i

indefinitely. approacheszeroas a increases obviously The last expression Also

conitn The-p~

limaf(a) =ai

a2m+1

21+ a2th

satisfied. to justify sufficient The conditions (17) are therefore

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

268

H.

J. ETTLINGER.

The poles off(z) are to be foundwhere z~n

Zk =

1 =

0, or
1.

ek-

t2k+ 1,

2n J

k= 0, 1 *

2n

certainly We observe that the poles are on the unit circle and therefore inside the rectangleR as soon as a > 1.
z2m Co

1+z2n

C' 3s + + 4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-f4n-1 ~~+ z - e 2n z- e e2n ,,IT

7r

C1

C2n-1

Now
Ck = ]im2 2-Zk

1+

Zk) Zn 1) z2m 2nz2n2Mz2m-lZk

lim (2m +
z-Zk ______

Zk2m

2nZk2n-1

1 -Zk 2(m-n)+l 2n

2n
=
-2n

(2m+1-2n) [(2k+1) 2n]

(2k+1) (2m+1) 2n-.

Also

2n-1

Ck -E k=O
-

2n-1

~~n k=O
-

e(2k+l)

2m+ +

2n

7r

1 1

2n 1 i

-2-

e(2m+1)7ri 2m+1 2n 2m+1 7ri-~~~e

2m+1 2ne n

2=

-e

2m+1

i 2n sin

2m + 1

7r

And

= A = 2rE Msk k~o

2n- 1

2n sin

_r22

2m+ 7r

2n

_ .

Hence

JjX21

rX x2m

dx=

2n sin 2n+1 Then

sina 7r

7r

Let 2m + 1 = aand 2n =f.

tJO 1 + xI

rXxa-1dx=

a formulawhich Euler had obtained.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

CAUCHY' S PAPER

OF

1814

ON DEFINITE

INTEGRALS.

269

In a similarmannerotherformula are obtained by taking otherfunctions and other regions and integratingaround the correspondingrectangle.* We returnto considerthe method used by Cauchy in the second part to obtain the results of the fundamentaltheorem II and its immediate place, Cauchy adopts the "principal corollariesproved above. In the first regardingthe evaluation of value" definitionto remove any difficulties on the path of integration.t simple integralsdue to singularities Suppose we have ?'(x), the derivative of a real function+(x) of a real variable. Consider

(19)

f
ad

ct'(x)dx,

at (X) between c and d discontinuity where ?'(x) has a finiteor infinite but is continuous in c c x < X and X <x c d. By the "principal value" of (19) is meant A'd r pX-h l '(x)dx lim 0 q'(x)dx + Iq
h-o

Lc

JX4h
-

lim [h-*o

q(c) + g(X
4(c) - A,

h)

g(X + h) + 4(d)]

where

-ck(d)
13 -.O-

A = lim [4(X + h) - 4(X - h)].

of convergence ofan improper integral, Accordingto the moderndefinition conditionfor the existenceof this limit is a necessary but not a sufficient the convergence of (19); Cauchy takes it as his working definition.t Secondly, to definethe improperiterated integralswhich occur, Cauchy proceeds as follows. Let U(x, y) be a functionwhich is continuousin x and y and possesses a continuous partialderivativewithrespectto x everywhereinside a rectangle,R (O _ x _ a, 0 c y _ b), and on the boundary, singularity L, except at the point (0, 0) whereit possesses a non-removable but does not become infinite.? ThentI

(20)

fdy af

) dx =

limJI dy

where t > 0. This definition is totallyinadequate, since the simple integralobtained after a firstintegrationdoes not even come under the principal value and may even diverge. definition
* See O., pp. 289-295; G., pp. 118-122. t O.C., p. 402. O.C. Cf. exampleon p. 404,
_2

fU

dx

dz/z.

? Cf. examplesgivenby Cauchy,O.C., p. 394 and p. 397.

01 O.C.,p. 390.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions

270

H. J. ETTLINGER.

It is, however,to be noticed that these insufficient definitions do not impair the value of Cauchy's results, nor do they substantially affect the method. If the discontinuity occurs at a cornerof R, the method is not applicable in general, as Cauchy's formulaitself shows.* For the of equation (20) above is an attempt to "cut-out" the singudefinition larity. But this does not yield a convergentresult for this case. We cannot,therefore, put any real contentinto this particularresultfromour modernpoint of view and have, therefore, excluded it in our treatment. Cauchy, himself,makes no use of this equation in any of the numerous applicationsof the memoir. When the point of discontinuityoccurs inside of R at (X, Y), the rectangleis divided into fourpartst by the lines x = X, y = Y, and the iterated integral is separated in a manner corresponding to the double integrals over each of the four rectangles. When these four integrals are added together,we have in rather obscure form what amounts to the method which we have set forthabove. The singularpoint has been y Y + a, and a method of evaluating the line integralabout the small rectangleis given in the form: t A = lim lim J (21) U(X+ y)dy + jl
-

"cut-out"

by a small rectangle, x = X

i,

x = X + i, y = Y -77,

U(X (X -

y)dy-

U(X -

dy

The bracket is substantiallythe real part of our equation (11). The difference betweenour methodof evaluation of (11) and Cauchy's method of evaluating? (21) is a strikingexample of the economy of the complex variable formulation. When the point of discontinuity is on the boundary of R, the value of A is given by two termsIIof (21). In both cases the resultsafterintegration are identical with those of (12) and (13). In the historicalreviewofthe theoryoffunctions by Brilland Noether? a brieftreatmentof the historicalimportanceof the memoiris given but not froma criticalpoint of view as is here done.**
UNIVERSITY * OF TEXAS, ArTSTIN, TEXAS.

O.C. Cf. third equationunder(20), p. 412. t O.C., p. 396. t O.C. Cf. p. 397, equation (13). ? O.C., pp. 406-412. O.C., p. 400. ? Jahresbericht der deutschen Mathemataken-vereinigung, vol. 3 (1894), p. 165ff. ** The above paper has grown out of an investigation of Cauchy'sworkon definite integrals and residues in a Seminarcourseat HarvardUniversity. Some ofthe earlyworkwas donewith the collaboration ofDr. E. S. Allen.

This content downloaded from 187.101.182.82 on Sun, 24 Mar 2013 18:24:03 PM All use subject to JSTOR Terms and Conditions