Abels Complete Work Houzel PDF

The Legacy of Niels Henrik Abel The Abel Bicentennial, Oslo 2002
Springer-Verlag 2004. (Editors O. A. Laudal and R. Piene)
The Work of Niels Henrik Abel
Christian Houzel
1 Functional Equations
2 Integral Transforms and Definite Integrals
3 Algebraic Equations
4 Hyperelliptic Integrals
5 Abel Theorem
6 Elliptic functions
7 Development of the Theory of Transformation
of Elliptic Functions
8 Further Development of the Theory of Elliptic Functions
and Abelian Integrals
9 Series
10 Conclusion
References
During his short life, N.-H. Abel devoted himself to several topics characteristic
of the mathematics of his time. We note that, after an unsuccessful investigation
of the influence of the Moon on the motion of a pendulum, he chose subjects in
pure mathematics rather than in mathematical physics. It is possible to classify these
subjects in the following way:
1. solution of algebraic equations by radicals;
2. new transcendental functions, in particular elliptic integrals, elliptic functions,
abelian integrals;
3. functional equations;
4. integral transforms;
5. theory of series treated in a rigourous way.
The first two topics are the most important and the best known, but we shall
see that there are close links between all the subjects in Abels treatment. As the
first published papers are related to subjects 3 and 4, we will begin our study with
functional equations and the integral transforms.
22 C. Houzel
1 Functional Equations
In the year 1823, Abel published two norwegian papers in the first issue of Ma-
gasinet for Naturvidenskaberne, a journal edited in Christiania by Ch. Hansteen.
In the first one, titled Almindelig Methode til at finde Funktioner af een variabel
Strrelse, naar en Egenskab af disse Funktioner er udtrykt ved en Ligning mellom
to Variable (uvres, t. I, p. 110), Abel considers a very general type of functional
equation: V(x, y, , f, F, . . . , , f , F , . . . ) = 0, where , f, F, . . . are
unknown functions in one variable and , , , . . . are known functions of the two
independent variables x, y. His method consists in successive eliminations of the
unknown , f, F, . . . between the given equation V = 0 and the equations obtained
by differentiating this equation with constant, then with constant, etc. If, for
instance = const, there is a relation between x and y, and y may be considered as
a function of x and the constant value of ; if n is the highest order of derivative of
present in V , it is possible to eliminate and its derivatives by differentiating V
n + 1 times with constant. We then eliminate f and its derivative, and so on, until
we arrive at a differential equation with only one unknown function of one variable.
Naturally, all the functions, known and unknown, are tacitly supposed indefinitely
differentiable.
Abel applies this to the particular case = f(x, y, , ), where f, , and
are given functions and is unkown; he gets a first order differential equation
with respect to . For instance, the functional equation of the logarithm log xy =
log x + log y corresponds to the case where (x, y) = xy, (x, y) = x, (x, y) = y
and f(x, y, t, u) = t + u; differentiating with xy = const, we get 0 = x x y y,
from which, with y = const, we get x = xc , where c = y y. In the same way, the
functional equation for arctangent,
x+y
arctan = arctan x + arctan y,
1 xy
x+y
corresponds to (x, y) = 1xy , (x, y) = x, (x, y) = y and f(x, y, t, u) = t + u;
differentiating with constant gives 0 = (1 + x 2 ) x (1 + y2 ) y, whence x =
c
1+x 2
if c = (1 + y2 ) y.
When (x, y) = x, (x, y) = y and f(x, y, t, u) = t u, we get first

y x x y = 0,
y x
x
whence x as a known function of x if y is supposed constant. For (x, y) = x + y,
x y
this gives x = c = y , so log x = cx(for (0) = 1) and x = ecx ; for
x
(x, y) = xy, x = xc , so log x = c log x ((1) = 1) and x = x c .
All these examples were classical as is the next one, coming from mechanics. The
law of composition of two equal forces making an angle 2x leads to the functional
equation
The Work of Niels Henrik Abel 23
x y = (x + y) + (x y); (1)
where x is the ratio of the resultant force to one of the two equal forces. Differ-
entiating with y + x = const, one gets x y x y = 2 (x y); another
differentiation, with x y = const, gives x y x y = 0. If y is regarded as
constant, this gives x +cx = 0 and x = cos(x +) with , and constant.
From (1), one sees that = 2 and = 0 and the problem imposes 2 = 0, so
= 1 and (x) = 2 cos x.
Here is another case of application of Abels general method: the equation has
the form = F(x, y, x, x, . . . , f y, f y, . . . ), where is a given function of x
and y and , f, are unknown functions. By differentiating with constant, one
gets a relation between x, x, x, . . . and y, f y, f y, . . . , whence two differential
equations, with respect to and to f , considering successively y and x as constant;
if and f are determined, it is easy to determine by the functional equation.
In particular, if (x + y) = x f y + fy x, so that (x, y) = x + y, the
differentiation with constant gives x f y f y x = 0, and x = a sin(bx +c),
fy = a sin(by + c) then (x + y) = aa b sin(b(x + y) + c + c ) so that =
aa b sin(b + c + c ).
In the case of (x+y) = f(xy)+(xy), one gets 0 = f (xy)(yx)+2 (xy).

Abel takes xy = c as constant and writes = k, where = x y and k = f 2(c) ,

so = k + 2k 2 ; then he takes x y = c constant and writes f = c = 2c c , so
f = c + c . Finally
k
(x + y) = c + c xy + k + (x y)2
2
or = c +c x( x)+k + 2k (2x )2 = c + 2k 2 +k + x(c 2k)+(2k c )x 2
and we see that the condition c = 2k is necessary; = k + c + 2k 2 .
The third example is (x + y) = x f y + fx y, which gives
0 = x f y x f y + f x y fx y; (2)
if one supposes that f(0) = 1 and (0) = 0, one gets 0 = x x c + fx c by
making y = 0 (c = f (0) and c = (0)); so fx = kx + k x and, substituting
this value in (2) and making y constant: x + a x + bx = 0 etc.
Abel returned to the study of functional equations in the paper Recherche des
fonctions de deux quantites variables independantes x et y, telles que f(x, y), qui
ont la propriete que f(z, f(x, y)) est une fonction symetrique de z, x et y, published
in German in the first volume of Crelles Journal in 1826 (uvres, t. I, p. 6165).
The condition of the title characterises a composition law which is associative and
commutative; it may be written as f(x, y) = f(y, x), f(z, f(x, y)) = f(x, f(y, z)) =
f(y, f(z, x)) or
f(z, r) = f(x, v) = f(y, s) (3)
if f(x, y) = r, f(y, z) = v and f(z, x) = s. Differentiating with respect to x, to y
and to z and multiplying the results, one gets
24 C. Houzel
r v s r v s
= . (4)
x y z y z x
v
But, by the definition of v, the quotient of y by v
z is a function y when z is
s s
regarded as constant; in the same manner, x is the quotient of x by z , so (4)
r r
becomes x y = y x and this gives r as an arbitrary function of (x) + (y),
where is a primitive of . So f(x, y) = ((x) + (y)); putting this expression
in (3) and making z = y = 0 and x = p, one gets p = p + c, where
c = (0), and then f(x, y) = (x) + (y) + c or
f(x, y) = (x) + (y) (5)
where (x) = (x)+c. In other words, Abel finds that f is conjugate to the addition
law by the function : he has determined the one-parameter groups.
The second volume of Crelles Journal (1827) (uvres, t. I, p. 389398) contains
another paper of Abel on a functional equation:
x + y = (x f y + y fx) = (r), (6)
where r = x f y + y fx; this equation includes, as particular cases, thelaws of

addition for log ( fy = 12 y, x = x = log x) and for arcsin ( fy = 1 y2 ,
r r r r
x = x = arcsin x). One has x = r x , y = r y , so x y = y x
or
y ( f y + y f x) = x ( fx + x f y), (7)
whence, for y = 0,
a x ( fx + x) = 0, (8)
where a = (0), = f(0) and = f (0), a differential equation which deter-

mines if f is known. Substituting in (7), one gets ( fx + x)( fy + y f x) =
( fy + y)( fx + x f y) or
1 1
( fy f y f y y f y) = ( fx fx f x x f x) = m,
y x
necessarily constant. So
f x ( fx + x) + (mx fx) = 0, (9)
which determines f ; as this differential equation is homogenenous, it is easily

integreted by putting fx = xz, in the form log clog x = 12 log(z 2 n 2 )+ 2n log zn
z+n ,
2
where m = n and c is a constant of integration. One gets

c2n = ( fx nx)n+ ( fx + nx)n ,
x
with c = , then by (8) and (6) is verified if x = + (0). Abel explicitly
treats the case in which n = = 12 : fx = + 12 x, then x = a log( + x) + k
and x = 2k + a log(2 + x).

The relation 2n = ( fx nx)n+ ( fx +nx)n , which determines f , allows us to
express fx nx, and then x and fx, in terms of fx + nx = v; turning back to (8), this
a
gives x = n+ log(cnx +c fx). When n = 0, the relation which determines f takes
fx
x
the form e = fx and we have x = a ax 2a ax
log c+ fx , x = log c+ f ( x ) .

The equation (6) signifies that f x fy+y
fx
= fx fy and Abel verifies that it is
satisfied. Another particular case is that in which = . When m is finite, (9)
reduces to x f x fx = 0, so that fx = cx; when m is infinite and equals p , (9)
becomes x fx px fx = 0 and fx = px log cx. In this last case, one gets by (7)
y y x x = 0, whence x x = k constant and x = k log mx (a new m) and then
( pv log c2 v) = k log m 2 v.
A memoir left unpublished by Abel is devoted to the equation x + 1 = ( fx),
where f is given and unknown (uvres, t. II, p. 3639, mem. VI). Abel introduces
a function such that fy = (y + 1); one may take arbitrarily on the interval
[0, 1] and define on [0, +[ by (y + n) = f n (y) (and on ] , 0] by
(y n) = f n (y) if f is bijective). For x = y, the functional equation
becomes 1 + y = (y + 1), so that y = y + y where is any periodic
function of y with period 1. Denoting the inverse function of by `, Abel gets
x = `x + (`x).
y
As an example, he takes fx = x n and y = an , so that `x = log log log
xlog log a
n and

log log x log log a log log x log log a
x = + ,
log n log n
for instance x = logloglogn x if = 0 and a = e.

Abel treats in a similar manner the general equation F(x, ( fx), (x)) = 0,
where F, f and are given functions and is unknown. Supposing that fx = yt
and x = yt+1 or yt+1 = ( f yt ), one has F( f yt , u t , u t+1 ) = 0, where u t = yt ;
this difference equation has a solution u t = t and z = (yt ). For instance the
equation (x)2 = (2x) + 2 leads to (u t )2 = u t+1 + 2 and, if u 1 = a + a1 , this gives
t1 1
u t = a2 + 2t1 ; on the other hand yt+1 = 2yt , so that yt = c 2t1 (c constant)
a
and 2t1 = xc . Finally, x = bx + bx (b = a1/c ). As we see, this type of equations
is treated with a method different from the preceeding one, by reduction to finite
difference equation.
Another type of functional equation is related to the dilogarithm
x2 x3 xn
x = x + 2
+ 2 + ... + 2 + ... ,
2 3 n
which Abel studies in the posthumous memoir XIV (uvres, t. II, p. 189193) after
Legendres Exercices de Calcul integral. The study is based on the summation of
26 C. Houzel
the series (for |x| 1) in the form of an integral

dx
x = log(1 x) (10)
0 x
and Abel reproduces several functional equations given by Legendre, as for example
2
x + (1 x) = log x log(1 x).
6
But he adds a remarkable new property:

x y y x
= + (11)
1x 1y 1x 1y
y x log(1 y) log(1 x)
for (x, y) in the interior domain of the figure
a y
In order to prove (11), Abel substitutes 1a 1y for x in (10):

a y dy dy 1a y
= + log
1a 1 y y 1y (1 a)(1 y)

dy y dy
= log 1 + log(1 y)
y 1a y

dy a dy
log 1 + log(1 a)
1y 1y 1y

y dy a
= y log 1
1a 1y 1y
log(1 a) log(1 y),
a
where the remaining integral is computed by taking z = 1y as variable:

dy a dz a
log 1 = log(1 z) = z = + const.
1y 1y z 1y
The constant of integration is determined by taking y = 0 and is found to be a.

Abel was the first mathematician to give a general and (almost) rigourous proof
of Newtons famous binomial formula
m(m 1) 2 m(m 1)(m 2) 3
(1 + x)m = 1 + mx + x + x + ... (12)
2 23
He published his demonstration in the first volume of Crelles Journal (1826,
Recherches sur la serie 1 + mx + m(m1)
2 x 2 + m(m1)(m2)
23 x 3 + . . . , uvres, t. I,
p. 218250). He uses an idea of Euler, already exploited by Lagrange and Cauchy:
writing (m) the second member of (12), one proves that
(m + n) = (m)(n), (13)
so that (m) = Am = (1 + x)m for m rational as was observed by Euler. Lagrange

extended this proof to every value of m admitting that is an analytic function of m.
Cauchy used an analogous strategy for m real and |x| < 1 using the continuity of ,
for which his proof was unfortunately incomplete. Abel considers the most general
case, with x and m complex, with |x| < 1 or |x| = 1 and Re m > 1 (if x = 1,
one needs Re m > 0).
For m = k + k i, (m) = f(k, k )(cos (k, k ) + i sin (k, k )), with f ,
continuous functions of k, k real. The continuity is almost established by Abel
in his theorem V, but this theorem is not entirely correct. The concept of uniform
convergence did not exist at that time and it was not easy to give a general theorem
for the continuity of the sum of a series of continuous functions. The functional
equation (13) becomes
f(k + , k + ) = f(k, k ) f(, ); (14)

(k + , k + ) = 2m + (k, k ) + (, ),
where m is an integer, which must be constant because of the continuity of . In

a first step, Abel treats the functional equation for ; putting k = (k, k + ) =
2m + (k, k ) + (0, ) he gets
k + = a + (k + ), (15)
with a = 2m + (0, k ) + (0, ), whence
k = ck + a, (16)
where c is a function of k , . Indeed, taking = k, 2k, . . . , k in (15) and adding

the results, Abel gets k = ( 1)a + (k) and = ((1) a) + a for k = 1,
28 C. Houzel

a natural integer; then, for k = (, N, = 0), = ( 1)a +

and = c + a, with c = (1) a. This formula is extended to the negative
values of k using (k) = 2a k and, by continuity, to every real value of k. So
(k, k + ) = ck + 2m + (0, k ) + (0, ), (17)
where c = (k , ), a function of k and . For k = 0, this gives
(0, k + ) = 2m + (0, k ) + (0, ),
a functional equation which may be treated as (15) and which has the solution
(0, k ) = k 2m,
with an arbitrary constant ; then (17) becomes
(k, k + ) = (k , ) k + (k + ) 2m,
also equal to 2m + (k, k ) + (0, ) = (k, k ) + by (14), so that (k, k ) =

Fk k + k 2m, with Fk = (k , ) independent of and F(k + ) = Fk =
F(0) = a constant. Finally
(k, k ) = k + k 2m. (18)

To treat the functional equation (14) for f , Abel writes f(k, k ) = e F(k,k ) and
F(k + , k + ) = F(k, k ) + F(, ), a functional equation analog to that for
with m = 0, so its solution is of the form F(k, k ) = k + k , with two arbitrary
constant , . Finally

(k + k i) = ek+ k (cos(k + k ) + i sin(k + k )) (19)
and it remains to determine the constants , , and .

For k = 1 and k = 0, (1) = 1 + x = 1 + cos + i sin , where = |x| < 1
and = arg x; this gives e cos = 1 + cos and e sin = sin , so that
1 sin
e = (1 + 2 cos + 2 ) 2 and tan = , = s + , (20)
1 + cos
with 2 s 2 and Z. Now, for k = 0 and any k, let p = f and q =
designate the real and the imaginary part of the series (k), which are continuous
functions of after Abels theorem IV (which is correct); one has
f = ek cos ks cos k ek sin ks sin k,

= ek sin ks cos k + cos ks sin k
and cos k = ek ( fcos ks+sin ks), sin k = ek (cos ks fsin ks),

independent of by continuity. For = 0, e = 1 and s = 0 after (19) whereas
f = 1 and = 0, so k = 0 and
k k
f = (1 + 2 cos + 2 ) 2 cos ks, = (1 + 2 cos + 2 ) 2 sin ks; (21)
this is Cauchys result for f + i = |1 + x|k (cos ks + i sin ks) = (1 + x)k .

Abel now considers the case in which m = in is purely imaginary; then the
series (12) is convergent for any value of n by dAlemberts rule (which is Abels
theorem II) and Abel states its continuity as a function of n as a consequence of his
theorem V. He writes the real and imaginary parts of the series in the form
p = 1 + 1 cos 1 + . . . + cos + . . .
and q = 1 sin 1 + . . . + sin + . . . ,
where = 1 2 . . . , = + 1 + 2 + . . . + and
ni + 1
= (cos + i sin ).

From (19) he knows that p = e n cos n and q = e n sin n; in order to determine

e n cos n1 e n sin n 1
resp. , he takes the limits of n resp. n for n = 0. As
1
and ( 2; for = 1, 1 = 2 ), he gets n and ( + ) 2
so
1 1
= cos 2 cos 2 + a3 cos 3 . . . ,
2 3
1 2 1
= sin + sin 2 3 sin 3 + . . .

2 3
f1
Now, computing in the same manner the limits, for k = 0, of k and k , one gets
from (21)
1 1
= cos 2 cos 2 + 3 cos 3 . . . (22)
2 3
1 2 1 3
and = sin sin 2 + sin 3 . . . ,
2 3
so that = and = . The sum (19) of the series (12) for m = k + k i is

ekk (cos(k + k ) + i sin(k + k ))
with and as in (20). Let us interpret Abels result: writing + i = log(1 + x),
one gets
m log(1 + x) = (k + ik )( + i) = k k + i(k + k ),
so that (m) = (1 + x)m .

Comparing (20) and (22), Abel gets
1 1 1
log(1 + 2 cos + 2 ) = cos 2 cos 2 + 3 cos 3 . . .
2 2 3
30 C. Houzel
and
sin 1 1
arctan = sin 2 sin 2 + 3 sin 3 . . . ; (23)
1 + cos 2 3
by making tend toward 1, 12 log(22 cos ) = cos 12 cos 2 13 cos 3. . .

and 12 = sin 12 sin 2 + 13 sin 3 . . . for < < . If = 2 and
1 1 in (23), one gets Gregorys series arctan = 13 3 + 15 5 . . .
Taking x = i tan and m real in the binomial series, Abels finds

m m(m 1) 2 m(m 1)(m 2)(m 3) 4
cos m = (cos ) 1 (tan ) + (tan ) . . . ,
12 1234

m(m 1)(m 2)
sin m = (cos )m m tan (tan )3 + . . .
123
for 4 4 (for = 4 , m must be > 1).

Now, taking |x| = 1 and m > 1, he finds as the real part of
(1 + x)m (cos i sin ) :
m m(m 1)
cos + cos( ) + cos( 2) + . . .
1 12
m m
= (2 + 2 cos ) cos
2 + m
2
where is an integer such that | 2| (with the restriction m > 0 in

case
of equality).
The substitutions = 2x and = mx, mx + 2 , m x + 2 or
m x + 2 2 give Abel various formulae, for instance
m m(m 1)
(2 cos x)m cos 2m = cos mx + cos(m 2)x + cos(m 4)x + . . .
1 12
m m(m 1)
(2 cos x)m sin 2m = sin mx + sin(m 2)x + sin(m 4)x + . . .
1 12
for 2 2 x 2 + 2 . Abel was the first to prove rigourously such formulae
for m non integer; in a letter to his friend Holmboe (16 January 1826, uvres, t. II,
p. 256), he states his result and alludes to the unsuccessful attempts of Poisson,
Poinsot, Plana and Crelle.
Other examples of functional equations in Abels work may be mentioned, as
the famous Abel theorem (see 5), which may be interpreted in this way. In a letter
to Crelle (9 August 1826, uvres, t. II, p. 267), Abel states his theorem for
genus 2
in a very explicit manner: he considers the hyperelliptic integral (x) = (+x)dx

P(x)
where P is a polynomial of degree 6; then Abels theorem is the functional equation
(x1 ) + (x2 ) + (x3 ) = C ((y1 ) + (y2 )), where x1 , x2 and x3 are independant
variables, C is a constant and y1 , y2 are the roots of the equation
c2 a
c22 + 2c1 a4

2 x1 x2 x3
y x1 x2 x3 y + = 0,
2c2 a5 2c2 a5
with P(x) = a + a1 x + a2 x 2 + a3 x 3 + a4 x 4 + a5 x 5 + x 6 and c + c1 x j + c2 x 2j +

x 3j = P(x j ) for j = 1, 2, 3. Abel says that this functional equation completely

characterises the function .
Abel discovered how to express the elliptic functions as quotients of two entire
functions of the type of Weierstrass -function; there is an allusion to that in the
introduction to his Precis dune theorie des fonctions elliptiques, published in the
fourth volume of Crelles Journal (1829, uvres, t. I, p. 527528) and in a letter to
Legendre (25 November 1828, uvres, t. II, p. 274275). The elliptic function ()
is defined by

dx
= , where (x, c) = (1 x 2 )(1 c2 x 2 ),
(x, c)
0
and = f where the entire functions and f are solutions of the system of
functional equations ( + ) ( ) = ( f )2 ( f)2 , f( + )
f( ) = ( f f )2 c2 ( )2 . This system is partially solved in a notebook
of 1828, with x and y in place of and ; supposing odd and f even and
taking the second derivative with respect to x at x = 0, Abel finds the equations
f y + f y ( f y)2 = a( fy)2 c2 b(y)2 and y + y + ( y)2 = b( fy)2 a(y)2
with a = f(0) f (0) and b = ( 0)2 . If it is supposed that a = 0 and b = 1,
this reduces to ( f y)2 f y fy = c2 (y)2 , ( y)2 y y = ( fy)2 . Again
differentiating four times at x = 0, Abel obtains the derivatives of f up to the 4th
order and , but his computation, aimed to find differential equations for f and ,
stops here.
Two posthumous papers by Abel are devoted to differential equations of Riccati
type. In the first one, Sur lequation differentielle dy + ( p + qy + ry2 )dx = 0, ou
p, q et r sont des fonctions de x seul (uvres, t. II, p. 1925), Abel shows how to
transform this equation in another one of the form dy + (P + Q y2 )dx = 0. Two
methods are proposed. The first one, by putting y = z + r with r = 2rq , which
q2 dq 1 dr q
gives dz + (P + Qz 2 )dx = 0 with P = p 4r dx 2r + dx 2r 2 and Q = r.

The second
one, which is classical,

by putting y = zr with
r = e qdx

; this
qdx qdx qdx qdx
gives P =
pe and Q = re . Abel observes
that when
pe = are or

qdx ar 1 dr dp 2
e = p , the equation, which is written dy+ p+ 2 rdx pdx y+ry dx = 0,

may be integrated in finite terms, giving y = rp tan

r pdx . For example,
2
2
the equation dy + 1x yx dx = 0 has a solution of the form y = 1cx 1+cx 2
and
m 1 y n 2

the equation dy + x + 2 (n m) x + x y dx = 0 has a solution of the form
mn
1

2
y = x 2 tan c + m+n+2 x 2 (m+n+2) ; in the case in which n = m 2, this
32 C. Houzel
solution becomes y = x m+1 tan(log k x). Another easy case of integration

is given
p q
2 a2 c
pdx
by the relations c = 2a = r; in this case y = a + a2 c 1+e 2 2
.
c
a c pdx
1e c
Abel explains how to solve the equation when a particular solution y
is known.
(q+2ry )dx
Putting y = z+y , he finds dz+((q+2ry )z+rz 2 )dx = 0 and y = y +
e
(q+2ry ) .
e rdx

1 ay 2
For example the equation dy + x 2 + x + cy dx = 0 has the particular solution

1a 2
y = 1a 1 1

2c 2c c x and this leads to the general solution

2 1 (1a)2 4c
1a 1a 1 1 kx
y= + .
2c 2c c x C ck 2 x (1a) 4c
2
(1a) 4c
Other cases of integration are found by Eulers method of integrating factor: the
2 ))
expression zdy + z( p + qy2 )dx is a complete differential when x z
= (z( p+qy
y
r r
or, if z = er , when x = ( p + qy2 ) y + 2qy. Abel tries with r = a log( + y)
with a constant and , functions of x only. He finds the conditions a a p =
a 2q = aq + 2q= 0, where , are the derivatives of , . Thus the

equation dy + y2 dx = 0 admits the integrating factor z = (+y) 1
2 and the
1
solution y = +

.
2 C 2 dx

In the second paper, Abel considers the differential equation
(y + s)dy + ( p + qy + ry2 )dx = 0,
which is reduced to the form zdz +

(P + Qz)dx = 0 by the substitution
y = + z with = s and

= e rdx . One has P = ( p qs + rs2 )e2 rdx
ds

and Q = q 2rs
dx e rdx . If P = 0, this equation has the solution z =

ds

2rs + dx q e rdx dx so that the equation
(y + s)dy + (qs rs 2 + qy + ry2 )dx = 0

ds

has for solution y = s + e rdx 2rs + dx q e rdx dx. When Q = 0, the

equation in z has the solution z = 2 (qs p rs2 )e2 rdx dx and the equation

ds
(y + s)dy + p + 2rs + y + ry2 dx = 0
dx

2
rdx
has for solution y = s + e rdx 2 rs2 p + sds dx e dx.
In order that z = er be an integrating factor for the equation
ydy + ( p + qy)dx = 0,
r r
we must impose y x ( p + qy) y q = 0. For r = + y, this gives the conditions
d d

dx = dx q = p + q = 0, so = c, = c qdx and cp
+ q = 0.
For r = + y + y2 , one finds = c, = 2c qdx, q + 2c p qdx = 0

and = 2c qdx qdx

qdx

qdx
. When q = 1, we find that the equation ydy +

1 1 2c (x+y+a)2
c(x+a) + y dx = 0 admits the integrating factor x+a e . More generally,
for r = + 1 y + 2 y2 + . . . + n yn , one finds n + 2 conditions d n
dx = 0 =
dn1 dn2 d
dx nqn = dx (n1)qn1 n pn = . . . = dx q1 2 p2 = q+ p1 = 0
for the n + 1 coefficients k ; so there is a relation between p and q. For n = 3, Abel
finds

q + 6c p qdx qdx + 3c p pdx = 0.
A function r = leads to the conditions d

1
+y
2 d
+ 2q =
dx + q = dx q
2
2 q p = 0 and the equation ydy + C
+ 12 q qdx + qy dx admits

( qdx )2

1
the integrating factor e +y with =
1 C
and = + 12 .
qdx ( qdx )2

Another form tried by Abel is r = a log( + y); he finds

a that ydy
a+1 (a+1)c

a2
q qy dx = 0 has the integrating factor a qdx + cy . More gen-

erally r = a log(y + ) + a log(y + ) gives a new form of differential equation
integrable by the factor er .
2 Integral Transforms and Definite Integrals
The second Norwegian paper of Abel, titled Oplsning af et Par Opgaver ved Hjelp af
bestemte Integraler (1823, uvres, t. I, p. 1127), studies in its first part the integral
x=a
ds

equation a = (ax)n where is a given function, s an unknown function of x

x=0
and n < 1.
In the case where n = 12 , s is interpreted as the length of a curve to be found,
along which the fall of a massive point from the height a takes a time equal to a.
Let the curve be KC A, the initial position of the falling body be the point C and
its
initial velocity be 0; when the falling body is in M its velocity is proportional to
a x, where a is the total height AB and x is the height A P. So the fall along an
ds
infinitesimal arc MM takes a time dt proportional to ax , where s = AM is the
curvilineal abscissa along the curve, and the total duration of the fall is proportional
x=a

ds
to the integral
ax
.
x=0
Abels equation is probably the first case of an integral equation in the history of
mathematics; before that, Euler had introduced in his Institutiones Calculi Integralis
the general idea to solve a differential equation by a definite integral, for instance by
34 C. Houzel
the so called Laplace transform and Fourier (1811) and Cauchy (1817) had studied
the Fourier transform and its law of inversion.
Abel
supposes that s has a development in power series with respect to x:
s = (m) x m ; differentiating and integrating term by term, he obtains a =

a m1
a m1
1 m1
m(m) x(ax)dxn . One has m x(ax)dxn = mamn t(1t)dtn = (1n)(m+1) mn

(mn+1) a ,
0 0 0
using the Eulerian function , for which Abel refers to Legendres Exercices de
Calcul integral; so
(m + 1)
a = (1 n) (m) amn .
(m n + 1)
(k) ak ( is implicitly supposed to be analytic); by identification,

Let now a =
(k+1) (k)
1 t k dt
Abel gets (n+k) = (1n)(n+k+1) (k) = n(1n) (1t)1n
, so that
0
1

(m) m xn (k) (xt)k dt
s= x =
n (1 n) (1 t)1n
0
1 1
xn (xt)dt x n sin n (xt)dt
= 1n
=
n (1 n) (1 t) (1 t)1n
0 0

x
1 (xt)dt
and, in the particular case where n = 12 , s =

1t
.
0
n
Abel applies this result in the case where a = ca (c constant, and the expo-
nent n not to be confused with that of ax in the general problem, which is now 12 ), in
1
1 n dt
which s = Cx n+ 2 , with C = c t 1t ; then dy = ds2 dx 2 = dx kx 2n1 1,
0
2
where k = n + 12 C 2 , so

y = dx kx 2n1 1 = k + x k 1
in the particular case where n = 12 ; in this case, the curve KC A solution of the
problem is a straight line. The isochroniccase, where a = c constant is another
interesting case; here n = 0 and s = C x (C = 2c ), equation characterising the
cycloid. This problem was initially solved by Huygens (1673).
Turning back to the general case Abel gives another interpretation of the solution
as a derivative of of non-integral order n. Indeed, if x = (m) x m and if k is

a natural integer,
dk (m + 1) mk
= (m) x ;
dx k (m k + 1)
in which the right hand side is still meaningfull when k is not a natural integer, and
then
1
(m + 1) 1 t m dt
= ,
(m k + 1) (k) (1 t)1+k
0
1

1 (xt)m dt
(m)
1

1 (xt)dt
so that the right hand side becomes x k (k) (1t)k
= x k (k) (1t)k
, whence
0 0
d n xn

1 (xt)dt 1 d n
the definition of dx n
= n (1t)1n
and the solution s = (1n) dx n of the initial
0
1
problem. The derivative of order n of s = x is naturally (1n) x, which means
that
a
dn 1 xdx
n
= (n < 1);
da (1 n) (a x)n
0
d 12 s
for n = 12 , x = 1.
dx 2
The idea of a derivative of non-integral order comes from Leibniz; it was based
on the analogy, discovered by Leibniz, between the powers and the differentials in the
celebrated formula for d n (xy), which has the same coefficient as (x+ y)n = pn (x+ y)
in Leibniz notation. The general binomial formula, with exponent e non necessarily
integral, suggests to Leibniz a formula for d e (xy) as an infinite series (letter to
the Marquis de lHospital, 30 September 1695). Abels procedure is an extension

1
dx(lx)e
d n (z e ) en 0
of a formula given by Euler in 1730: dz n = z
1 , where e and n are
dx(lx)en
0
arbitrary numbers and l notes the logarithm. At Abels time, some other authors also
considered derivatives of arbitrary order, as Fourier and Cauchy, but the theory really
began with Liouville in 1832 and Riemann in 1847.
At the end of this part, Abel reports that he has solved the more general integral
equation a = (xa) fx dx, where and f are given functions and is unknown.
36 C. Houzel
Abel published a German version of this study in Crelles Journal (vol. I, 1826,
uvres, t. I, p. 97101). He finds the solution without any use of power series,

1 1
starting from the Eulerian integral of the first kind y(1y)dyn = (1n)
(+1n) , which gives
0
a z 1 dz (1n) n
(az)n = (+1n) a and
0
x a x
da z 1 dz (1 n) an da
=
(x a)1n (a z) n ( + 1 n) (x a)1n
0 0 0
x
= n (1 n) x = n (1 n).
( + 1)

x da

a f z.dz
Then, if fx = x d, one has (xa)1n (az)n = n (1 n) fx and
0 0
x a
sin n da f zdz
fx = .
(x a)1n (a z)n
0 0
x=a
ds

Therefore, in the original problem a = (ax)n , one has

x=0
x x a
sin n ada sin n da ds
= = s.
(x a)1n (x a)1n (a x)n
0 0 0
In this paper, there is no mention of derivatives of non-integral order.

The second part of the Norwegian paper is devoted to the proof of the integral
formula:
+ +
2y v2 y2 2 2
(x + y 1) + (x y 1) = e vdv (x + t)ev t dt,

+ 2
t 2 + 14 y2
1

giving as a particular case cos y =

e t dt when t = et , x = 0. Abel

uses the developments
x 2 x

(x + y 1)+(x y 1) = 2 x y + y4 . . . ,
12 1234
2
t t3
(x + t) = x + t x + x + x + . . .
12 123
and the definite integrals
+ +
2n+1

v2 t 2 2n 2 2 1 2n
e t dt = 2
, ev y v2n dv = y2n1 .
v2n+1 2

The last two parts of the paper give summation formulae by means of definite
integrals. From the development et 11 = et + e2t + e3t + . . . and the value
kt 2n1
t 2n1 dt
e t dt = (2n)
k 2n , Abel deduces et 1
= (2n)(2n); the Eulerian formula
0 0
2n1 2n
(2n) = 2(2n+1)
An , where An is the n-th Bernoulli number, then gives An =
2n1
t 2n1 dt
2n t dt 2n

22n1 2n et 1
= 22n1 et 1
. Using these values in the EulerMacLaurin sum
0 0
x
x = x.dx 12 x + A1 12x A2 1234

formula + . . . and Taylor series for

x 2t 1 , Abel finds

x + 2t 1 x 2t 1

1

dt
x = xdx x + t 1
. (24)
2 2 1 e
0
This formula was already published in 1820 by Plana in the Memoirs of the Turin
Academy; Plana found it by the same type of formal manipulations as Abel. It was
rigorously established by Schaar in 1848, using Cauchys calculus of residues.
As particular applications of this formula, Abel gives the values of some definite

mtdt
integrals: for x = emx , sin
et 1
= em11 m + 12 , for x = 1x ,
0

tdt 1 1 tdt
1 2
= 2 log x 2 +3+ 1 2

2 t
x + 4 t (e 1) x x 1 + 4 t (et 1)
0 0

eat eat 1
and for x = sin ax, et 1
dt = a cot a.
0
The second Abels summation formula is
(x + 1) (x + 2) + (x + 3) (x + 4) + . . . (25)

1 dt (x + t 1) (x t 1)
= x + 2 .
2 et et 2 1
0
In order to obtain this, Abel puts a priori the first member equal to 12 x

+ A1 x +
cx 1
A2 x + . . . with unknown coefficients A1 , A2 , . . . ; when x = e , one sees
that A2 = A4 = . . . = 0 and 12 tan 12 c = A1 c A3 c3 + A5 c5 . . . . On the other

ct ct
hand, 12 tan 12 c = eet e
et
dt after Legendre (Exercices de Calcul Integral, t. II,
0
p. 186), so the series for ect ect gives the A2n+1 in the form of integrals and the
38 C. Houzel
1
reasoning ends as for the first formula. As an application, Abel takes x = x+1 and

gets (1+t 2 )(etdtt et ) = 12 log 2 14 .

0
In the second volume of Magasinet for Naturvidenskaberne (1825), Abel pub-
lished another derivation of the formula (24) and he extended it to the case of iterated
sums (uvres, t. I, p. 3439):
n
n n1
x = An1,n n x dx n An2,n (n 1) x dx n1
1
+ . . . + (1)n1 x dx + (1)n x
2

n1 Pdt (x + t 1) (x t 1)
+2(1)
e2t 1 2 1
0

n1 Qdt (x + t 1) + (x t 1)
+2(1)
e2t 1 2
0
where the coefficients A0,n , A1,n , . . . , An1,n are defined by the development of pn =
2 n1
1
(ev 1)n in the form (1)
n1
A0,n p + A1,n ddvp + A2,n ddv2p + . . . + An1,n ddvn1p and
P = A0,n A2,n t 2 + A4,n t 4 . . . , Q = A1,n t A3,n t 3 + A5,n t 5 . . . ; by derivating
pn , Abel establishes recursive relations between the Ak,n : A0,n+1 A0,n = 0,
A1,n+1 A1,n = n1 A0,n , A2,n+1 A2,n = n1 A1,n , . . . , An1,n+1 An1,n = n1 An2,n ,
An,n+1 = n1 An1,n . The proof of this formula is based on the expression of as
a Laplace transform: x = evx fv dv, which naturally restricts the generality; it
gives
n
fv
x = evx dv.
(ev 1)n
As an example, for x = eax and n = 2, this formula gives:

1 1 1 1 dt sin at tdt cos at
= + 2 2 2 .
(ea 1)2 2 a a e2t 1 e2t 1
0 0
1
Another example, with x = x2
and n = 1, leads to

1 1 1 1 1 tdt
2
+ 2
+ 2
+ . . . = 2 + + 4a ;
a (a + 1) (a + 2) 2a a (e2t 1)(a2 + t 2 )2
0
2 3

tdt
in particular, for a = 1, 6 = 2 +4 (e2t 1)(1+t 2 )2
.
0
A posthumous paper of Abel is devoted to the study of the Laplace transform; its
title is Sur les fonctions generatrices et leurs determinantes (uvres, t. II, p. 6781,
mem. XI) and the study is purely formal. Abel writes an arbitrary function of
several variables in the form:
(x, y, z, . . . ) = exu+yv+z p+... f(u, v, p, . . . )dudvd p . . . ,
and he calls the generating function of f and f the determinant function of ,

in abbreviation = fg f and f = D . The following properties of the transform
are established in the case of one variable only: linearity, effect of a translation
D(x + a) = eavDx and fg(eav Dx) = (x + a), effect of derivations or inte-
n n
n
grations D ddxx = vn Dx, fg(vn Dx) = ddxx

n n , D xdx n = vn Dx and

n
fg(vn Dx) = xdx n , effect of finite differences or iterated sums
Dn x = (ev 1)n fv, fg((ev 1)n fv) = n x, Dn (x) = (eva 1)n fv

and fg((ev 1)n fv) = n (x).
Abel also states the effect of the composition of a translation, a derivation and
a certain number of finite differences. More generally, if the operator is defined by

d n (x + ) d n (x + )
(x) = An, + A n , + ... , (26)
dx n dx n
where An, , An , , . . . are constant coefficients, one has D(x) = v Dx where

v = An, vn ev + An , vn ev + . . . ,
and Abel considers the composition of an arbitrary number of operators of the type
of .
Abel clearly understood how the Laplace transform gives a symbolic calculus
on the operators (26); he uses this calculus to obtain developments in series. For
instance, he explains that the Taylor series for (x + ) amounts to the development
v2 2 v3
ev = 1+v+ 12 + 123 3 +. . . in the determinant function. A polynomial relation
between the multiplicators , 1 , . . . , associated to operators , 1 , . . . , gives
an analogous relation between the operators themselves. Let us consider the operator
x = (x + ) + ax; one has Dx = (ev + a) fv where f is the determinant
function of . Since
n(n 1) n2 2v
(a + ev )n = an + nan1 ev + a e + ...
2
n(n 1) 2 (n2)v
= env + nae(n1)v + a e + ... ,
2
n(n 1) n2
n x = an x + nan1 (x + ) + a (x + 2) + . . .
2
n(n1) 2
= (x +n)+na(x +(n1))+ a (x +(n2))+. . . ;
2
40 C. Houzel
Abel writes down both forms, which are the same for n a natural integer, but which
may be extended (under different conditions for the convergence of the series) to
other values of n; he says nothing about that, but he may have envisaged this type of
extension as we saw that he was interested by derivatives of non-integral order and we
know that Cauchy defined pseudo-differential operators with constant coefficients us-
ing the Fourier transform (1827). For a = 1, n x = (x +n)n(x +(n 1))
+ n(n1)
2 (x + (n 2)) . . .
Now let 1 x = (x + 1 ) + a1 x, so that D1n x = yn fv with y = ev1 + a1 ;
1
if z =ev + a then y = a1 + (z a) and it is possible to get a development
yn = Am z m . Therefore 1n x = Am m x. In the case where 1 = ,
n(n 1)
1n x = (a1 a)n x + n(a1 a)n1 x + (a1 a)n2 2 x + . . .
2
n(n 1)
= n x + n(a1 a)n1 x + (a1 a)2 n2 x + . . .
2
For a1 = 0, (x + n) = n x nan1 x + n(n1) 2 a
2 n2
x + . . . and if moreover
n n1 n(n1) n2
a = 1, (x + n) = x + n x + 2 x + . . . , a formula given
by Euler (1755).
When x = (x +)ax and 1 x = cx +k dx v
dx , Dx = (e a) fv = z fv
and D1 x = (c + kv) fv = y fv; as y = c + log(z + a) = c + k log a +
n n n k

k z 1 z2 1 z3 n
Am z m , which

a 2 a2 + 3 a3 . . . , one may write a development y =
gives 1n x = Am m x. For example, if c = 0, a = k = 1 and n = 1, dx

dx =
1 1 2 1 3

x 2 x + 3 x . . . , a formula given by Lagrange (1772). Starting
from a formula of Legendre:
(vcv )2 (vcv )3
bv = 1 + lb vcv + lb(lb 2lc) + lb(lb 3lc)2 + ... ,
2 23
in which he makes b = e and c = e , Abel obtains in the same way
d(x + ) ( 2) d 2 (x + 2)
(x + ) = x + + (27)
dx 2 dx 2
2 3
( 3) d (x + 3)
+ + ...
23 dx 3
2
and, in particular, x = (0) + x () + x(x2)
2 (2) + x(x3)
23 (3) + . . .
m
Abel published the special case of (27) in which x = x , m a natural integer, in
the first issue of Crelles Journal (uvres, t. I, p. 102103); there, he proves the
formula by induction on m and he observes that, when = 0, the result reduces to
the binomial formula. Another special case given in the posthumous memoir is that
in which x = log x; then
3 2

1 2 1
log(x +) = log x + + + +. . .
x + 2 x +2 x +2 3 x +3 x +3
2

and in particular log(1+) = 1+ + 12 1+2

1 1+
1+3 + 13 1+3

1 1+
1+3 +. . .
2 2 3 1 223 4
When = 3, this reduces to log(1 + 2) = 1+ + 3 (1+3)3
+ 4 (1+4)4
+ ...;
2 1 n2 n1
for example, log 3 = 1 + n n+1 n+1 .
n3
n n x
Abel also considers the developments of n x, ddxx
n and
d (e x)
dx n in power series
with respect to n; they are respectively obtained from the developments of
(ev 1)n = exp(n log(ev 1)), vn = en log v , and (1 + v)n = en log(1+v) .
The coefficients respectively contain the powers of log(ev 1), log v and log(1+v),
so we must identify the operators respectively defined by
x = fg(log(ev 1) fv), log v fv and (1 + v) fv;
these operators are respectively x = x + d x, 1 x 12 12 x + 13 13 x

. . . , where 1 x = x x, and x 12 x + 13 x . . .
In the continuation of the paper, Abel expresses this last operator in the integral
form
0
et dt
x = ((x t) x),
t

which is obtained in the following manner: the equality
a
1
e(1v)t dt = (ea eav ea ea v )
1 v
a
= ea (eav + veav + 2 v2 eav + . . . )

ea (ea v + vea v + 2 v2 ea v + . . . )
leads to
a
et (x t)dt =
a
ea ((x a) + (x a) + 2 (x a) + 3 (x a) + . . . )

ea ((x a ) + (x a ) + 2 (x a ) + 3 (x a ) + . . . ),

0
from which Abel deduces x x + 2 x 3 x + . . . = et (x + t)dt

and, integrating with respect to , x 12 2 x + 13 3 x 14 4 x + ... =

0 et dt
t ((x + t) x).
42 C. Houzel
Other classical relations between functions give Abel relations between opera-
tors. So the Fourier series 12 = cos v cos 2v + cos 3v . . . leads to
x = (x + ) + (x ) (x + 2) (x 2) + (x + 3) + (x 3)
(x + 4) (x 4) + . . .
1

dtsin(vt)
and the formula (ev 1)1 (v)1 + 2 =2 e2t 1
leads to
0

1 1

dt (x + t 1) (x t 1)
x xdx + x = 2 ,
2 e2t 1 2 1
0

dtcos(vt) v
which is formula (24). From the formula 1+t 2
= 2e given by Legendre
0
(Exercices de Calcul integral, t. II, p. 176), Abel deduces

dt (x + t 1) + (x t 1)
= (x ), (28)
1 + t2 2 2
0

dt 1 1
for instance (1+t 2 )(2 t 2 +x 2 )
= 2 x(x) for x = x (where it is easy to verify that
0
must be taken as +); when x = x1n , (x+t 1)+(xt
2
1)
= z n cos n, where

dt cos(n arctan tx ) 1
z = x 2 + 2 t 2 and = arctan tx , so that 1+t 2 n = 2 (x+)n or
0 (x 2 +2 t 2 ) 2

2
x n1 (cos )n cos nd
= , (29)
2 (x + )n (x sin )2 + ( cos )2
0
which reduces to

2

= (cos )n cos nd (30)
2n+1
0
when = x.

dtsin at

tdtsin at a
From the integrals t(1+t 2 )
= 2 (1 ea ) and 1+t 2
= 2e also given
0 0

dt (x+t 1)(xt 1)
by Legendre, Abel deduces 2 (x (x )) = t(1+t 2 )

2 1
,
0

dt
tdt (x+t 1)(xt 1) 1)(xt 1)
2 (x) = 1+t 2
2 1
and
2 x = t (x+t
2 1
;
0 0

2
1

d n1
for x = xn this gives sin (cos ) sin n = 2 by putting t = x tan .
0
In an addition to this paper (Sur quelques integrales definies, uvres, t. II,

xn
p. 8286, mem. XII), Abel develops (cos )n cos n and (x+) n in power series with
respect to n and, comparing the coefficients of the powers of n in (29), he gets the
values of some definite integrals:

2 2
1 d 1 x log cos d
= 2 2 2 2
, log = ,
2 x x sin + cos 2 x x + x2 sin2 + 2 cos2
0 0

2 2
((log cos )2 2 )d

1 x
log = .
2 x x+ x 2 sin2 + 2 cos2
0
t
Putting x = (log x)n and x = tan in (28), he gets
n n
2
d log cosx + 1 + log cosx 1

x 2 sin2 + 2 cos2 2
0

= (log(x + ))n
2x

2 n n
and d log cosx + 1 + log cosx 1 = (log 2)n when x =
0
= 1.
More generally, putting t = tan u in (28) we get

2

du((x + 1 tan u) + (x 1 tan u))
0

2
= (x + ) and du((1 + 1 tan u) + (1 1 tan u)) = (2) when
0
xm
x = = 1; for x = 1+x n , this gives

2
(cos u)nm (cos mu(cos u)n + cos(n m)u) 2m
du = .
(cos u)2n + 2 cos nu(cos u)n + 2 2 1 + 2n
0
m
In (30) Abel replaces n by a fraction n and he puts n = , so that

cos 2n
2
1

m
n dy
m = (cos n) cos md =
n (y)m fy ,
2n 2 n 1 y2
0 1
44 C. Houzel
where y = yn n(n1)2 yn2 (1 y2 ) + n(n1)(n2)(n3)

234 yn4 (1 y2 )2 . . . and
fy = ym m(m1)
2 ym2 (1 y2 )+ m(m1)(m2)(m3)
234 ym4 (1 y2 )2 . . . ; for instance

cos 8
sin 8

1 2 + 8y4 ydy dz 4 1 8z 2 + 8z 4 .

4
8
4 = 1 8y 2
=
2 1y
1 0
All this early work of Abel gives evidence of his carefull study of Legendres
Exercices de Calcul integral, which also were his source of inspiration for the theory
of elliptic integrals.
A paper on a related subject was published by Abel in the second volume
of Crelles Journal (1827, uvres, t. I, p. 251262) under the title Sur quelques
integrales definies. It contains

some applications of the relation discovered by Abel
y2 dy 1 dy2
da y1 da = e
pda
, where y1 and y2 are two solutions of the linear dif-
2
1 (x+a) +1 dx
ferential equation ddau2 + p dy da + qy = 0. For instance y 1 = x 1 (1x)1
and
0

1 (x+a)++ dx
y2 = x (1x)
are solutions of the hypergeometric equation
0
d2 y

+ + dy ( + 1)( + + )
+ + y = 0,
da2 a 1+a da a(a + 1)
and this leads to the relation
1 1
dx(x + a) +1 dx(x + a)++ 1
( + + )
x 1 (1 x)1 x (1 x)
0 0
1 1
dx(x + a) dx(x + a)++
( + 1) = Ca+ (1 + a)+ ,
x 1 (1 x)1 x (1 x)
0 0
where the constant C is determined by making a = :
1 1
C = ( + 1) dx x 1
(1 x) 1
dx x (1 x)
0 0
= (cot + cot ).
x dx 1 dx
In the same way y1 = and y2 = (1+x)1x(x+a)
(1+x) (x+a) ++ 1 are so-
0 0
2
lutions of the hypergeometric equation dda2y + + + dy
a 1a da +
(1)
a(1a) y=0

x dx (1)(++ 1)
x 1 dx

and we get the relation (1+x) (x+a) = (1+x)1 (x+a)++ 1

.
0 0

x dx +1

x dx
The function y3 = (1+x) (x+1a) = (1 a) (1+x) (1+(1a)x) is a third
0 0
solution of the same hypergeometric equation, which, combined with y1 =

dx
a +1 (1+x)x (1+ax) , gives
0

x dx x dx
a
(1 + x) (1 + ax) (1 + x) +1 (1 + (1 a)x)
0 0

x dx x dx
+(1 a)
(1 + x) (1 + (1 a)x)
(1 + x) +1 (1 + ax)
0 0
(1 )(1 )
= ( + + 1).
( + 1)

x dx

x 1 dx
When = 1, this relation becomes a (1+x) (1+ax)1
(1+x) +1 (1+(1a)x)
+
0 0

x dx

x 1 dx
(1a) (1+x) +1 (1+ax)1
(1+x) (1+(1a)x) =
sin ; in particular, for = = 12 ,
0 0

dx dx
a
x(1 + x)(1 + ax) x(1 + x)3 (1 + (1 a)x)
0 0

dx dx
+(1 a) . = 2.
x(1 + x)(1 + (1 a)x) x(1 + x)3 (1 + ax)
0 0
As Abel observes, these integrals are elliptic and the change of variable x = tan2
transforms the preceding relation in

2 2
d d cos2
a
1 (1 a) sin2 1 a sin2
0 0

2 2
d cos2

d
+(1 a) = ,
1 a sin2 1 (1 a) sin2 2
0 0
which is equivalent to Legendres famous relation between the complete integrals of

the first two kinds (Exercices de Calcul integral, t. I, p. 61). Legendre had proved
this relation by a very similar method.

x 1 (1x)1 dy

Starting from the integral y = dxx(x+a) + , Abel finds that da +
1+a + a y
0
x (1x)

a daa1 (1+a)1
= a(1+a)(x+a) + , so that y a (1 + a) = C x (1 x) (a+x)+
, where
0

C is independent of a and is found to be (+) by making a = . Thus
46 C. Houzel
x
dx x 1 (1 x)1
= a (1 + a)
( + ) (x + a)+
0
a
da a1 (1 + a)1
+x (1 x) ;
(a + x)+
0
when + = 1, this gives

x a
(1 + a) dx x 1 (1 x) x da a (1 + a)1
+ = .
a1 x +a (1 x)1 a+x sin
0 0

1
eax x 1 (1 x)1 dx (, > 0) is a solution of the conflu-
The integral y =
0
2

ent hypergeometric equation dda2y + +a + 1 dy
da + a y = 0, and so is

y1 = eax x 1 (1 x)1 dx
1

= ea eax x 1 (1 + x)1 dx = ea a+1 ex x 1 (a + x)1 dx
0 0
(a > 0). Abel derives from that the formula

1
ax 1
= e x (1 x) 1
dx ex x 1 (a + x) dx
0 0
1
ax
a e x (1 x) 1
dx ex x 1 (a + x)1 dx,
0 0
and, for = 1 ,
1
dx ax x a
= e ex dx 1 +
sin x 1x x
0 0
1
x dx x a
a dx eax e 1+ .
1x x+a x
0 0

2
As a last example, Abel considers the integrals y = eaxx x 1 dx and y1 =
0
axx 2 1 d2 y
e x dx ( > 0), solutions of the differential equation da2
12 a dy 1
da 2 y
0
= 0, which is related to the so called Weber equation. The corresponding relation is
2
1 +1 a axx 2 1 2
e4 = e x dx eaxx x dx
2 2 2
0 0

2 2
+ eaxx x dx eaxx x 1 dx.
0 0
In a posthumous paper Les fonctions transcendantes a12 , a13 , a14 , . . . , a1n

exprimees par des integrales definies (uvres, t. II, p. 16), Abel gives integral
formulae for these finite sums, extended from 1 toa 1. He also studies their
dn 1 1
continuation to non integral values of a and n. As dan a = (1)n 2 3 n an+1
,
one has
1
n1 d n1 L(a)
= (1)
an 2 3 (n 1)dan1
1
1 x a1 1
where L(a) = a = x1 dx. From this Abel deduces
0
1
1 1 1
1 a1
1 x
L(a, ) = = l dx (31)
a () x1 x
0
2 1 2 3 1 3
for any value of . Substituting x a1 = 1(a1) l 1x + (a1) l x (a1)

2 23 lx +
1
. . . and 1x = 1 + x + x 2 + . . . , he obtains

a1 1 1 1
L(a, ) = 1 + +1 + +1 + +1 + . . .
1 2 3 4
2
(a 1) 1 1 1
( + 1) 1 + +2 + +2 + +2 + . . .
12 2 3 4
3
(a 1) 1 1 1
+ ( + 1)( + 2) 1 + +3 + +3 + +3 + . . .
123 2 3 4
( + 1)
= (a 1)L ( + 1) (a 1)2 L ( + 2)
2
( + 1)( + 2)
+ (a 1)3 L ( + 3) . . . ,
23
1 1 1
where L () = L(, ) = 1 + 2 + . . . is the zeta function.
+ 3 + 4
1
m
m 1

1 l 1y
Putting a instead of a in (31), Abel deduces L a , = () y1 dy +
0
1
a

1 ym1 l 1y ym1 A A
() ya 1 dy or, writing ya 1 = 1cy + 1c y + . . . in the hypothesis
0
m 1 < a,
48 C. Houzel
m
L , = a (AL (, c) + A L (, c ) + A L (, c ) + . . . )
a
c c2 c3
where L (, c) = 1 + 2 + 3 + 4 + ...

1 1
The following paper Sur lintegrale definie x a1 (1 x)c1 l 1x dx (uvres,
0
t. II, p. 713) is related to the same subject; it gives developments in series for
ac
La and associated functions. When = 1, the integral is equal to (a+c) . As
the logarithmic derivative of a is equal to La C, where C is the Euler con-

1 ac
stant, Abel deduces from this x a1 (1 x)c1lxdx = (La L(a + c)) (a+c) ,
0
1 a1 c1 ac
x (1 x) l(1 x)dx = (Lc L(a + c)) (a+c) . For c = 1, this gives
0
1
1
x a1lxdx = a12 , x a1l(1 x)dx = L(1+a)
a . Developing (1 x)c1 in se-
0 0
ries, Abel obtains
a c 1 1 (c 1)(c 2) 1
(La L(a + c)) = 2 (c 1) 2
+
(a + c) a (a + 1) 2 (a + 2)2
(c 1)(c 2)(c 3) 1
+ ... (32)
23 (a + 3)2
1 a a(a+1) a(a+1)(a+2)
For example, if c = 1 a, La sin a = a2 + (a+1)2 + 2(a+2)2 + 23(a+3)2 + . . . ,
2
which becomes 2 log 2 = 22 + 32
+ 253 2 + 2235
372
+ 23357
3492
+ . . . when a = 12 for
L 12 = 2 log 2. When a = 1 x and c = 2x 1, L(1 x) L x = cot x and
(1 x)(2x 1) 1 2x 2 (2x 2)(2x 3)

cot x = 2
2
+
x (1 x) (2 x) 2(3 x)2
(2x 2)(2x 3)(2x 4)
+ ...
2 3(4 x)2
L(a+c)La
From (32) Abel deduces an expression of L(a+c)Lc as a quotient of two series and,
a(a1) a(a1)(a2)
making c = 1, L(1 + a) = a 22
+ 232
. . . . Thus
cot a = L(1 a) La

a(a + 1) (a 1)(a 2)
= 2a 1 +
22

a(a + 1)(a + 2) (a 1)(a 2)(a 3)
+ + ... .
2 32
The integral of the title, with an integer, is obtained by successive differentiations
with respect to a:
1 a1 1
(c1)(c2)

x (1 x)c1 l 1x dx = a1 c1 1
1 (a+1) + 12
1
(a+2) . . . .
0
Taking the successive logarithmic derivatives,

1 Abel sees
that this integral has an
1
expression in terms of the sums La, L a = a2
, L
a = a3
, . . . ; for example
1 3
a1 c1 1
x (1 x) l dx = 2(L (a + c) L a)
x
0
+3(L (a + c) L a)(L(a + c) La)

3 a c
+(L(a + c) La) .
(a + c)

1 1 1
The successive differentiations of the equality lx dx = with respect to
0

1 1 1 n d n

1 n
give the formula l 1x ll x dx = dn , whence (lz)n ez dz = n+1 dd
n
0 0

ex dx = 1
1
by a change of variable. Abel deduces from this the formulae
0

1
ex dx = 12
1 1
(n = 0) and l x L C (n = 1; C is the Euler
0
constant), which leads to

enx x 1lxdx = (L C log n).
n
0
A third posthumous paper is titled Sommation de la serie y = (0) + (1)x +

(2)x 2 + (3)x 3 + . . . + (n)x n , n etant un entier positif fini ou infini, et (n) une
fonction algebrique rationnelle de n (uvres, t. II, p. 1418). Abel decomposes
B
in terms of one of the forms An , (a+n) . He has first to sum f(, x) = x +
xd f(1,x)
2 x 2 + 3 x 3 + . . . + n x n ; this is done using the identities f(, x) = dx and
n)
f(0, x) = x(1x1x . Then Abel considers
x2

1 x xn dx x 1 F( 1)
F = + + + . . . + = ,
a (a + 1) (a + 2) (a + n) x
n+1
for which F(0) = 1x
1x . The formula (10) for the dilogarithm is thus obtained
when = 2, n = and a = 1.
3 Algebraic Equations
We know that in 1821 Abel thought he had found a method to solve the general quintic
equation by radicals; when he discovered his error and proved that such a solution
was impossible, he wrote a booklet in french with a demonstration, Memoire sur les
50 C. Houzel
equations algebriques, ou lon demontre limpossibilite de la resolution de lequation

generale du cinquieme degre (Christiania, 1824; uvres, t. I, p. 2833).
The impossibility of an algebraic solution for the general quintic equation had
already been published by P. Ruffini (1799, 1802, 1813), but his demonstration was
incomplete for he supposed without proof that the radicals in a hypothetical solution
were necessarily rational functions of the roots. Abel, who did not know of Ruffinis
work, began with a proof of this fact.
Supposing the root of
y5 ay4 + by3 cy2 + dy e = 0 (33)
of the form
1 2 m1
y = p + p1 R m + p2 R m + . . . + pm1 R m , (34)
1
with m a prime number and p, p1 , . . . , pm1 , R of an analogous form (R m is
a chosen exterior radical in a hypothetical solution by radicals and it is supposed
that it is not a rational function of a, b, . . . , p, p1 , . . . ), Abel first replaces R by pRm
1
in order to have an expression of the same form with p1 = 1. Putting (34) in the
1 2 m1
equation, he gets a relation P = q + q1 R m + q2 R m + . . . + qm1 R m = 0, with
coefficients q, q1 , . . . polynomial in a, b, c, d, e, p, p2 , . . . , R. These coefficients
are necessarily 0 for otherwise the two equations z m R = 0 and q + q1 z + . . . +
qm1 z m1 = 0 would have some common roots, given by the annulation of the
greatest common divisor
r + r1 z + . . . + rk z k
of their first members. Since the roots of z m R = 0 are of the form z, where
z is one of them and is an m-th root of 1, we get a system of k equations
r + r1 z + . . . + k rk z k = 0 (0 k 1 and 0 = 1), from which it is possible
to express z as a rational function of r, r1 , . . . (and the ). Now the rk are rational
with respect to a, b, . . . , R, p, p2 , . . . and we get a contradiction for, by hypothesis,
z is not rational with respect to these quantities.
The relation P = 0 being identical, the expression (34) is still a root of (33)
1 1
when R m is replaced by R m , an arbitrary m-root of 1, and it is easy to see that
the m expressions so obtained are distinct; it results that m 5. Then (34) gives us
1 1 1 1
m roots yk (1 k m) of (33), with R m , R m , . . . , m1 R m in place of R m , and
we have
1
p = (y1 + y2 + . . . + ym ),
m
1 1
R m = (y1 + m1 y2 + . . . + ym ),
m
2 1
p2 R m = (y1 + m2 y2 + . . . + 2 ym ),
m
... ,
m1 1
pm1 R m = (y1 + y2 + . . . + m1 ym );
m
1
this proves that p, p2 , . . . , pm1 and R m are rational functions of the roots of (33)
1 2 n1
(and ). Now if, for instance R = S + v n + S2 v n + . . . + Sn1 v n , the same
1
reasoning shows that v n , S, S2 , . . . are rational functions of the roots of (33) and
continuing in this manner, we see that every irrational quantity in (34) is a rational
function of the roots of (33) (and some roots of 1).
Abel next shows that the innermost radicals in (34) must be of index 2. Indeed
1
if R m = r is such a radical, r is a rational function of the 5 roots y1 , y2 , . . . , y5 and
R is a symmetric rational function of the same roots, which may be considered as
independent variables for (33) is the general quintic equation. So we may arbitrarily
1
permute the yk in the relation R m = r and we see that r takes m different values;
a result of Cauchy (1815) now says that m = 5 or 2 and the value 5 is easily
excluded. We thus know that r takes 2 values and, following Cauchy, it has the form
1
v(y1 y2 )(y1 y3 ) . . . (y2 y3 ) . . . (y4 y5 ) = vS 2 , where v is symmetric and S
is the discriminant of (33).
1 m
1
The next radicals are of the form r = p + p1 S 2 , with p, p1 symmetric. If
1
1 m
1
1 m
r1 = p p1 S 2 is the conjugate of r, then rr1 = p2 p21 S 2 = v must be
symmetric (otherwise m would be equal to 2 and r would take on 4 values, which is
not possible). Thus
1
1 m
1
1 m
r + r1 = p + p1 S 2 + v p + p1 S 2 =z
takes m values which implies that m = 5 and z = q + q1 y + q2 y2 + q3 y3 + q4 y4 ,

with q, q1 , . . . symmetric. Combining this relation with (33), we get y rationally in
z, a, b, c, d and e, and so of the form
1 2 3 4
y = P + R 5 + P2 R 5 + P3 R 5 + P4 R 5 , (35)
1
with P, R, P2 , P3 and P4 of the form p+ p1 S 2 , p, p1 and S rational in a, b, c, d and e.
1
1
1 5
From (35) Abel draws R 5 = 15 (y1 + 4 y2 + 3 y3 + 2 y4 + y5 ) = p + p1 S 2 ,
where is an imaginary fifth root of 1; this is impossible for the first expression
takes 120 values and the second only 10.
Euler (1764) had conjectured a form analogous to (35) for the solutions of the
quintic equation, with R given by an equation of degree 4. In a letter to Holmboe
(24 October 1826, uvres, t. II, p. 260), Abel states that if aquinticequation is
algebraically solvable, its solution has the form x = A + 5 R + 5 R + 5 R + 5 R
where R, R , R , R are roots of an equation of degree 4 solvable by quadratic
radicals; this is explained in a letter to Crelle (14 March 1826, uvres, t. II, p. 266)
for the case of a solvable quintic equation with rational coefficients, the solution
1 2 4 3 1 2 4 3 1 2 4 3 1 2 4 3
being x = c+ Aa 5 a15 a25 a35 + A1 a15 a25 a35 a 5 + A2 a25 a35 a 5 a15 + A3 a35 a 5 a15 a25 , where
52 C. Houzel

a = m + n 1 + e2 + h(1 + e2 + 1 + e2 ),

a1 = m n 1 + e2 + h(1 + e2 1 + e2 ),

a2 = m + n 1 + e h(1 + e2 + 1 + e2 ),
2

a3 = m n 1 + e2 h(1 + e2 1 + e2 ),
A = K + K a+ K a2 + K aa2 , A1 = K + K a1 + K a3 + K a1 a3 , A2 = K + K a2 +
K a + K aa2 and A3 = K + K a3 + K a1 + K a1 a3 , and c, h, e, m, n, K, K , K
and K are rational numbers.
Abel published a new version of his theorem in the first volume of Crelles
Journal (1826, uvres, t. I, p. 6687). In the first paragraph of this paper, Abel
defines the algebraic functions of a set of variables x , x , x , . . . They are built
from these variables and some constant quantities by the operations of addition,
multiplication, division and extraction of roots of prime index. Such a function
is integral when only addition and multiplication are used, and is then a sum of
monomials Ax m 1 x m 2 . . . It is rational when division is also used, but not the
extraction of roots, and is then a quotient of two integral functions. The general
algebraic functions are classified in orders, according to the number of superposed

radicals in their expression; a function f(r , r , . . . , n p , n p , . . . ) of order ,

with r , r , . . . , p , p , . . . of order < and f rational, such that none of the nk pk

is a rational function of the r and the other n p , is said to be of degree m if it

contains m radicals nk pk . Such a function may be written f(r , r , . . . , n p) with

p of order 1, r , r , . . . of order and degree m 1, and f rational;
1 2 n1
it is then easy to reduce it to the form q0 + q1 p n + q2 p n + . . . + qn1 p n , with
coefficients q0 , q1 , q2 , . . . rational functions of p, r , r , . . . , so of order and
1
degree m 1, p n not a rational function of these quantities. Abel carries out the
supplementary reduction to the case q1 = 1. In order to do this, he chooses an index
such that q = 0 and puts qn p = p1 , which will play the role of p. The starting
point of his preceding paper has been completely justified.
In the second paragraph, Abel proves that if an equation is algebraically solvable,
one may write its solution in a form in which all the constituent algebraic expressions
are rational functions of the roots of the equation. The proof is more precise than
that of the 1824 paper, but follows the same lines. The coefficients of the equation
are supposed to be rational functions of certain independent variables x , x , x , . . .
In the third paragraph, Abel reproduces the proof of Lagranges theorem (1771)
according to which the number of values that a rational function v of n letters may
take under the n! substitutions of these letters is necessarily a divisor of n! and
Cauchys theorem (1815) which says that if p is the greatest prime number n, and
if v takes less than p values, then it takes 1 or 2 values. Indeed it must be invariant
for any cycle of p letters, and it is possible to deduce from this that it is invariant for
any cycle of 3 letters and from this by any even substitution. Thus, as Ruffini had
proved, a rational function of 5 variable cannot take 3 or 4 values. Abel then gives,
following Cauchy, the form of a function v of 5 letters x1 , x2 , . . . , x5 which takes 2
values: it may be written p + q, where p and q are rational symmetric functions

and
= (x1 x2 )(x1 x3 ) . . . (x4 x5 )
is the square root of the discriminant. Indeed if v1 and v2 are the two values of v,
v1 + v2 = t and (v1 v2 ) = t1 are symmetric and v1 = 12 t + 2t12 . Finally, Abel gives
the form of a function of 5 quantities which takes 5 values: it is r0 + r1 x + r2 x 2 +
r3 x 3 + r4 x 4 , where r0 , r1 , r2 , r3 and r4 are symmetric functions of the five quantities
and x is one of them. Indeed this is true for a rational function of x1 , x2 , x3 , x4 , x5
which is symmetric with respect to x2 , x3 , x4 , x5 . Now if v is a function which takes
5 values v1 , v2 , v3 , v4 , v5 under the substitutions of x1 , x2 , x3 , x4 , x5 , the number
of values of x1m v under the substitutions of x2 , x3 , x4 , x5 is less than 5, otherwise it
would give 25 values under the substitutions of x1 , x2 , x3 , x4 , x5 and 25 does not
divide 5!. If = 1, v is symmetric with respect to x2 , x3 , x4 , x5 and the result is
true; it is also true if = 4 for the sum v1 + v2 + v3 + v4 + v5 is completely
symmetric and v1 + v2 + v3 + v4 is symmetric with respect to x2 , x3 , x4 , x5 , so
v5 = v1 + v2 + v3 + v4 + v5 (v1 + v2 + v3 + v4 ) is of the desired form. It is
somewhat more work to prove that cannot be 2 or 3. Eliminating x between the
equations
(x x1 )(x x2 )(x x3 )(x x4 )(x x5 ) = x 5 ax 4 + bx 3 cx 2 + dx e = 0
and r0 + r1 x + r2 x 2 + r3 x 3 + r4 x 4 = v (a quantity taking 5 values), one obtains
x = s0 + s1 v + s2 v2 + s3 v3 + s4 v4
where s0 , s1 , s2 , s3 and s4 are symmetric functions. The paragraph ends with the
following lemma: if a rational function v of the 5 roots takes m values under the
substitutions of these roots, it is a root of an equation of degree m with coeffficients
rational symmetric and it cannot be a root of such an equation of degree less than m.
The fourth paragraph finally gives the proof of the impossibility of a solution by
1
radicals. As in the preceding paper, Abel proves that an innermost radical R m = v
in a hypothetical solution has an index m (supposed prime) equal to 2 or 5; if m = 5,
one may write
1 2 3 4
x = s0 + s1 R 5 + s2 R 5 + s3 R 5 + s4 R 5
1 1
and s1 R 5 = (x1 + 4 x2 + 3 x3 + 2 x4 + x5 )
5
where is a fifth root of 1, and the second member takes 120 values, which is
impossible for it is a root of the equation z 5 s15 R = 0. So m = 2 and R = p + qs
with p, q symmetric and s = (x1 x2 ) (x4 x5 ); the second value is R=

5
1
p qs, so p = 0. Then, at the second order appear radicals + s2 = R 5 with

, symmetric as well as = 5 2 2 s2 ; p = 5 R + 5 takes 5 values so that
R
54 C. Houzel
1 2 3 4
x = s0 + s1 p + s2 p2 + s3 p3 + s4 p4 = t0 + t1 R 5 + t2 R 5 + t3 R 5 + t4 R 5
with t0 , t1 , t2 , t3 and t4 rational in a, b, c, d, e and R. From this relation, one deduces

11
(x1 + 4 x2 + 3 x3 + 2 x4 + x5 ) = p ,
t1 R 5 =
5

where is a fifth root of 1; p 5 = t15 R = u + u s2 and ( p 5 u)2 = u 2 s2 , an
equation of degree 10 in p , whereas p takes 120 values, a contradiction.
Abel reproduced this demonstration in the Bulletin de Ferussac (1826, t. 6,
uvres, t. I, p. 8794).
In a short paper published in the Annales de Gergonne (1827, t. XVII, uvres,
t. I, p. 212218), Abel treated a problem of the theory of elimination: given two
algebraic equations
y = p0 + p1 y + p2 y2 + . . . + pm1 ym1 + ym = 0
and y = q0 + q1 y + q2 y2 + . . . + qn1 yn1 + yn
with exactly one common solution y, compute any rational function fy of this
solution rationally as a function of p0 , p1 , . . . , pm1 , q0 , q1 , . . . , qn1 . He denotes
the roots of by y, y1 , . . . , yn1 and the product of the
y j with j = k by Rk
f yk yk Rk
(y0 = y). As y = 0, Rk = 0 for k 1 so that f y = y k Rk
, where is any
Fy
rational function. If f y = y , with F and polynomial, one may take = to get

fy = Fyk Rk .
yk Rk
Abel proposes a better solution, based on the observation that R, being
a symmetric function of y1 , y2 , . . . , yn1 , may be expressed as R = 0 +
1 y + 2 y2 + . . . + n1 yn1 , with coefficients 0 , 1 , 2 , . . . , n1 polyno-
mial in p0 , p1 , . . . , pm1 , q0 , q1 , . . . , qn1 , and the same is true for Fy R =
t0 + t1 y + t2 y2 + . . . + tn1 yn1 . Naturally, Rk = 0 + 1 yk + 2 yk2 + . . . + n1 ykn1
and Fyk Rk = t0 + t1 yk + t2 yk2 + . . . + tn1 ykn1 . Now taking y = 1 y , we have
Rk 1 yk y2 yn1
k k
= 0 + 1 + 2 + . . . + n1
yk yk yk yk yk
= n1
Rk Fyk tn1 Fy
and, in the same way, yk = tn1 , so that Fy = n1 . For a rational function F y
tn1
(where F is not the derivative of F!), the value is
tn1
, where tn1 is the coefficient
n1 n1 n2
of y in F y R. In the case of f y = y, let R = y + y + . . . ; then
Ry = yn + yn1 + . . . = ( qn1 )yn1 + . . .

so that y = qn1 + .
In his researches about elliptic functions, published in the second and the third
volume of Crelles Journal (182728, uvres, t. I, p. 294314 and 355362) Abel
met some new cases of algebraic equations solvable by radicals. Such equations
were known from the time of A. de Moivre (1707), who showed that the equation
of degree n (odd integer) giving sin na is solvable by radicals, when sin a is known,
the formula involving sin 2
n in its coefficients. Then Gauss (1801) proved that the
2i
cyclotomic equation of degree n 1 (n a prime number) giving the n-th root e n is
also solvable.
Abel developed an analogous theory for elliptic functions instead of circular

x
functions. Let x = u be defined by u = 2 dt2 2 2
(e, c given real param-
0 (1c t )(1+e t )
eters); then is a uniform meromorphic function in the complex domain, with two
independent periods
1/c
dx
2 = 4 ,
(1 c2 x 2 )(1 + e2 x 2 )
0
1/e
dx
2i = 4i .
(1 e x 2 )(1 + c2 x 2 )
2
0
Abel discovered that the equation of degree n 2 giving na when (a) is known

is solvable by radicals, the formula involving n , being a period, in its coef-

2
ficients. The equation of degree n 21 giving the non-zero values of 2 n may be

decomposed in n + 1 equations of degree n1 2 , all solvable, by means of an equation

of degree n + 1, which, in general, is not solvable
by radicals. For certain singular
moduli ec , for instance when ec = 1, 3 or 2 3, the equation of degree n + 1 is
also solvable; Gauss already knew this lemniscatic case, where c = e = 1.
The base for these results is Eulers theorem of addition for elliptic inte-
fF+ fF
grals, which gives, in Abels notation ( + ) = 1+e 2 c2 2 2 , where f =

1 + e . So the roots of the equation for n are

2 2
1 c , F = 2 2

(1)m+ n + m+i n , |m|, || n12 and they are rational functions of

(), f, F, where = n . Abel defines

2m
2i
1 = + , = 1 +
n1
n n1
n
|m| 2 || 2

2i
and 1 = 1 ,
n1
n
|| 2
where is an n-th root of 1; he proves, by means of the additiontheorem, that 1

n
and ()n +(1 )n are rational functions of , so that = A + A2 B n with
n1
2
+ 2m + 2m

A and B rational in . Indeed, 1 = + n n is ra-
m=1
56 C. Houzel

tional with respect to = x and 1 2i

n = R R (1 c2 x 2 )(1 + e2 x 2 ),
where R and R are rational in x, so that and 1 have the same form and 1

is deduced from by changing
the sign of
the radical (1 c2 x 2 )(1 + e2 x 2 ). Now

2k i k 2k i
1 + 2k 2k 2k

n = 1 , + n + n = and 1 + n + n =

k 1 so that 1 and ()n +(1 )n , which are rational in , take the same
value when is replaced by any other root of the considered equation and are
therefore rational in . The n 1 different values of give n 1 values A j and B j
for A and B and one has

2ki 1 k n
1 + = + j A j + A2j B nj . (36)
n n j
2m 2m

m 1 + m 1

Then Abel uses 2 = n , 3 = n ,
|m| n1
2 |m| n1
2
such that 2 3 and (2 )n + n
(3 ) are rational functions of 1 . He gets
n
2 = C + C 2 Dn with C, D rational in 1 and

1 n
= 1 + C j + C 2j Dnj . (37)
n j
The radicals in (36) and (37) are not independent (otherwiseeach formula
should
give n n1 different values). Indeed, if k = k 1 + 2i , 1k =

n
2 || n1

2i 2 2
2i

k 1 k
= k k

n , where = cos n + i sin n , + n
|| n1
2
2i k 1k k 1k
and 1k + = k 1k so that

n ( 1 )k
+ (11 )k
and ( 1 )kn
+ 1
(1 )kn
are
rational functions of . As

( 1 )n = A1 + A21 B1n and (11 )n = A1 A21 B1n ,

it is easy to deduce k = ( 1 )k (Fk + Hk A21 B1n ), with Fk and Hk rational

k
in , that is n Ak + A2k Bkn = (A1 + A21 B1n ) n (Fk + Hk A21 B1n ). In the

k
same way, n Ck + Ck2 Dkn = (C1 + C12 D1n ) n (K k + L k A21 B1n ), with K k
and L k rational in .
For the other problem (division of the periods by an odd prime n), the roots are

2 m i n1
, 0 m, , (m, ) = (0, 0);
n 2
Abel groups them according to the points m: of the projective line P 1 (Fn ), which
gives n + 1 groups of n1 n1
2 roots each: (1, 0), (m, 1), 1 2 , 0 m n 1.
m+i

He puts r = 2 2
. If p = 2 n

n , r,m = n is any rational

k n1
symmetric function of the 2 n , 1 k 2 , = m + i, then r = r1
and r,m = r1,m . Consider now the equation of degre n + 1 of which the roots
are r1 and r1,m (0 m n 1). The sum of the k-th powers of these roots is
2 k k
n1 times the sum of the (r ) and of the (r,m ) , so it is rational with respect to
e and c, and the same is true for the coefficients of the considered equation of degree
n + 1. We thus see that the equations of degree n1 2 respectively giving the r and
the r,m for m fixed have coefficients given by an equation of degree n + 1. In fact
one such auxiliary equation of degree n + 1 is sufficient, for when a function such as
p = r1 is given, any other such function q = r1 is a rational function of p. Abel
proves this by a method due to Lagrange, determining r1 and the r1,m as solutions
of the linear system
(r1 )k r1 + (r1,0 )k r1,0 + (r1,1 )k r1,1 + . . . + (r1,n1 )k r1,n1 = sk
where the sk are easily seen to be symmetric with respect to r1 , r2 , . . . , r1,0 , . . . and
so rational functions of e and c.
2 k
In order to solve the n + 1 equations of degree n1
2 , with roots n , Abel

procedes as Gauss did for the cyclotomy: putting = n and a primitive root
modulo n, the roots may then be written 2 ( ), 0 n3
2 and the Lagrange
resolvant is
n3 n3
() = 2 () + 2 () + . . . + 2 2 2 ,
where is a n1 2 n1
2 -th root of 1. It is a rational function of () and its 2 -th power v
n1
is symmetric with respect to the 2 roots. Thus v is known when the equation of
degree n + 1 is solved. Varying , we get n1 2 values vk for v and

2 2
n1

n1
() = p n3 + 2 v1 + . . . + 2 v n3 ,
n1 2 2
n3
where p n3 = 2 () + 2 () + . . . + 2 2 corresponds to = 1 and
2
n1
2 vk
is symmetric. The n 1 radicals are not independent: sk =
n1
k is a rational
2 v1
function of 2 () which remains invariant when is replaced by m , so it is a rational

function of the root of the auxiliary equation of degree n + 1.
When e = c = 1, = and there exists a complex multiplication, that is
a formula
(m + i) = T,
4
for m, integers and m + odd, with T a rational
function if (). This permits
the solution by radicals of the equation giving n whenever n is a prime number
58 C. Houzel
of the form 4 + 1. Indeed, consider n = 2 + 2 , where and are relatively prime

integers such that + is odd, and there exists integers m, t such that 1 = 2m nt;
m
then +i m
+ i = 2m
n and

2m
= + t = (1)t ,
n n n

so it is sufficient to consider the equation for +i = . This equation is

equivalent to ( + i) = 0 and its roots are x = m+i +i , m, integers; it is

n1
easy to reduce these roots to the form x = +i , || 2 = 2 and to see that,
in this form, they are all distinct. Moreover, the equation has no multiple roots, and
finally, since is an odd function, we have to consider an equation of degree 2, with
roots 2 (), 2 (2), . . . , 2 (2) or 2 (), 2 (), 2 (2 ), . . . , 2 (21 ), where
is a primitve root modulo n. Then, as for the preceding problem, one sees that
1 2
2 m 1 m 2 2m 2 (21)m 21
( ) = 2 A + v + s2 v + s21 v 2 , where
v = (2 () + 2 () + 2 2 (2 ) + . . . + 21 2 (21 ))2 ,
2 () + k 2 () + 2k 2 (2 ) + . . . + (21)k 2 (21 )
sk = ,
(2 () + 2 () + 2 2 (2 ) + . . . + 21 2 (21 ))k
A = 2 () + 2 () + 2 (2 ) + . . . + 2 (21 )
are rational with respect to the coefficients of the equation, so of the form a + bi
with a and b rational numbers. Abel oberves that, when n is a Fermat prime number
N1
2 N +1, all the radicals in the solution are of index 2 for 2 = 2 N1 and 2 = 1. He
applies these results to the division of the lemniscate of polar equation x = cos 2
(x distance to the origin, polar angle), for which the elementary arc is dx 4 .
1x
All these examples of solvable equations (Moivre, Gauss, elliptic functions) gave
Abel models for a general class of solvable equations; following Kronecker, we call
them Abelian equations and they are the object of a memoir published in the fourth
volume of Crelles Journal (1829, uvres, t. I, p. 478507). To begin with, Abel
defines (in a footnote) the notion of an irreducible equation with coefficients rational
functions of some quantities a, b, c, . . . considered as known; his first theorem states
that if a root of an irreducible equation x = 0 annihilates a rational function fx of
x and the same quantities a, b, c, . . . , then the it is still true for any other root of
x = 0 (the proof is given in a footnote).
The second theorem states that if an irreducible equation x = 0 of degree has
two roots x and x1 related by a rational relation x = x1 with known coefficients,
then the given equation may be decomposed in m equations of degree n of which
the coefficients are rational functions of a root of an auxiliary equation of degree
m (naturally = mn). First of all, the equation (x1 ) = 0 with the theorem I
shows that (x) = 0 for any root x of x = 0; so x = 2 x1 , 3 x1 , . . . are all
roots of x = 0. If m x1 = m+n x1 (the equation has only a finite number of
roots), or n ( m x1 ) m x1 = 0, we have n x x = 0 for any root of x = 0

by the theorem I and, in particular n x1 = x1 ; if n is minimal with this property,
x1 , x, . . . , n1 x1 are distinct roots and the sequence ( m x1 ) is periodic with period
n. When > n, there exists a root x2 which does not belong to this sequence; then
( m x2 ) is a new sequence of roots with exactly the same period for n x2 = x2 and
if k x2 = x2 for a k < n, we should have k x1 = x1 . When > 2n, there exists
a root x3 different from the m x1 and the m x2 , and the sequence ( m x3 ) has a period
n. Continuing in this way, we see that is necessarily a multiple mn of n and
that the roots may be grouped in m sequences ( k x j )0kn1 ( j = 1, 2, . . . , m).
Note that this proof is analogous to that of Lagrange establishing that a rational
function of n letters takes, under the substitutions of these letters, a number of values
which divides n!. In order to prove his second theorem, Abel considers a rational
symmetric function y1 = f(x1 , x1 , . . . , n1 x1 ) = Fx1 of the first n roots and the
corresponding y j = Fx j = F( k x j ) (2 j m); for any natural integer , the
sum y1 + y2 + . . . + ym is symmetric with respect to the mn roots of x = 0, so
it is a known quantity and the same is true for the coefficients of the equation with
the roots y1 , y2 , . . . , ym . Since the equation with the roots x1 , x1 , . . . , n1 x1 has
its coefficients rational symmetric functions of x1 , x1 , . . . , n1 x1 , each of these
coefficients is a root of an equation of degree m with known coefficients. In fact
one auxiliary equation of degree m is sufficient: this is proved by the stratagem of
Lagrange already used by Abel for the division of the periods of elliptic functions
(Abel notes that it is necessary to choose the auxiliary equation without multiple
roots, which is always possible).
When m = 1, = n; the roots constitute only one sequence x1 , x1 , . . . , 1 x1
and the equation x = 0 is algebraically solvable as Abel states it in his theorem III.
We now say that the equation x = 0 is cyclic. This result comes from the fact that
the Lagrange resolvant
x + x + 2 2 x + . . . + 1 1 x
(x any root of the equation, -th root of 1) has a -th power v symmetric with

respect to the roots. We now get x = 1 (A + v1 + v2 + . . . + v1 ),
where v0 , v1 , . . . , v1 are the values of v corresponding to the diverse -th roots

of 1 ( = 1 for v0 ) and A = v0 . The 1 radicals are not independent for if

= cos 2 2
+ 1 sin and if

vk = x + k x + 2k 2 x + . . . + (m1)k 1 x,

the quantity vk ( v1 )k = ak is a symmetric function of the roots of x = 0,
so it is known. Abel recalls that this method was used by Gauss in order to solve
the equations of the cyclotomy. The theorem IV is a corollary of the preceding one:
when the degree is a prime number and two roots of x = 0 are such that one of
them is a rational function of the other, then the equation is algebraically solvable.

As a1 = v1 v1 = a, does not change when is replaced by its complex
conjugate, it is real when the known quantities are supposed to be real. Thus v1 and
v1 are complex conjugate and
60 C. Houzel

v1 = c + 1 a c2 = ( ) (cos + 1 sin ),
and so
+ 2m

+ 2m
v1 = cos + 1 sin .

So in order to solve x = 0 it suffices to divide the circle in equal parts, to divide

the angle (which is constructible) in equal parts and to extract the square root of
. Moreover, Abel notes that the roots of x = 0 are all real or all imaginary; if
is odd they are all real.
The theorem VI is relative to a cyclic equation x = 0 of composite degree =
m 1 m 2 m = m 1 p1 . Abel groups the roots in m 1 sequences ( km 1 + j x)0k p1 1
(0 j m 1 1) of p1 roots each. This allows the decomposition of the equation in
m 1 equations of degree p1 with coefficients rational functions of a root of an auxiliary
equation of degree m 1 . In the same way, each equation of degree p1 = m 2 p2 is
decomposed in m 2 equations of degree p2 using an auxiliary equation of degree m 2 ,
etc. Finally, the solution of x = 0 is reduced to that of equations of respective
degrees m 1 , m 2 , . . . , m . As Abel notes, this is precisely what Gauss did for the
cyclotomy. The case in which m 1 , m 2 , . . . , m are relatively prime by pairs is
particularly interesting. Here for 1 k an auxiliary equation f yk = 0 of degree
m k allows to decompose x = 0 in m k equations Fk ( j x, yk ) = 0 of degree n k = mk
(0 j m k 1). Since x is the only common root of the equations Fk (x, yk ) = 0
(for km p x = m q x with k n p 1 and n q 1 implies km p = m q and then
k = = 0 if p = q), it is rational with respect to y1 , y2 , . . . , y . So, in this case, the
resolution is reduced to that of the equations f 1 y1 = 0, f 2 y2 = 0, . . . , f y = 0
of respective degrees m 1 , m 2 , . . . , m and with coefficients known quantities. One
may take for the m k the prime-powers which compose .
All the auxiliary equations are cyclic as is x = 0, so they may be solved by the
same method. This follows from the fact that if
y = Fx = f(x, m x, 2m x, . . . , (n1)m x)
is symmetric with respect to x, m x, 2m x, . . . , (n1)m x, so is F(x). Then, by

Lagranges stratagem F(x) is a rational function y of y.
Abel ends this part of the memoir with the theorem VII, relating to a cyclic
equation of degree 2 : its solution amounts to the extraction of square roots. This
is the case for Gauss division of the circle by a Fermat prime.
The second part deals with algebraic equations of which all the roots are rational
functions of one of them, say x. According to Abels theorem VIII, if x = 0 is
such an equation of degree and if, for any two roots x and 1 x the relation
1 x = 1 x = 0 is true, then the equation is algebraically solvable. Abel begins
by observing that one may suppose that x = 0 is irreducible. So that if n is the
period of ( k x), the roots are grouped in m = n groups of n roots. Each group
contains the roots of an equation of degree n with coefficients rational functions
of a quantity y = f(x, x, 2 x, . . . , n1 x) given by an equation of degree m
with known coefficients, which is easily seen to be irreducible. The other roots
of the equation in y are of the form y1 = f(1 x, 1 x, 2 1 x, . . . , n1 1 x) =
f(1 x, 1 x, 1 2 x, . . . , 1 n1 x) (by the hypothesis), so rational symmetric with
respect to x, x, 2 x, . . . , n1 x and (again by Lagranges stratagem) rational in y:
y1 = y. Now if y2 = 1 y = f(2 x, 2 x, 2 2 x, . . . , n1 2 x),
1 y = y2 = f(1 2 x, 1 2 x, . . . , n1 1 2 x)
= f(2 1 x, 2 1 x, . . . , n1 2 1 x) = 1 y
so that the equation in y has the same property as the initial equation x = 0 and it
is possible to deal with it in the same manner. Finally x = 0 is solvable through
a certain number of cyclic equations of degrees n, n 1 , n 2 , . . . , n such that =
nn 1 n 2 n , this is Abels theorem IX. In the theorem X, Abel states that when

= 11 22 . . . with 1 , 2 , . . . , prime, the solution amounts to that of 1
equations of degree 1 , 2 equations of degree 2 , . . . , equations of degree , all
solvable by radicals.
As an example, Abel applies his general theorem to the division of the circle
in = 2n + 1 equal parts, where is a prime number; the equation with roots
cos 2 4 2n
, cos , . . . , cos has rational coefficients and it is cyclic. If m is a prim-
itive root modulo , the roots are x, x, . . . , n1 x where x = cos 2 = cos a and
x = cos ma, polynomial of degree m. As Gauss has proved, the division of the
circle is reduced in parts is reduced to the division of the circle in n parts, the
division of a certain (constructible) angle in n parts and the extraction of a square
root of a quantity
= |(x + x + 2 2 x + . . . + n1 n1 x)
(x + n1 x + n2 2 x + . . . + n1 x)|.
1 1 1 2 n1 1 1
It is not difficult
to compute = 2 n 4 2 ( + + . . . + ) = 2 n + 4 so the
square root is conformally to Gauss result. After his notebooks, we know that
Abel also wanted to apply his theory to the division of periods of elliptic functions
with a singular modulus, precisely in the case where = 2n + 1.
On the 8 of October 1828, Abel sent the statement of three theorems on algebraic
equations to Crelle.
A. Given a prime number n and n unknown quantities x1 , x2 , . . . , xn related by
the relations (x1 , x2 , . . . , xn ) = (x2 , x3 , . . . , x1 ) = . . . = (xn , x1 , . . . , xn1 ) = 0,
where is a polynomial of degree m, the equation of degree m n m obtained
by elimination of n 1 of the quantities and division by the factor (x, x, . . . , x)
n
is decomposable in m nm equations of degree n, all algebraically solvable, with
n
the help of an equation of degree m nm . Abel gives, as examples, the cases where
n = 2, m = 3 and n = 3, m = 2; in these cases m n m = 6 and the equation of
degree 6 is algebraically solvable.
B. If three roots of an irreducible equation of prime degree are so related that one
of them is rationally expressed by the other two, then the equation is algebraically
solvable.
62 C. Houzel
This theorem is given, as a necessary and sufficient condition, by E. Galois as

an application of his Memoire sur les conditions de resolubilite des equations par
radicaux (1831) and it was at first interpreted as the main result of this memoir.
C. If two roots of an irreducible equation of prime degree are so related that
one of them is rationally expressed by the other, then the equation is algebraically
solvable.
This statement is the same as that of the theorem IV in the 1829 memoir (which
was composed in March 1828).
Abel left uncompleted an important paper Sur la theorie algebrique des equations
(uvres, t. II, p. 217243). In the introduction, he explains in a very lucid way his
method in mathematics, saying that one must give a problem such a form that it
is always possible to solve it. For the case of the solution by radicals of algebraic
equations, Abel formulates certain problems:
(1) To find all the equations of a given degree which are algebraically solvable.
(2) To judge whether a given equation is algebraically solvable.
(3) To find all the equations that a given algebraic function may satisfy.
Here an algebraic function is defined, as in the 1826 paper, as built by the
operations of addition, subtraction, multiplication, division and extraction of roots
of prime index. There are two types of equations to consider: those for which the
coefficients are rational functions of certain variables x, z, z , z , . . . (with arbitrary
numerical coefficients; for instance the general equation of a given degree, for which
the coefficients are independent variables) and those for which the coefficients are
constant; in the last case the coefficients are supposed to be rational expressions in
given numerical quantities , , , . . . with rational coefficients. An equation of the
first type is said to be algebraically satisfied (resp. algebraically solvable) when it
is verified when the unknown is replaced by an algebraic function of x, z, z , z , . . .
(resp. when all the roots are algebraic functions of x, z, z , z , . . . ); there are anal-
ogous definitions for the second type, with algebraic function of x, z, z , z , . . .
replaced by algebraic expression of , , , . . . .
In order to attack his three problems, Abel is led to solve the following ones To
find the most general form of an algebraic expression and To find all the possible
equations which an algebraic function may satisfy. These equations are infinite in
number but, for a given algebraic function, there is one of minimal degree, and this
one is irreducible.
Abel states some general results he has obtained about these problems:
(1) If an irreducible equation may be algebraically satisfied, it is algebraically
solvable; the same expression represents all the roots, by giving the radicals in
it all their values.
(2) If an algebraic expression satisfies an equation, it is possible to give it such
a form that it still satisfies the equation when one gives to the radicals in it all
their values.
(3) The degree of an irreducible algebraically solvable equation is the product of
certain indexes of the radicals in the expression of the roots.
About the problem To find the most general algebraic expression which may
staisfy an equation of given degree, Abel states the following results:
(1) If an irreducible equationof prime degree is
algebraically solvable, its roots
are of the form y = A + R1 + R2 + . . . + R1 , where A is rational and
R1 , R2 , . . . , R1 are roots of an equation of degree 1.
This form was conjectured by Euler (1738) for the general equation of degree
.
(2) If an irreducible equation of degree , with prime, is algebraically solvable,
either it may be decomposed in equations of degree of which the
coefficients depend
on an equation
of degree , or each root has the form

y = A + R1 + R2 + . . . + R , with A rational and R1 , R2 , . . . , R roots
of an equation of degree 1.
(3) If an irreducible equation of degree not a prime-power is algebraically solv-
able, it is possible to decompose in a product of two factors 1 and 2 and
the equation in 1 equations of degree 2 of which the coefficients depend on
equations of degree 1 .
(4) If an irreducible equation of degree , with prime, is algebraically
solvable,
its roots may be expressed by the formula y = f( R1 , R2 , . . . , R ) with
f rational symmetric and R1 , R2 , . . . , R roots of an equation of degree
1.
A corollary of (3) is that when an irreducible equation of degree =

1 1 2 2 . . . (1 , 2 , . . . , prime) is algebraically solvable, only the radicals

necessary to express the roots of equations of degrees 1 1 , 2 2 , . . . , appears in
the expression of the roots. Abel adds that if an irreducible equation is algebraically
solvable, its roots may be found by Lagranges method. According to this method,
an equation of degree is reduced to the solution of (1)! () equations of degree ()
( the Euler function) with the help of an equation of degree (1)!
() (Abel text leaves
a blank at the place of these numbers). Abel announces that a necessary condition
for the algebraic solvability is that the equation of degree (1)!
() have a root rational
with respect to the coefficients of the proposed equation; if is a prime number, this
condition is also sufficient.
The first paragraph of the paper explains the structure of algebraic expressions,
as was done in the published 1826 article; this time, the order of such an expression
is defined as the minimum number of radicals necessary to write it. In the second
paragraph, a polynomial
yn + Ayn1 + A yn2 + . . . = (y, m)
is said to be of order m when the maximum order of its coefficients A, A , . . . is m.

1 2 1 1

The first theorem states that an expression t0 +t1 y1 1 +t2 y1 1 +. . .+t1 1 y1 1 , with
t0 , t1 , . . . , t1 1 rational with respect to a 1 -th root of 1 and radicals different
1

from y1 1 , is 0 only if t0 = t1 = . . . = t1 1 = 0. The second theorem states
that if an equation (y, m) = 0 of order m is satisfied by an algebraic expression
64 C. Houzel

y = p0 + p1 1 y1 + . . . of order n > m, it is still satisfied by the expression with
2
1 y1 , 1 y1 , . . . instead of 1 y1 , where is a 1 -th root of 1. After the third
theorem when two equations (y, m) = 0 and 1 (y, n) = 0 have a common root, the
first one being irreducible and n m, then 1 (y, n) = f(y, m) (y, m). Then the
fourth theorem says that 1 (y, n) is divisible by the product (y, m) of (y, m)
and the polynomial (y, m), (y, m), . . . , (1) (y, m) obtained by successively

replacing in (y, m) the outermost radical y1 by y1 , 2 y1 , . . . ( -th

root of 1); this comes from the fact that (y, m), (y, m), . . . , (1) (y, m) are
relatively prime by pairs. In the fifth theorem, Abel states that if (y, m) = 0 is

irreducible, so is (y, m) = 1 (y, m ) = 0.
1 1

Now if am = f ymm , ym1
m1
, . . . , of order m, is a root of an irreducible

equation (y) = 0, must be divisible by y am , and so also by (y am ) =
(y, m 1 ) (theorem
IV), which is irreducible (theorem
V). It now follows that is
divisible by (y, m 1 ) = 1 (y, m 2 ) and by 1 (y, m 2 ) = 2 (y, m 3 ), etc., with
m > m 1 > m 2 > . . . Finally, we arrive at some m +1 = 0 and (y, 0) divides
(y) and has rational coefficients, so that = . This leads to the degree of ,
for that of (y, m 1 ) is m , that of 1 (y, m 2 ) is m m 1 , . . . and that of is
m m 1 . . . m = . This is the third general result of the introduction, with the
further explanation that the index of the outermost radical is always one of the factors
of the degree . The first general result of the introduction is also a consequence of
that fact, as the fact that an algebraic expression solution of an irreducible equation
of degree takes exactly values.
In the third paragraph, Abel first deals with the case in which is a prime
1 2 1
number; then m = and am = p0 + p1 s + p2 s + . . . + p1 s , with s = ym ;
1 1 1 1
giving to the radical s its values s , s , . . .
, 1 s ,
where is a -th root
of 1, we get values z 1 , z 2 , . . . , z for am and, as the given equation has only
roots, we cannot get new values by replacing the p j or s by other values pj and s
obtained by changing the value of the radicals they contain. Now if
1 1 1 1 1
p0 + p1 s + . . . + p1 s = p0 + p1 s + . . . + p1 1 s ,
we see that different values 0 , 1 , . . . , 1 of the root correspond to different

values of the root of 1. Writing the corresponding equalities and adding, we obtain
p0 = p0 , so p0 = p0 and then
1 1
p1 s = p1 s (0 + 1 1 + 2 2 + . . . + 1 +1 ) + . . .
So
1 1
1 2 1
s = f , p0 , p0 , p1 , p1 , . . . , s , s = q0 + q1 s + q2 s + . . . + q1 s
1 2 1
and s = t0 + t1 s + t2 s + . . . + t1 s , although Abels given proof of the
fact that t1 = t2 = . . . = t1 = 0 is not quite complete. In the notes at the end
of the second volume of Abels Works, Sylow has explained how to complete the

proof (p. 332335) in order to finally obtain p 1 s = p s for some between 2 and

1; this shows that p1 s is root of an equation of degree 1.
Changing s, it is possible to get p1 = 1 and then, we have as usual p0 =
1 2
1 1 1
1 + z 2 + . . . + z ) a known quantity, s = (z 1 +
(z z 2 + . . . + z ), p2 s

= 1 (z 1 + 2 z 2 + . . . + 2 z ), . . . ; this gives
1
1
p2 s = (z 1 +2 z 2 +. . .+2(1) z )(z 1 +1 z 2 +. . .+(1) z )2 ,

2
1
p3 s = (z 1 +3 z 2 +. . .+3(1) z )(z 1 +1 z 2 +. . .+(1) z )3 , etc.

By the usual Lagrangian stratagem, Abel proves that q1 = pm s is a rational

function of s and the known quantities for 2 m 1. The distinct values
of s are of the form pm sm with 2 m 1; Abel shows that the irreducible
equation of which s is a root is cyclic of degree dividing 1, with roots s, s1 =

s, . . . , s1 = 1 s, where s = ( fs) sm , f rational, 2 m 1 and
a divisor of 1. He finally arrives at the following form for the root z 1 of y = 0:
1 1 1 m m
z 1 = p0 + s + s1 + . . . + s1

+ 1 s s + 1 s1 s1
m m2 m2 m2

+ . . . + 1 s1 s1 + 2 s s + 2 s1 s1 + . . . + 2 s1 s1

m 1 m 1 m 1

+ . . . + 1 s s + 1 s1 s1 + . . . + 1 s1 s1 ,
where the j are rational functions and
1 1 m m 2 m (1)

s = Aa a1 a2 . . . a1 ,
1 m m 2 m 3 1

s1 = A1 a a1 a2
. . . a1 ,
1 m (1) 1 m m (2)

. . . s1 = A1 a a1 a2 . . . a1 ,
generalising the form communicated to Crelle in march 1826. Naturally a,a1 ,. . .,a1
are roots of a cyclic equation of degree dividing 1, but Abel does not say anything
about it, that part of the paper being almost reduced to computations. Kronecker (1853)
rediscovered this result, and stated it more precisely, also studying the form of the roots
of cyclic equations.
The last part of the paper contains computations to establish the second state-
ment relative to the problem To find the most general algebraic expression which
may satisfy an equation of given degree; Sylow gives an interpretation of these
computations at the end of the volume (p. 336337).
66 C. Houzel
4 Hyperelliptic Integrals
Abel studied Legendres Exercices de Calcul integral at the fall of 1823 and this
book inspired him a series of new important discoveries; we already saw some of
them. A memoir presented in 1826 to the Royal Society of Sciences in Throndhjem
(uvres, t. I, p. 4060) is devoted to a generalisation of Legendres formula for
the exchange of the parameter and the argument in elliptic integrals of the third

fx xdx
kind. Abel considers an integral p = e xa taken from x = c, where f is
(n)
a rational function and x = k(x + ) (x + ) . . . (x + (n) ) with , , . . .
rational numbers; derivating with respect to the parameter a and comparing with the
fx x
derivative of exa with respect to x, he obtains

dp a
f a + p= (38)
da a
e fx x e fc c ( p) p

+ + p a e fx x x p p 2 dx
xa ca
( p) e fx xdx ( p) ( p)

e fx xdx
+
a+ ( p) x+ ( p) ( p)
(a + ( p) ) p +2
(x + ( p) ) p
( p)
if fx = ( p) x p + . When f is polynomial (( p) = 0) and

( p) ( p)
(x+ )
x = (x + )(x + ) . . . (x + (n) ),
there is another formula
a e fx x x e fc c c

dp
f a+ p= (39)
da a a(a x) a(a c)

ap
+ ( p, p ) e fx x x p dx,
a
( p+ p +1)

( p+1) ( p+ p +2) + f

where ( p, p ) = + 23...( p+ p +1) (F, F , . . . denoting the values at
23...( p+ p +2)
x = 0 of the successive derivatives
of faa function Fx).

a

e pe fa
As d p a + f a pda a = a , taking c such that e fc c = 0 in

(38) or such that e fc c c = 0 in (39), Abel gets

e fa
fx
e fa da

e xdx
e fx x (40)
a xa (a x)a
fa p
e a da
= p ( p) e fx x x p p 2 dx
a
e fa da
fx
e xdx
( p)
(a + ( p) )a x + ( p)
fa
e fx xdx

( p) ( p) e da
+
( p)
(a + ( p) ) p +2 a (x + ( p) ) p
and
e fa
fx
e fa da

e x.dx
e fx x x (41)
a x a (a x)a a
fa p
e a da
= ( p, p ) e fx x x p dx
a a
when f is a polynomial; the integrals with respect to a must be taken from a value
fa
which annihilates e a .
Abel gives special cases of these formulae, for instance when is the constant 1;
if more-over fx = x n , one has
xn an
an e dx xn e da
e e
x a ax

n n n n
=n ea an2 da ex dx + ea an3 da ex xdx + . . .

n n
+ ea da ex x n2 dx .
When fx = 0, (40) gives

1

da xdx
x
(a x)a a xa

da xdx da xdx
= +
+ ...
(a + )a x+ (a + )a x +

da xdx
+ (n) (n)

(a + )a x + (n)
1

xdx
da

a p da
( p, p ) aa x x p dx. If,

and (41) a xa x x (ax)aa =
in this last formula, = = . . . = (n) = m, as x = (x)m , ( p, p ) =

( p + 1 + m( p + p + 2))k( p+ p +2) , where k( j) is the coefficient of x j in x, so
1 (x)m dx

m+1 da
(x)
(a)m xa (a x)(a)m+1

a p da

= k( p+ p +2) ( p + 1 + m( p + p + 2)) (x)m x p dx,
(a)m+1
equality which reduces to
dx da
a x (42)
(x a) x (a x) a
p p
1 ( p+ p +2) a da x dx
= ( p p )k
2 a x
when m = 21 and this gives an extension to hyperelliptic integrals of Legen-

dx
dres formula. If, for example x = 1 + x n , one has 1 + an (xa) 1+x n
68 C. Houzel
da p
n p 2 dx
a da x

(n 2 p 2)

1 + x n
(ax) 1+an
= 2 1+an
1+x n
. The ellip-
2 2
tic case corresponds to x = (1 x )(1 x ) and leads to

dx
(1 a2 )(1 a2 )
(x + a) (1 x 2 )(1 x 2 )

da
(1 x 2 )(1 x 2 )
(a + x) (1 a2 )(1 a2 )
x 2 dx

da
=
(1 a2 )(1 a2 ) (1 x 2 )(1 x 2 )
2
a da dx

2
(1 a )(1 a ) 2 (1 x )(1 x 2 )
2
or, with x = sin and a = sin ,

d
cos 1 sin2
(sin + sin ) 1 sin2

d
cos 1 sin2
(sin + sin ) 1 sin2
sin2 d

d
=
1 sin2 1 sin2
2
sin d

d
.
2
1 sin 1 sin2
The formula (40) with fx = x gives
ea ex xdx ea da ea da ex xdx

ex x = ( p)
a xa (a x)a (a + ( p) )a x + ( p)

that is, for x = x 2 1:

ea da ea ex dx x 2 1

x
e 1 x2
(a x) a2 1 a2 1 x a

1 ea da ex dx x 2 1

=
2 (a + 1) a2 1 x+1

1 ea da ex dx x 2 1

+ .
2 (a 1) a2 1 x1
Let us turn back to the formula (41) with = = . . . = (n) = m, but with f
any polynomial:
e fa
fx
e (x)m dx e fa da

fx m+1
e (x)
(a)m xa (a x)(a)m+1

= (( p + p + 2) ( p+ p +2) + ( p + 1 + m( p + p + 2))k( p+ p +2) )
fa p
e a da
e fx (x)m x p dx,
(a)m+1
that is
e fx dx e fa da
e fa a e fx x
(x a) x (a x) a

( p+ p +2) 1 ( p+ p +2)
= ( p + p + 2) + ( p p )k
2
fa p
e a da
e fx (x)m x p dx
(a)m+1

when m = 12 ; if moreover fx = x and x = 1 x 2 , this gives ea 1 a2
ex
ea

sin d
dx da or cos esin sine +sin sin

= ex 1 x 2 = cos e
(xa) 1x 2 (ax) 1a2

sin d
sine +sin
(integrals from , = 2 ).
Abel also applies these formulae to definite integrals: the formula (40) with f
polynomial gives
x
e fx xdx
(43)
xa
x
x
e fa a p da

( p+ p +2)

fa
= e a ( p + p + 2)
e fx xdx
a
a x
e fa da e fx xdx

e fa a ( p)
(a + ( p) )a x + ( p)
a
e fa
when x , x annihilate e fx x and a annihilates a . For fx = 0, this gives
x x
xdx da xdx
= a ( p)
xa (a + ( p) )a x + ( p)
x a x

x e fx dx fa ( p+ p +2)

fa p

x

e fx x p dx.

and for x = 1, xa =e ( p + p + 2) e a da
x a x

e fa
The formula (43), with a = 0 gives
70 C. Houzel
a x

( p) e fa da e fx xdx

(a + ( p) )a x + ( p)
a x
a x

( p+ p +2) e fa a p da
=
( p + p + 2) e fx xdx,
a
a x

a eka da

x ekx xdx
( p)

for instance (a+( p) )a
x+( p)
= 0 for fx = kx and x = 1,
a x
a x
( p+ p +2) fa p

( p + p + 2)
e a da e fx x p dx = 0.
a x
Using now (41) with x , x annihilating e fx x x, Abel gets
x x
e fx xdx e fa a p da

= e fa a ( p, p ) e fx x x p dx
xa a a
x a x
and, when = = . . . = (n) = m,
x x
e fx (x)m dx e fa a p da

= e fa (a)m ( p, p ) e fx (x)m x p dx.
x a (a)m+1
x a x
(44)
For fx = 0 and m = 12 , this gives
x p x p
dx 1 ( p+ p +2) a da x dx
= ( p p )k
(x a) x 2 a a x
x a x
expressing the periods of an hyperelliptic integral of the third kind by means of the
periods of the integrals of the first two kinds; in the elliptic case,
x = (1 x 2 )(1 x 2 ),
one has
x
dx
(1 a2 )(1 a2 )
(x a) (1 x 2 )(1 x 2 )
x
x
x 2 dx

da
=
(1 a2 )(1 a2 ) (1 x 2 )(1 x 2 )
a x
x
a2 da

dx

(1 a2 )(1 a2 ) (1 x 2 )(1 x 2 )
a x

with x , x and a = 1 or 1 .
From (44) with x = 1 x 2n , x = 1, x = 1 and a = 1, Abel deduces
1
dx
(x a)(1 x 2n )m
1

1+2 p
(m + 1) 2n

a2n2 p2 da
= (2 p + 1 2mn)
n(1 a2n )m m + 1 + 1+2 p (1 a2n )1m
2n 1
1
for m > 0. If, for example m = 2 and n = 3, this gives
1 1
dx 2 a4 da
= 62
(x a) 1 x 6 3 1 a6 3 1 a6
1 1
5
2 da
+ 64 .
3 1 a6 3 1 a6
1

e fa

a e fa a p da
x fx
( p, p ) e x x p dx = 0; for instance,

Now if a = 0, aa
a x
(n)
when x = (x +) (x + ) . . . (x +(n) ) and a = (a+)1 (a+ )1 . . .
(n)
(a + (n) )1 , with 1 < , , . . . , (n) < 0, one has
a x
a p da x p dx
( p, p ) =0
a x
a x

if x = ( p) , x = ( p ) , a = (q) and a = (q ) . When = = . . . =
(n) = 12 , denoting by x a polynomial of degree n, with roots , , , . . . and by

p dx
F( p, x) the integral xx , this relation becomes
72 C. Houzel

( p p )k( p+ p +2) F( p, )F( p, , )

+ ( p p )k( p+ p +2) F( p, )F( p , )

= ( p p )k( p+ p +2) F( p, )F( p, , )

+ ( p p )k( p+ p +2) F( p, )F( p , ),
1
and, x = (1 x 2 )(1c2 x 2 ) (elliptic case), = 1, = 1, = c and = 1c ,
we find

1 1
F(1)E = E(1)F ,
c c
dx x 2 dx

where Fx = and Ex = .
0 (1x 2 )(1c2 x 2 ) 0 (1x 2 )(1c2 x 2 )
A short posthumous paper Sur une propriete remarquable dune classe tres
etendue de fonctions transcendantes (uvres, t. II, p. 4346, mem. VIII) contains
some of the same formulae. Abel starts from a differential equation 0 = sy + t dy
dx ,
x fx x
where t = x and s = fx are polynomials; then, for r = xa (xa)2 =
a fa
(xa)2 xa + R, with

1 1 1
R= a f a + a f a (x a)
2 3 2

1 1
+ a f a (x a)2 + . . . ,
24 23
and y = x solution of the differential equation, one has

yx yx dz
rydx = or
= a fa z + Rydx,
xa xa da

ydx dp
where z = xa . If z = pa, this equality becomes Rydx xx

xa = a a da
and we have
1

xdx da R x
p= x x = dxda
a xa (a x)a a a a
am da

= ((n + 1)m+n+2 m+n+1 ) x n xdx,
a a
where k (resp. k ) is the coefficient of x k in x (resp. fx); the origin of integration

1
in x (resp. in a) must annihilate x x (resp. a ). Note that, up to a constant factor,

fx
p
x = e x dx has the form (x)m (x
e
m
1 ) 1 ...
, where p is a rational function (not the
preceding p !) and 0 < m, m 1 , . . . < 1. When x = 1x , fx 12 x = 0 and
m = 12 (m + 1)m+1 and we find the formula (42).
The following paper in the second volume of Abels Works (p. 4754, mem. IX)
extends this theory to the case of a linear differential equation of order m with
polynomial coefficients:
dy dm y
0 = sy + s1
+ . . . + sm m , (45)
dx dx
sk = k (x) polynomial. Abel looks for a function r such that
dy d m2 y d m1 y
rydx = vy + v1 + vm2 m2 + tsm m1 ,
dx dx dx
1 dv
with t a given function, taken equal to xa . He finds that v1 = s t dx for
d(s+1 t)
0 m 1 (with v1 = r and vm1 = tsm ), so that v1 = s t dx +
d 2 (s+2 t)
dx 2
. . . and
d(s1 t) d 2 (s2 t) d m (sm t)

r = st + 2
. . . + (1)m .
dx dx dx m
(a)
Now s t = xa + R , where R is polynomial in x, so
a 1 a 2 a m a
r = + + (3) + . . . + (m + 1) + ,
x a (x a)2 (x a)3 (x a)m+1
m dm R
with = R dR 1

ydx
dx + . . . + (1) dx m . Thus the integral z = xa
satisfies a dif-
dz dm z
ferential equation a z + 1 a da + . . . + m a da m = ydx, where
m2 m1
= vy + v1 dy d y d y
dx + vm2 dx m2 + tsm dx m1 and = 0 (where 0 is the value of
at the origin of integration); Abel writes the solution of this equation in terms of
a fundamental system of solutions of (45).
In the second part of the paper, Abel wishes to find coefficients 1 , 2 , . . . , m

1 2 mm

depending on a such that z = xa + (xa)2 + . . . + (xa)m ydx satisfy a differ-
dz

ential equation of the form z + da = + ydx; he writes induction relations

between the and, supposing = constant, a differential equation to deter-
mine .
The first article on Abelian integrals published by Abel in Crelles Journal
(1826, uvres, t. I, p. 104144) is devoted to the search for differential forms dx ,
R

p+q R
with and R polynomials, such that their integrals have the form log pq R
= z,
pqdR+2( pdqqd p)R M
with p, q polynomials. As dz =
( p2 q 2 R) R
, one must have = N with

dq dp dp
M = pq dR 2 2
dx +2 p dx q dx R and N = p q R. It follows that q = 2 dx p Ndx
dN
dN
and p Ndx must be polynomials. If
N = (x + a)m (x + a1 )m 1 . . . (x + an )m n ,
dN m m1 mn
Ndx = x+a + x+a1 + ... + x+an and we see that p is divisible by
(x + a)(x + a1 ) (x + an ).
74 C. Houzel
Hence p = (x + a)(x + a1 ) (x + an ) p1 and

(x + a)m (x + a1 )m 1 (x + an )m n = p21 (x + a)2 (x + a1 )2 (x + an )2 q 2 R
which shows that m = m 1 = . . . = m n = 1 if p and q are relatively prime (R is
supposed to be square free) and that R = (x + a)(x + a1 ) (x + an )R1 = NR1 .
Now p1 , q, N and
R1 are determined
by the equation p21 N q 2 R1 = 1 and so
dq dp
= p1 q dR
dx + 2 p dx q dx R1 .
Abel studies p21 N q 2 R1 = 1, or more generally p21 N q 2 R1 = v (where v is
polynomial of degree less than the mean of the degrees of N and R1 ), as a Diophantine
equation in the ring of polynomials. The first observation is that N p21 and R1 q 2 must
have the same degree and so R = (NR1 ) = 2(q p1 + R1 ) (where designates
the degree) is an even number, so we can put N = n m and R1 = n + m. By
Euclidian division R1 = Nt + t , with t = 2m and t < n m. Using the method
of indeterminate coefficients, Abel shows that there exist t1 and t1 such that t1 < m
and t = t12 +t1 . The equation now becomes ( p21 q 2 t12 )N q 2 s = v, with s = Nt1 +t
2
and shows that pq1 and t12 differ by a rational fraction of degree less than t1 , and
as a consequence the polynomial part of pq1 is t1 , say t1 . Thus p1 = t1 q + with
2
< q and one can verify that q t1sN = Rs12N sv2 ; if R = R1 N = r 2 + r
q
with r < r, we see as before that t1sN and rs have the same polynomial part, so
q
and r+ts1 N have the same polynomial part 2 and q = 2 + 1 with 1 < .
Now the Diophantine equation becomes
s1 2 2r1 1 s12 = v
where s1 = N + 4t1 N 4s2 and r1 = 2s t1 N; as r12 + ss1 = R, r1 = n and
s, s1 < n. 2 2
Our equation may be written 1 rs11 = rs11 + ss1 + s v2 which shows
1 1
r1 r1
that 1 s1 and s1 have the same polynomial part up to sign. Writing 1 for the
polynomial part of rs11 , the polynomial part of 1 is 21 and = 21 1 + 2 with
2 < 1 . The equation can now be rewritten s2 12 2r2 1 2 s1 22 = v, where
r2 = 21 s1 r1 and s2 = s + 4r1 1 4s1 21 , from which it is easy to see that
r2 = r1 > s2 . Continuing in this way, we obtain
2
sn n1 2rn n1 n sn1 n2 = (1)n1 v
with n1 = 2n n + n+1 , n integral part of rsnn , rn = 2n1 sn1 rn1 and
sn = sn2 + 4rn1 n1 4sn1 2n1 ; as > 1 > . . . > n > . . . , there is an
2
m such that m = 0, giving the equation sm m1 = (1)m1 v. The sequence (k ) is
determined by the Euclidian algorithm and m1 is the g.c.d. of and 1 , and so 1
if p and q are relatively prime and v = (1)m1 sm .
In the initial problem, we had v = 1, so we must take sm constant. As the sk are
of degree k, this gives n 1 conditions on the 2n coefficients of R1 and N, once the

p1 1 d p1
index m is chosen. Thus q = t1 + 1 and = p1 dN
dx + 2N dx : q,
2+
..
21 + . + 2 1
m1
of degree n 1.
Abel adds some observations. The first is that one may put N = 1 in the problem,
for

p+q R 1 p21 N + q 2 R1 + 2 p1 q R 1 p + q R
log = log 2 = log
pq R 2 p1 N + q 2 R1 2 p1 q R 2 p q R
with p = p21 N +q 2 R1 and q = 2 p1 q. In that case, the equation takes the Pell-Fermat

form p 2 q 2 R = 1 and = q2 ddxp . The second is that, if kk is the k-th convergent
of the continued fraction for pq1 , k2 N k2 R1 = (1)k1 sk and so, putting

z k = k N + k R1 and z k = k N k R1 ,

zk rk + R z k1
we get z k
= rk R z k1
. Lastly,

p1 N + q R1 t1 N + R1 r1 N + R
log = log + log
p1 N q R1 t1 N R1 r1 N R

rm N + R
+ . . . + log .
rm N R
Abel also shows that

1
dx = 2( t1 dN + Ndt1 + dr1 + . . . + drm ds . . . m1 dsm1 ).
2
2
p1 R1 v p1
As q = N + q 2 N , one sees that the continued fraction for q is obtained from

R1 1
that for N . Let us suppose that N = 1, so that t1 = r and . An R = r+
..
. 2+ 21 +
1
2
easy computation shows that rm2 + sm sm1 = rm1 + sm1 sm2 = . . . = r12 + ss1 =
2
R = r + s. If sm = a constant, we have rm+1 r 2 = s asm+1 and, since
2
rm+1 = r > s, sm+1 , this implies rm+1 = r and sm+1 = as . This shows that the
r
polynomial part m+1 of sm+1m+1
is equal to a and sm+2 = as1 , rm+2 = r1 and so
n
on, in general we have rm+n = rn1 , rmn = rn , smn = a(1) sn1 and mn =
n1
a(1) n1 . This shows that the continued fraction is periodic with partial quotients
r, 2, 21 , . . . , 2a 1 , 2a, 2ra , 2a, 2a 1 , . . . , 2, 2r, 2, . . . ; if m = 2k 1 is odd,
k
sk1 = a(1) sk1 and a = 1. Conversely, one sees that if R has a continued
fraction of the preceding form, sm = a. Abel may conclude
by a criterion for the

existence of a polynomial such that dx = log y+ R with y rational:

R y R
R must
76 C. Houzel
1
be of the form r + 1
2+
..
21 + .+ 1
1
21 + 1
2+ 1
2r+ 1
2+
..
21 + .
1
and then y=r+ 1
.
2 +
..
21 + .+ 1
1
21 + 1
2+ 2r
Conversely, Abel proves that sk = sk2 implies rk = rk1 , k = k2 , rk+1 =

rk2 , sk+1 = sk3 etc. and generally rk+n = rkn1 , k+n = kn2 , sk+n = skn2 ;
for n = k1, this gives s2k1 = s1 = 1. If sk = csk1 , he finds k = 1c k1 , rk+1 =
k
rk1 , sk+1 = 1c sk2 , . . . , s2k = c(1) . The integral found is
2 1

(dr + dr1 + . . . + drk1 + drk ds 1 ds1 . . . k1 dsk1 )
R 2

r+ R r1 + R rk1 + R 1 rk + R
= log + log + . . . + log + log
r R r1 R rk1 R 2 rk R
when m = 2k and
2 1

(dr + dr1 + . . . + drk1 ds 1 ds1 . . . k2 dsk2 k1 dsk1 )
R 2

r+ R r1 + R rk1 + R
= log + log + . . . + log
r R r1 R rk1 R
when m = 2k 1.
In the elliptic case, where R is of degree 4 = 2n, n = 2 and the sm are of
degree 1, so there is only one condition to write in order to have sm = const.
If R = (x 2 + ax + b)2 + ex, we have r = x 2 + ax + b, s = ex, = x+a e
2
and then r1 = x 2 + ax b, s1 = 4b x + 4ab e e
e + 1, 1 = 4b x 16b2 and s2 =
2 3
2
2
e
ae e e ae e
4b2
+ 16b 3 x 4b 2 4b + 16b2 b . In order to make s1 constant, we have only to

x 2 +ax+ R
put b = 0 and we find that (3x+a)dx , with R = (x 2 +ax)2 +ex.

2 2
= log 2
x +ax R
(x +ax) +ex
In order to make s2 constant, we put e = 4ab, R = (x 2 + ax + b)2 4abx and find

x 2 + ax + b + R

(4x + a)dx
= log
(x 2 + ax + b)2 4abx x 2 + ax + b R

1 x 2 + ax b + R
+ log .
2 x 2 + ax b R
Abel also computes the case in which s3 is constant, which is given by

e = 2b(a a2 + 4b)
and
3
1

(5x + a2 + 4b)dx x 2 + ax + b + R

2 2
= log
x 2 + ax + b R
2 2
(x + ax + b) 2bx(a a + 4b) 2

x 2 + ax b + R
+ log ,
x 2 + ax b R

and the case where s4 is constant, which leads to e = b(3a a2 + 8b) and

(6x + 32 a 12 a2 + 8b)dx

(x 2 + ax + b)2 b(3a + a2 + 8b)x

x 2 + ax + b + R x 2 + ax b + R
= log + log
x 2 + ax + b R x 2 + ax b R

1 x 2 + ax + 14 a(a a2 + 8b) + R
+ log .
2 x 2 + ax + 14 a(a a2 + 8b) R
At the end

of the memoir, Abel states a theorem according to which, whenever
an integral dx
, and R polynomials, may be expressed by logarithms, it is always
R
in the form A log p+q R , with A constant, p and q polynomials.
pq R
Chebyshev (1860) and Zolotarev (1872) studied the same problem in the elliptic
case looking for arithmetical conditions on the coefficients of R, these latter supposed
to be integers.
The first text written by Abel on elliptic functions (between 1823 and 1825), with
the title Theorie des transcendantes elliptiques (uvres, t. II, p. 87188), also deals
with this problem but it was not published by Abel. In the first chapter, Abel studies

Pdx
the conditions under which an elliptic integral R
, with P a rational function
2 3 4
and R = + x + x + x + x , is an algebraic function. At first taking P
polynomial, he observes that this algebraic function must be rational in x and R,
so ofthe form Q + Q R with Q and Q rational; since dQ is rational, we may write
Pdx
d(Q R) = R
. The function Q is a polynomial otherwise its poles would remain

as poles in the differential: Q = f(0) + f(1)x + . . . + f(n)x n and d(Q R) = S dxR ,
with
dQ 1 dR
S=R + Q = (0) + (1)x + . . . + (m)x m .
dx 2 dx
This gives

1
( p) = ( p + 1) f( p + 1) . . . + p + f( p) + p f( p 1)
2

1
+ p f( p 2) + ( p 1) f( p 3)
2
78 C. Houzel
m
and m = n +3. Abel draws the conclusion that the integrals x dx

R
may be expressed
as linear combinations of those with 0 m 2 and an algebraic function; but
dx
xdx
x 2 dx
, and are independent when the reductions admitted involve only
R R R
m
algebraic functions. The reduction of x dx
R
is given by a system of m 2 linear
equations ( p) = 0 for 3 p m 1, (m) = 1 to determine the f( p)
(0 p m 3) and the formulae
1 3
(0) = f(1) + f(0) , (1) = 2 f(2) + f(1) + f(0) ,
2 2
5 3
(2) = 3 f(3) + f(2) + 2 f(1) + f(0) .
2 2
For instance
x 4 dx

5 1 5 1

dx xdx
= +
R 24 2 3 R 12 2 2 R
11
2
x 2 dx

5 2 5
+ x R.
8 2 3 R 12 2 3
When the values found for (0), (1) and (2) are 0, the integral is alge-
4 25 3 15 2 2

x 4 dx
braic; for instance, when R = 125 256
3
3 + 32 2 x + 16 x + x + x ,
4
R
=

5 11
12 2 3 x R.
In a completely analogous manner, Abel reduces (xa)dxm R to a linear combi-

xdx
x 2 dx

nation of dxR , dx

R
, R , (xa) R
and an algebraic function Q R, Q having

only one pole in a: Q = (1) (2) (m1)
xa + (xa)2 + . . . + (xa)m1 . Indeed d(Q R) = S R
dx

with
(1) (2) (m)
S = (0) + (1)x + (2)x 2 + + 2
+ ... + ,
x a (x a) (x a)m

1 1
(0) = a + a2 (1) ( + 4a)(2) (3),
2 2
1
(1) = (1), (2) = (1)
2
and ( p) = ( p1)( p1) ( p 12 )( p) p( p+1) ( p+ 12 )( p+2)
( p+1)( p+3); here = +a+a2 +a3 +a4 , = +2a+3a2 +4a3 ,
= + 3a + 6a2 , = + 4a and = , so that R = + (x a) +
(x a)2 + (x a)3 + (x a)4 . In order to get the announced reduction, we
determine the ( p) by a linear system ( p) = 0 for 2 p m 1, (m) = 1;
then (0), (1), (2) are given by the preceding formulae and
1 3
(1) = (1) (2) (3) 2 (4).
2 2
For instance,
a2 + 12 a
2
dx dx xdx x dx
2
= + +
(x a) fx fa fx 2 fa fx fa fx

1 f a

dx fx
, (46)
2 fa (x a) fx (x a) fa
where R = fx. This reduction does not work if = fa = 0, which gives (m) = 0
and in that case, we must take Q with a pole of order m in a and we see that
dx

2
is reducible to dxR ,

xdx
and xdx
(xa)m R R R
even for m = 1:

2a2 + a 2 x 2 dx 2

dx dx xdx R
= + + .
(x a) R f a R f a R f a R f a x a
(47)
The next task for Abel is to find the possible relations between integrals of the
form

dx
.
(x b) R
It is easy to see that the only possible relations have the form

dx dx
(0) + (1)
(x a) R (x a ) R

dx dx
+(2) + (3)

(x a ) R (x a ) R
A A A

A
= R + + + ,
x a x a x a x a
where a, a , a , a are the roots of R. Using the preceding reduction and the

x 2 dx
fact that dxR , are independent, Abel finds A = 2(0)

xdx
R
and
R f a , A =
2 2
2(1) 2(2) 2(3) 2
f a , A = f a , A = f a , A(2a + a) + A (2a + a ) + A (2a +
a ) + A (2a 2 + a ) = 0 and A + A + A + A = 0; it is possible to choose
A = 0 and this gives the relation

dx dx dx
(0) + (1) + (2)
(x a) R (x a ) R (x a ) R
A A

A
= R + +
x a x a x a
with
1
(0) = (a a )(a a )(a a )(a a )(a + a a a ),
2
1
(1) = (a a)(a a )(a a )(a a)(a + a a a ),
2
1
(2) = (a a)(a a )(a a )(a a )(a + a a a ).
2
80 C. Houzel
Abel looks for linear relations between

2
dx xdx x dx dx dx
, , and , ;
R R R (x a) R (x a ) R
)(aa )
)(a a )
using (46), he finds dxR = (aa dx

+ (aaa dx

a +a aa (xa) R +a aa (xa ) R
and
x 2 dx

xdx
+
R 2 R
a (a a a a ) f a

dx
=

2(a a)(a + a a a ) (x a) R
a(a a a a ) f a

dx
+

2(a a )(a + a a a ) (x a ) R

a (a a a a ) a(a a a a )

R
+ .
(a a )(a + a a a ) x a x a
dx

When a + a = a + a , these relations loose their sense and give (xa) R

+

dx 2 R
(xa ) R
= (a a)(a a )(xa)(xa ) .
In the second chapter of his memoir, Abel studies the integration of elliptic
integrals by logarithmic functions. Such a function must be of the form

T = A log(P + Q R) + A log(P + Q R) + . . . + A(n) log(P (n) + Q (n) R)
with P, Q, P , Q , . . . polynomials and

A, A , . . . constant,
or, subtracting from dT
P+Q R P +Q R
a rational differential T = A log PQ R + A log P Q R + . . . Then dT is a sum

2N dP P dN
of terms of the form M dx
N R with M = A
dx
Q
dx
, N = P 2 Q 2 R; the fraction MN
has only poles of order 1, these poles are not roots of R and it is easy to see that its
polynomial part is of degree 1. Finally

dx xdx dx dx
T = k + k + L + . . . + L () ,
R R (x a) R (x a() ) R
(48)

2 dx
x
and cannot be reduced to other integrals by means of logarithms.
R
Let us suppose that T contains r + 1 logarithmic terms; looking at the degrees
of P, Q, P , Q , . . . and at the corresponding number of indeterminate coefficients
in T , Abel sees that the minimum value of is 2 and that r may be taken equal to 0.
Moreover, one may take
P = f + f x + f x 2

of degree 2, Q = 1 and N of degree 2; then f = , f =
2
and f =
k(2 4)+2k
.
2(k 4k)
For k = 0 and k = 1, one has

xdx
dx
dx
= (G + H K ) + (G H K )
R (x K ) R (x + K ) R
2
+ 2 x + x + R

1
+ log ,
2 +
x + x 2 R 2
2 3 2 2 2 2
where G = 42(+ +4 4
4 +82 42 ) , H = 4 , K = 4
482 3 .
It is possible to have N of degree 1 when (4 ) + 4 (4 2 ) 2 2 2
2
322 643 = 0; then f = 4 1

xdx
dx

8
and R
= 3 ( ) (x+) R

x+x 2 + R
+ 4 2
+ 31 log 2

x+x 2 R
, where = 8 and = 2 .
+
2
Abel finds another reduction, writing R = ( p + qx + rx 2 )( p + q x + x 2 ), P =
f( p + q x + x 2 ) and Q = 1 and choosing f such that
N = ( f 2 r)( p + q x + x 2 )(x a)2 ;

M 1
then N = 1 + L xa with
pq q p + (rq q)a2 q q f 2
L= , a= .
(rq q)a 2( f 2 r)
This leads to the equation f 4 (q 2 4 p ) f 2 (2qq 4 p4

p r)+q 2 4 pr = 0 and
2 f 2 r
+ A log f(f( pp +q x+x )+ R

dx
dx
to the relation R = L (xa) R +q x+x 2 ) R
, where A = f(rq q) .
Another formula is found by supposing N = k(x a)4 ; then
( p + p p p )a2 2( p p p p )a + p p ( p + p ) p p ( p + p ) = 0,
M 1
where p, p , p , p are the roots of R. In that case, N = 1 L xa , where
2( f + a f + a2 f )
L=
, f = p p p p + ka4 ,
f + 2a f

p + p + p + p + 4ka
f = , f = 1 + k
2 1+k
( p+ p p p )2
and k = (2( p + p )4a)(2( p+ p )4a) ; so

dx

(x p)(x p )(x p )(x p )

dx
=L
(x a) (x p)(x p )(x p )(x p )

f + f x + f x 2 + (x p)(x p )(x p )(x p )
+A log ,
f + f x + f x 2 (x p)(x p )(x p )(x p )
82 C. Houzel
1
with A = . All this work is inspired by Legendres reduction
2 ( p+ p 2a)( p + p 2a)
of elliptic integrals to canonical forms as it is presented in the Exercices de Calcul
integral but Abels study is deeper and more general for he investigates all the
possible relations between such integrals and proves the independance of the three
canonical kinds.
x m +k(m1) x m1 +...+k
Abel also studies the general case, where M N = x m +l (m1) x m1 +...+l ; if Q is of
degree n, P must be of degree n + 2 and m 2n + 4 which is the degree of N.
With the notations R = x, P = Fx, Q = fx, a, a , . . . , a(m1) roots of N (with
multiplicities , , . . . , (m1) ), one has

Fa( j) = fa( j) a( j) (0 j m 1),
whence, by successive derivations, a linear system to determine the coefficients of
P and Q. Then x m + k(m1) x m1 + . . . + k takes in a( j) the value A a( j) a( j) ,
where
dN
x = (x a)(x a ) . . . (x a(m1) ) ;
Ndx
this gives a linear system to get k, k , . . . in function of A, a, a , . . . For instance,
when = = . . . = (m1) = 1, m = 2n + 4 = 4 if Q = 1 and Abel finds, for
the coefficients of P,
a a a aa a
f = i
a + i
a
(a a )(a a )(a a ) (a a)(a a )(a a )
aa a aa a
+i
a + i

a ,
(a a)(a a )(a a ) (a a)(a a )(a a )

i a i a i a
f = + + ,
(a a )(a a ) (a a)(a a ) (a a)(a a )

i a i a
f = + (a + a ) f ,
a a a a
1
where i, i , i , i are equal to 1, and A = (a+a +a +a ) f +2 f . When m = 2,
2 3
Q = 1 and P R = C(x a)(x a ) , he finds
1 2a a ( a )2
f = ,
8 a a
1 a a 2a a ( a )2
f = ,
2 a 4 a a
a a a 2 2a a ( a )2
f =
a + ,
2 a 8 a a
1
A=
(a + 3a ) f + 2 f
a ( a )2
and a, a related by a a = a + 12 (a a ) a + 18 (a a )2 2a a .
2 2 2
When P R = C(x a) (x a ) ,
1 a 1 a
f = + ,
4 (a a ) a 4 (a a) a
1 a a 1 a a
f = + ,
2 (a a ) a 2 (a a) a

1 aa a 1 aa a a a a a
f = + ,
4 a a a 4 a a a a a
2
A = a a

a +
a
and a, a related by ( p + p + p + p )aa ( p p p p )(a + a ) + p p (

p + p )
p p ( p + p ) = 0, where p, p , p , p are the roots of R. So dxx =

2b+2b x dx P+ x a a+a a a+ a
(xa)(xa ) x + A log Px , with b = 2 a
a , b = 2
a a
. In
+ +
a a a a
2 2
a third case P R = C(x p)(x a)(x a ) and P = (x p)( f + f x) and a
is function of a.
The last case considered by Abel is that in which m = 1. Here P 2 Q 2 R =

C(x a)2n+4 and M x+k
N = xa with k = a A a. The coefficients of P and Q
are determined by a linear system and then a is given by an algebraic equation; this
leads to

1 1

dx dx P+Q R
= log .
(x a) R A a R a PQ R
Abel observes that the equation P 2 Q 2 R = C(x

a)
2n+4
is equivalent
1 to
2 2 1
P Q R = C, where F(xa) = (xa) P xa , f(xa) = (xa)n Q xa
n+2
1
and (x a) = (x a)4 R xa .
As we know, the same equation is met in the problem to express (k+x)dx

R
by

P+Q R M
a logarithm A log PQ R
; here N = x + k, so N is constant and may be taken
dP
as 1. The conditions of the problem are x + k = 2A Qdx , 1 = P 2 Q 2 R; the first
method proposed by Abel to determine P = f + f x + . . . + f (n+2) x n+2 and Q =

e + e x + . . . + e(n) x n is that of indeterminate coefficients. The first condition gives

e(n) f e(n)
A = (2n+4) f (n+2)
, k = (n+2)e f (n+2)
and the second gives a system of 2n + 5 equations
between the 2n + 4 coefficients e( p) , f ( p) : f 2 e2 = 1, . . . , f (n+2)2 e(n)2 = 0.
The compatibility of this system imposes a relation between the coefficients , , ,
and of R; for instance, when n = 0, so that Q = e and P = f + f x + f x 2 , one
has
2
2 f f e2 = f + 2 f f e2 = 2 f f e2 = 0,

2
whence f = , f = , f = 2
,e =
and
2 2 2 2 2 2 2 2 2
2 2
= 4 + , A = 41 , k =
4 .
84 C. Houzel
But it is possible to get a linear system for the coefficients e( p) , f ( p) : if
Fy = fyn+2 + f yn+1 + . . . + f (n+2) ,

fy = eyn + e yn1 + . . . + e(n)
and y = y4 + y3 + y2 + y + ,

the second condition is (Fy)2 ( f y)2 y = y2n+4 and it gives Fy = fy y
when y = 0. The system is obtained by differentiating 2n + 3 times this relation at

y = 0. When n = 0, one finds f = ce, f = c e, f = c2 e, 0 = c , where c( p) =

d p y 2
dy p y=0 and = + 2 as above; when n = 1, the system is 0 = 2c + c e ,
4 e

0 = 4c + c ee , 0 = 5c + c ee , whence
c c 2c c = 2c c 5c c = 0.
Without restricting the generality, we may take = 1 and = ; the preceding

equations then give = 2, = 3 and finally

xdx
(49)
x + 2x + 3x 2 x +
4 3

1 x 3 + 3x 2 2 2 + (x + 2) x 4 + 2x 3 3x 2 x +
= log .
6 x 3 + 3x 2 2 2 (x + 2) x 4 + 2x 3 3x 2 x +
Abel proposes another way to study the equation P 2 Q 2 R = 1; he writes

it P + 1 = P 2 R , P 1 = Q 2 R , where P Q = Q and R R = R. Then
P = 12 (P 2 R + Q 2 R ) and 2 = P 2 R Q 2 R ; with R = x 2 + 2qx + p, R =

x 2 + 2q x + p and P , Q constant, one finds q = q , P = Q = 2 , P =
p p
2x 2 +4qx+ p+ p 2
p p , Q= p p , k = q and A = 14 , so

(x + q)dx

(x + 2qx + p)(x 2 + 2qx + p )
2

1 2x 2 + 4qx + p + p + 2 R
= log .
4 2x 2 + 4qx + p + p 2 R
x+m x+m
With P = c , Q = c , one finds 2q = r + m m, 2q = r + m m ,
1 1 1 2
p= r(3m m) + m 2 m mm ,
2 2 2
1 1 2 1 2
p = r(3m m ) + m m mm ,

2 2 2
1 1 2 3
2c2 = r(m m)3 + (m m )(m 3 m 2 m m m + m ),
2 2
where r = q + q , and then
(x 2 + 2mx + m 2 )(x 2 + 2qx + p) c2

P= ,
c2
x 2 + (m + m )x + mm
Q= ,
c2
1
k = (3r m m).
4

If we impose k = 0, r = m+m
3 , m = 2q + q, m = 2q + q , p = q 2qq
2
and p = q 2 2qq ; we have

xdx

(x + 2qx q 2qq )(x 2 + 2q x q 2 2qq )
2 2

1 (x +q+2q ) x 2 +2qx q 2 2qq +(x +q +2q) x 2 +2q x q 2 2qq
= log .
4

(x +q+2q ) x 2 +2qx q 2 2qq (x +q +2q) x 2 +2q x q 2 2qq
The second method to study the equation P 2 Q 2 R = 1 is that used by

Abel in his published memoir for the more general case of hyperelliptic integrals:
putting R = r 2 + s, with r of degree 2 and s of degree 1, the equation becomes
P 2 Q 2r 2 Q 2 s = 1 and it shows that P = Qr + Q 1 with deg Q 1 < deg P.
Then Q 21 + 2Q Q 1r Q 2 s = 1 or, if r = sv + u, with v of degree 1 and u constant,
Q 21 +2Q Q 1 u + Qs(2vQ 1 Q) = 1; thus Q 2 = Q 2vQ 1 if of degree < deg Q = n
and
s1 Q 21 2r1 Q 1 Q 2 sQ 22 = 1,
with s1 = 1 + 4uv, r1 = r 2u, deg Q 1 = n 1 and deg Q 2 = n 2. Iterat-

ing the process, one gets equations s21 Q 22+1 2r2 Q 2 Q 2+1 s2 Q 22 = 1,
s2 +1 Q 22 +1 2r2 +1 Q 2 +1 Q 2 +2 s2 Q 22 +2 = 1, with deg Q p = n p; this
gives sn Q 2n = (1)n+1 , Q n and sn constant. The induction relations to determine
the sm are
sm = sm2 + 4u m1 vm1 , rm = rm1 2u m1 = sm vm + u m . (50)

2
A consequence of these relations is that sm1 sm + rm2 = sm1 sm2 + rm1 , so that
this quantity does not depend on m and
sm1 sm + rm2 = ss1 + r12 = r 2 + s = R; (51)

k
as sn = is constant, it is easy to see that rnk = rk , snk = sk1 (1) , vnk =
k1
vk1 (1) and u nk = u k1 . For n = 2+1 and k = +1, this gives = 1 and
u = 0; for n = 2, u 1 +u = 0. The Q m are determined from Q n by the induction
relations Q m = 2vm Q m+1 + Q m+2 and we see that r, 2v, 2v1 , . . . , 2vn1 are the
partial quotients of the continued fraction for QP , which is obtained by truncating

that for R. Putting rm = x 2 + ax + bm , sm = cm + pm x, vm = (gm + x) p1m and
qm = b bm , Abel draws from (50) and (51) the relations
86 C. Houzel
1 2
2 p2 + (a p 2c)qm1 qm2 qm1 cm1 c + qm1 qm
qm = 2
, =
qm1 pm1 p
and pm pm1 = 2qm ;
b b
since u m = m 2 m+1 = 12 (qm+1 qm ) and gm = a cpmm , these relations allow
to determine rm , sm , u m and vm if we know the qm , which are determined by an
2 p+c2
induction relation starting from q = 0, q1 = 2 b p ac
p2
and are rational functions
of a, b, c, p.
Abel applies this method to the elliptic integral 2 (x+k)dx2

. The condi-
(x +ax+ p) + px+c
tion sn = constant is equivalent to pn= 0 and it leads to qn= 0 and qnk = qk . The
1 1 c c1 cn1
coefficient k is equal to n+2 a + n+2 p + p1 + . . . + pn1 and the polynomials P
and Q are determined by the continued fraction. When c = 0, Abel finds the results
published in his 1826 paper, using
p( p + 4ab)
q1 = 2b, q2 = ,
8b2
2b(16b3 p( p + 4ab))
q3 = ,
( p + 4ab)2
4b p( p + 4ab)( p2 + 6ab p + 8a2 b2 8b3 )
q4 = .
(16b3 p( p + 4ab))2

(y+k
)dy P +Q R

From a relation = A log , Abel deduces

R P Q R

x + k dx P+Q R
= A log
x +l R PQ R
1
through the change of variable y = x+l ; he finds k = l + k1 , A = Ak and an
algebraic equation to determine l in function of the coefficients of R. Indeed, when
R = (y2 + ay + b)2 + c + py and R = (b2 + c)(x 4 + x 3 + x 2 + x + ),
2ab + p = (b2 + c)( 4l), a2 + 2b = (b2 + c)( 3l + 6l 2 ),
2a = (b2 + c)( 2l + 3l 2 4l 3 )
and 1 = (b2 + c)( l + l 2 l 3 +l 4 ). From this Abel deduces,

with l instead of
1 1 P+Q R

dx
dx
l, (xl) R = l+k R log PQ R , which gives a new
(2n+4) 2 3 4
+l+l +l +l
proof of (49) when l + k = .
In the third chapter of the Theorie des transcendantes elliptiques, Abel shows
dx

that the periods of an integral of the third kind p = (xa) R

are combinations of

dx
xdx
x 2 dx
the periods of the integrals R , R and R . Taking the integral from a value
x = r which annihilates R = fx, differentiating with respect to a and using (46), he
obtains

d p 1 f a 1

fx dx
+ p= + (A + Bx + Cx 2 )
da 2 fa (a x) fa fa fx
where A = a2 12 a, B = 12 and C = . From this he deduces

da

da

dx
p fa fx = (A + Bx + Cx 2 ) + constant
(a x) fa fa fx
and the constant is seen to be 0 by making a = r. Thus

dx da
fa fx
(x a) fx (a x) fa
1 1
( 2 x + x 2 )dx ( 2 a + a2 )da

da dx
=
fa fx fx fa
which the formula (42) for the case of elliptic integrals. When r is another root of

r dx

da
r ( 12 x+x 2 )dx
r dx
( 12 a+a2 )da
fx, one obtains fa (xa) fx
=
fa

fx
fx
fa
. And
r r r r r

r
r ( 12 x+x 2 )dx
r
r ( 12 a+a2 )da

if r is a third root of fx, da = dx .
fa fx fx fa
r r r r
Abel finds new relations between periods starting from

P+Q R P + Q R
s = A log + A log + ...
PQ R P Q R

B + Cx dx dx
= dx + L + L + ...
R (x a) R (x a ) R
(cf. (48)) which gives, by integrating from r to r :
r
B + Cx
s s = dx
fx
r
r

( 2 a + a 2 )da
1 1
dx L ( 2 a + a2 )da L
+ + . . .
fx fa fa fa fa
r r r
r
1
( 2 x + x 2 )dx

+ L da
+
L da
+ . . .
fx fa fa fa fa
r r r
The end of the Theorie des transcencantes elliptiques (p.

173188)dxis devoted
to the proof that an integral of the third kind (n) =
2 2 2 2 (1+nx ) (1x )(1c x )
may be

transformed in a linear combination of the integral of the first kind
dx
F = 2 2 2
, some logarithms (or arctangents) of algebraic functions
(1x )(1c x )
88 C. Houzel
and another integral of the third kind (n ) with a parameter n arbitrarily large
or arbitrarily close to a certain limit, and more generally to the relations between

integrals of the third kind with different parameters. Let us consider s = arctan QR ,
with R = (1 x 2 )(1 c2 x 2 ) and Q = x(a + bx 2 ); we have ds = M dx
N R with

dQ
N = Q 2 + R and M = 12 Q dR
dx R . If we impose that N = k(1 + nx 2 )(1 + n 1 x 2 )2 ,
dx
we find that k = 1, b = n n, a = (1 + n 1 ) 1 + n n 1 n = (n) and
2 1 L
n 1 = ( 1 + n n)( c + n n) = f(n). Then M L
N = A + 1+nx 2 + 1+n 1 x 2

with A = 2a n1 + n21 b, L = n 1n a and L = 2 n 2a. Thus

n R 2a n (2n + n 1 )
(n) = arctan F (52)
n1 a n ax + bx 3 n1a n

(2a 2 n) n
+ (n 1 ) + C
n1 a n
ax + bx 3
= F + (n 1 ) + arctan
R

n 2a n2nn n) n = (n).
with = n 1a n = (n), =

n 1 a n
1
= (n), = (2a2n 1 a n
It is easy to see that n 1 > 4n and that is an increasing function when
both upper signs are chosen in f(n). Thus, iterating the operation, we arrive at
a parameter n m as large as we wish, with m equivalent to 1n m , m remain-
ing between 0 and 1 and lim m = 0, lim m = 4. On the contrary, when both
lowersigns are chosen,
n m decreases
and its limit is the root k of the equation
k = ( k + 1 k)( k + c2 k). Applying (52) to (k), we obtain
3
2a + 3 k 1 ax k 2 x 3
(k) = F arctan .
3(a + k) 3(a + k) R

The formulae n 1 = ( 1 + n + n)( c2 + n n) and

n 1 = ( 1 + n n)( c2 + n + n)
respectively lead to values of n m between c2 and c and between 1 and c. Abel
also studies the case in which n is negative.
The transformed parameter n 1 is given by an equation of degree 4; inversely, one
has
(n 21 c2 )2
n= .
4n 1 (n 1 + 1)(n 1 + c2 )
When the sequence (n m ) is periodic, the integrals (n m ) may be expressed as
combinations of F and some arctangents.
Abel finds other relations as

m (n 1 ) A 1 Q R
(n) = (n 1 ) F+ arctan ,
(n) nn 1 (n) (n) P
with P, Q polynomials such that

2 2 2 2 (1 + n)(c2 + n)
P + Q R = (1 + nx ) (1 + n 1 x ) , (n) = ,
n
A constant and n 1 = (n) a certain function. When for instance P = 1 + bx 2 and

2 1+n)
Q = ex, (n) = c(c c +n)(1 n .
An essential discovery of Abel in the theory of elliptic functions is that these func-
tions, obtained by inverting elliptic integrals, have 2 independent periods in the com-
plex domain. In his posthumous memoirProprietes remarquables de la fonction y =
x determinee par lequation fydydx (a y)(a1 y)(a2 y) . . . (am y) = 0,
f y etant une fonction quelconque de y qui ne devient pas nulle ou infinie lorsque
y = a, a1 , a2 , . . . , am (uvres, t. II, p. 4042), he shows that the
function x,
f ydy
which is the inverse function of the hyperelliptic integral x =
y
where
y = (a y)(a1 y)(a2 y) . . . (am y), must have each of the numbers 2( k )
as period, where k is the values of the integral corresponding to y = ak . Jacobi later
proved (1834) that a regular uniform function of one complex variable cannot have
more than 2 independent periods; thus the inverse function of a hyperelliptic integral
cannot be uniform when m > 4. The inversion problem for hyperelliptic integrals
or more generally for abelian integrals must involve functions of several complex
variable, as Jacobi (1832) discovered through his intertretation of Abel theorem.
Here Abel writes the Taylor series for the function :

(x + v) = y + v2 Q 2 + v4 Q 4 + v6 Q 6 + . . . + y(vQ 1 + v3 Q 3 + v5 q5 + . . . )
where the Q j do not have poles at the ak . Thus ( + v) = a + v2 Q 2 + v4 Q 4 +

v6 Q 6 + . . . ( + v) is an even function of v and (2 v) = v. In the same way
(21 v) = v and (2 21 + v) = v and so on.
5 Abel Theorem
The most famous of Abels results is a remarkable extension of Euler addition

theorem for elliptic integrals. It is known as Abel theorem and gives the corresponding
property for any integral of an algebraic function; such integrals are now called
abelian integrals. This theorem, sent to the french Academy of Sciences by Abel
in 1826 in a long memoir titled Memoire sur une propriete generale dune classe
tres-etendue de fonctions transcendantes, is rightly considered as the base of the
following developments in algebraic geometry. Due to the negligence of the french
Academicians, this fundamental memoir was published only in 1841, after the first
edition of Abels Work (1839).
In the introduction, Abel gives the following statement:
When several functions are given of which the derivatives may be roots of the
same algebraic equation, of which all the coefficients are rational functions of the
same variable, one can always express the sum of any number of such functions by
90 C. Houzel
an algebraic and logarithmic function, provided that a certain number of algebraic

relations be prescribed between the variables of the functions in question.
He adds that the number of relations does not depend on the number of the
functions, but only on their nature. It is 1 for the elliptic integrals, 2 for the functions
of which the derivatives contains only the square root of a polynmial of degree 6
as irrationality.
A second statement, which is properly Abel theorem says:
One may always express the sum of a given number of functions, each of
which is multiplied by a rational number, and of which the variable are arbitrary,
by a similar sum of a determined number of functions, of which the variables are
algebraic functions of the variables of the givent functions.
The proof of the first statement is short ( 13, p. 146150). Abel considers an
algebraic equation 0 = p0 + p1 y+ p2 y2 +. . .+ pn1 yn1 + yn = y with coefficients
polynomials in x; this equation is supposed to be irreducible. He introduces another
polynomial
y = q0 + q1 y + q2 y2 + . . . + qn1 yn1
in x, y, certain coefficients a, a , a , . . . of the polynomials q0 , q1 , . . . , q n1 being
indeterminates. The resultant r = y y . . . y(n) of and , where y , y , . . . , y(n)
are the roots of y = 0, is a polynomial in x, a, a , a , . . . , which may be decomposed
in r = F0 xFx where F0 x and Fx are polynomials in x and F0 x does not depend
on a, a , a , . . . . Let x1 , x2 , . . . , x be the roots of Fx = 0 and y1 , y2 , . . . , y
the corresponding common roots of the equations y = 0, y = 0. The yk are
rational functions of xk , a, a , a , . . . by the theory of elimination. Now let f(x, y)dx
be a differential form, with f a rational function of x, y. When Fx = 0, dx =
Fx
F x where F x is the derivative with respect to x and Fx the differential with
respect to a, a , a , . . . Thus f(xk , yk )dxk = f(xFk x,ykk ) Fxk is rational with respect

to xk , a, a , a , . . . and dv = f(x1 , y1 )dx1 + f(x2 , y2 )dx2 + . . . + f(x , y )dx is

rational with respect to a, a , a , . . . A consequence is that the function

f(x1 , y1 )dx1 + f(x2 , y2 )dx2 + . . . + f(x , y )dx = v (53)
is an algebraic and logarithmic function of a, a , a , . . . Now if there are indetermi-

nate coefficients a, a , a , . . . in y, they may be determined by arbitrarily choosing
couples (x j , y j ) of roots of y = 0 and writing the equations y j = 0; the other
yk are then rational functions of xk and the (x j , y j ).
Abel gives a cleaver way to do the computation ( 4, p. 150159), first writing
Fx = Fr0 x = Fry
0 xy
and
f(x, y)dx

1 r r (n) r (n)
= f(x, y ) y + f(x, y ) y + . . . + f(x, y ) (n) y
F0 xF x y y y
1 r
= f(x, y) y
F0 xF x y
where the sum is extended to the n roots y , y , . . . , y(n) . Then he writes f(x, y) =
f 1 (x,y)
f 2 x y where f 1 (x, y) is a polynomial in x, y, of degree n1 in y, f 2 x a polynomial
r
in x and y the derivative of y with respect to y. If f 1 (x, y) y y = R(1) y + Rx
yn1 , with R(1) y polynomial in x, y of degree n 2 in y and Rx a polynomial in
x, it is easy to see that
f 1 (x, y) r
y = Rx. (54)
y y
R1 x
Thus dv = f(x, y)dx = f 2 xF0 xF x ; here the sum is extended to the roots
xk of Fx = 0. Grouping the roots of F0 x and of f 2 x, we obtain dv = 1RxF
1x
x
where R1 x has no common root either with F0 x or with f 2 x and the roots of
1 x annihilate F0 x or f 2 x. If R2 x is the quotient andR3 x the remainder of the
R1 x
Euclidian division of R1 x by 1 x, one computes that F x is the coefficient of
1 R1 x
x in the expansion of 1 xFx in decreasing powers of x; this result comes from
1 1 1
the development of F = x F x in decreasing powers of . A rather more
complicated computation, based on the decomposition of R13xx in simple elements,
R3 x d 1 R1
gives
1 xF x = d1 () where the sum of the right hand side is
1 F
extended to the roots of 1 x and, for each , is the multiplicity of . Unfortunately
this result is incorrect and Sylow corrects it in the notes at the end of the second
volume of Abels Works (p. 295296). The correct result is
R3 x 1 d 1 R1 1

=
1 x F x d 1 (x )Fx
1 x f1 (x,y) y
where x = (x) . From (54), we draw R1 x = F2 x Fx y y where F2 x =
1 x
f2 xis a rational function of x independent of a, a , a , . . . as are 1 x, f 1 (x, y) and
f1 (x,y) y 1 d 1 F2 x f1 (x,y) y
y. Thus dv = F12xx

y y + dx 1 x y y where the
1
symbol denotes the coefficient of x in the expansion of the following function in
decreasing powers of x. Now the expression in the right hand side is integrable and
gives
F2 x f 1 (x, y) 1 d 1 F2 x f 1 (x, y)

v=C log y + log y .
1 x y dx 1 x y
(55)
In general F0 x = 1 and then F2 x = 1, 1 x = f 2 x. If for example f 2 x = (x )m ,
the formula (55) takes the form
f 1 (x, y)dx f 1 (x, y)
m
=C log y
(x ) y (x )m y
d m1 f 1 (, B)

1
+ log B
1 2 (m 1) d m1 B
92 C. Houzel
where B is the value of y when x = (the second term disappears if m = 0).

In the fifth paragraph (p. 159170), Abel studies under which conditions the right
hand side of (54) is a constant independent a , a , . . . He supposes

of a,f1 (x,y)dx that F0 x = 1;
f1 (x,y)
then 1 x = f 2 x must be constant and (x)m y = C y log y. In
order that this expression be constant, he finds that the following condition must
(k) )
be realised: sup h f1(x,y
y(k) < 1 where, for any function R of x, h R denotes the
1kn
highest exponent of x in the expansion of R in decreasing powers of x. This condition
is equivalent to h f 1 (x, y(k) ) < h y(k) 1 for 1 k n and Abel deduces from it
that h(tm y(k) m) < h y(k) 1 for 0 m n 1 and 1 k n if
f 1 (x, y) = t0 + t1 y + t2 y2 + . . . + tn1 yn1 .
A proof of this deduction is given by Sylow in the notes (uvres, t. II, p. 296
297). The condition now takes the form htm < inf (h y(k) mh y(k) ) 1 for
1kn
0 m n 1 and 1 k n. Abel arranges the y(k) in a way such that
h y h y . . . h y(n) . Thus, in general h(y(k) y() ) = h y(k) for > k and
h y(k) = h y + h y + . . . + h y(k1) + (n k)h y(k) . Now one sees that
inf (h y(k) mh y(k) ) = h y + h y + . . . + h y(nm1) ,

1kn
so that htm = h y + h y + . . . + h y(nm1) 2 + nm1 with 0 nm1 < 1.

Let us suppose that
m ()
h y( j) = , (56)
()
an irreducible fraction, for k(1) + 1 j k() , 1 (here k(0) = 0 and

k() = n). Since k() k(1) must be a multiple n () () of () , we have k() =
n + n + . . . + n () () . If k() n m 1 < k(+1) and = n m 1 k() ,
m (+1) + A(+1)

htm = n m + n m + . . . + n () m () 2 + , (57)
(+1)
where A(+1)
= (+1) k() + is the remainder of the division of m (+1) by
m +A
(+1) . For = 1, this shows that tn1 = 0 unless 2. This inequality

signifies that the quotient of m by is 2 or that m < 2 m , the least

possible value of being = E m + 1 (integral part of m + 1). In addition,

one must impose n 1 and < k = n (condition neglected by Abel). Now

if > n 1, m + 1 n and, since n, m is equal to n1 or to n1
1
, which

imposes to y to be of degree 1 with respect to x; in this case, f(x, y)dx = Rdy

with R rational in y, is algebraic and logarithmic in y. Sylow (uvres, t. II, p.
298) observes that the least possible value of is still in the case in which
m + 1 n , with the only exceptions of y of degree 1 with respect to x or

y = y2 + (Ax + B)y + Cx 2 + Dx + E; in these cases, f(x, y)dx may be reduced

to the integral of a rational function and is expressible by algebraic and logarithmic

functions. Finally, the abelian integrals leading to a constant in the right hand side
of (55) are of the form

(t0 + t1 y + . . . + tn 1 yn 1 )dx

y
where the degree htm of each coefficient tm is given by (57). Such a function involves
a number of arbitrary constant equal to = ht0 + ht1 + . . . + htn 1 + n =
ht0 + ht1 + . . . + htn2 + n 1. Using (57), Abel transforms this expression into
A0 m + A1 2m + A2 (n 1)m + An 1
= + + + . . . +

A
m + A1
2m + A2 (n 1)m + An 1

+ 0 + + + . . . +

+nm n
A
0 m + A 1 2m + A 2
(n 1)m + A
n 1
+ +
+
+ . . . +

+ (n m + n m )n
+ . . . . . . . . . . . . . . . . . . . . . n + 1.
() (() 1)
Since A() () ()
0 + A1 + . . . + An () () 1 = n
()
2 and n = n + n + . . . +
n () () , this finally gives
mn 1 m n 1

= n + n m n +
2 2

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

() () (1) (1)
+n m n + m n + ... + m n +
2
() ()
n (m + 1) n (m + 1) n (m + 1)
... + 1. (58)
2 2 2
Abel indicates some particular cases, first the case in which = 1 and
mn 1 m + 1
= n n + 1;
2 2

more particularly, if in addition = n, one has n = 1 and = (n 1) m 21 . In
the second particular case, = = . . . = () = 1 = n = n = . . . = n () and
= n, thus
94 C. Houzel
= (n 1)m + (n 2)m + . . . + 2m (n2) + m (n1) n + 1
where m (k) = h y(k) (1 k n); when h y(k) = h y for 1 k n 1, this gives
nh y

= (n 1) 1 .
2
Abel finally explains that the result remains true for the integrals of the form
f1 (x,y)dx

y even when F0 x is not constant, provided that f1(x,y)
y be finite whenever x
is replaced by a root of F0 x and y by the corresponding value B. In the final notes
(uvres, t. II, p. 298), Sylow says under which precise conditions the number
determined by Abel coincides with the genus p later defined by Riemann: the only
multiple points of the curve defined by the equation y = 0 must be at infinity in the
directions of the axes and no two expansions of the y(k) in decreasing powers of x
may begin by the same term.
As we have said, if there are indeterminate coefficients a, a , a , . . . in y, one
may choose arbitrarily couple (x j , y j ) of common roots to y = 0 and y = 0
and determine a, a , a , . . . by the linear system y j = 0, 1 j . If some
couple (x j , y j ) has a multiplicity k, one must replace the equation y j = 0 by
dy j d k1 y j
y j = = ... = = 0. We get a, a , a , . . . as rational functions of the
dx j dx k1
j
(x j , y j ) and we may substitute these functions in Fx. Abel ( 67, p. 170180) writes
Fx = B(x x1 )(x x2 ) . . . (x x )F (1) x where F (1) x
is a polynomial of degree
with coefficients rational in the (x j , y j ) and x = f(x, y)dx. According to
(53),
1 x1 + 2 x2 + . . . + x = v (+1 x+1 + . . . + x )
where x+1 , . . . , x are the roots of F (1) x = 0, so algebraic functions of x1 , . . . , x ,
and v is an algebraic and logarithmic function.
Now is of the form hq0 + hq1 + . . . + hqn1 + n 1 h F0 x + A with
0 A h F0 x and = hr h F0 x = hy + hy + . . . + hy(n) h F0 x. Thus
= hy + hy + . . . + hy(n) (hq0 + hq1 + . . . + hqn1 ) n + 1 A.
For any m, hy h(qm ym ) = hqm + mh y and, according to (56),
m ()
hy( j) hqm + m when k(1) + 1 j k() . (59)
()
Let us suppose that the maximum value of h(qm y( j)m ) for nk() m nk(1) 1
and k(1) + 1 j k() is obtained for m = :
m () m ()
hq + hq n1 + (n 1) or
() ()
m ()
hq hqn1 (n 1 ) ()

()
for k(1) k() 1. Thus hq hqn1 = (n 1 ) m() + () ()
+ A
where () ()
is a natural integer and 0 A < 1. The sum of these equations for
fixed and variable gives
m ()

() ()
n hq + ()

1
= (2n k() k(1) 1)n () () + A() () ()
0 + A1 + . . . + An () () 1
2
+ () () ()
0 + 1 + . . . + n () () 1 + hqn1k(1) + . . . + hqnk()
1
= (2n k() k(1) 1)n () ()
2
1
+ n () (() 1) + C + hqn1k(1) + . . . + hqnk()
2
where C = () () ()
0 + 1 + . . . + n () () 1 . Let us write the inequalities (59) for
k(1) + 1 j k() , m = and then sum up all these inequalities for variable.
This gives
hy + hy + . . . + hy(n)
hqn1 + hqn2 + . . . + hq0

1 1
+ (2n k() k(1) 1)n () () + n () (() 1) + C
=1
2 2
or hy +hy +. . . +hy(n) (hq0 +hq1 +. . . +hqn1 ) +C1 +C2 +. . . +C

where

1

n 1
= nm + n + n + . . . + n () () + n
2 2

1

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

n 1
+ n (1) m (1) + n () () + n (1)
2 2
() () ()
n 1 1
+ n () m () + n () .
2 2
We finally obtain n + 1 A + C1 + C2 + . . . + C and we remark
that, according to (58),
n + 1 = ,
so that A + C1 + C2 + . . . + C .
As Abel notes it, = A when C1 + C2 + . . . + C = 0 and, for each ,
96 C. Houzel
() m ()
hyk = hq + . (60)
()
He shows that, for a convenient choice of y, these conditions are realised. The first
one signifies that ()
= 0 for k
(1)
k() 1 and 1 or that
m ()
hqn1 = hq (n 1 ) A()
(61)
()
for k(1) k() 1. The degrees hqm will then be definite if we know the hq
and we have by (60),
m () m ()
hq + hq + (62)
() ()
m ()
for any and any . Abel puts ()
= and deduces from the preceding inequality
(1 ) hq hq1 (1 )1 .
Thus hq hq1 = (1 )(1 1 + (1 1 ) ) where 0 1 1,

and hq = hq1 + (1 2 )(1 1 + (1 1 )2 ) + (2 3 )(2 2 + (1 2 )3 ) +
. . . + (1 )(1 1 + (1 1 ) ). Inversely, for any choice of hq1 and
of the (between 0 and 1), these values of hq verify the inequalities (62). It is
then possible, using (61) and some work, to prove that (60) is verified; a narrower
limitation is imposed to the .
All this discussion was made in the hypothesis that the only condition limiting
the indetermination of the coefficients of the qm was that the polynomial F0 x divides
the resultant r. When more conditions are imposed to limit the number of the
indeterminate coefficients a, a , a , . . . , the minimum value of may be of the
form A B < A. In the final notes (uvres, t. II, p. 299300), Sylow explains
that A is the reduction due to the presence of singularities at a finite distance on the
curve y = 0 and that the additional reduction B is due to the eventual coincidence
of the initial terms in some of the y(k) . Moreover he explains how Abels formula
(58) may lead to a computation of A.
In the following paragraph 8 (p. 181185), Abel explicitly deals with the case
where y is of degree n = 13 in y, the degrees in x of the coefficients pm being
2 for m = 0, 2, 8; 3 for m = 1, 3, 6, 9; 4 for m = 4, 7, 10; 5 for m = 5 and
1 for m = 11, 12. He determines the exponents h y(k) by a method similar to

that of the Newton polygon and finds h y = h y = h y = m = 43 , n = 1;
m 1
h y(4) = h y(5) = h y(6) = h y(7) = h y(8) = = 5, n = 1; h y(9) = h y(10) =
m m
h y(11) = h y(12) = = 1
2 , n = 2; h y
(13)
= = 1, n = 1. These values

give = 38 and the limitations 10 1 12, 5 2 9, 1 3 4 and 4 = 0.
Choosing for instance 1 = 11, 2 = 6, 3 = 4, he finds A0 = 23 , A2 = 13 , A3 = 25 ,
A4 = 35 , A5 = 45 , A7 = 15 , A 1 1 12
9 = 2 , A10 = 0, A11 = 2 and then 85 1 17 ,
8
5 1
14 2 1 and 2 3 1. The values of the differences
hq6 hq11 , hq4 hq6 , hq0 hq4
are correspondingly limited and they may be: hq6 hq11 = 2, 3; hq4 hq6 = 0;
hq0 hq4 = 3, 2. It is now possible to determine all the degrees hqm knowing
= hq12 . The possible values of are 13 + 47, 13 + 48, 13 + 57 or 13 + 58.
The corresponding values of are 13 + 85, 13 + 86, 13 + 95 and 13 + 96. Thus
= 38 = for every choice.
Then ( 9, p. 185188) Abel extends his relation (53) in the form
h 1 1 x1 + h 2 2 x2 + . . . + h x = v
where the coefficients h 1 , h 2 , . . . , h are rational numbers. In the paragraph 10

(p. 188211), he deals with the
case in which y = yn + p0 where p0 is a polynomial
f 3 xdx
in x and the integral x = ym f 2 x where f 2 x and f 3 x are polynomials in x.

If p0 = r1 1 r2 2 . . . r where the polynomials r1 , r2 , . . . , r are squarefree and
1 2
relatively prime by pairs, let us put with Abel R = r1 n r2 n . . . r n , so that the
determinations of y are y(k) = k1 R(1 k n), being a
primitive n-th root of
1. The determinations of the integral x are of the form em Rf3mxdx
f 2 x where e is an
integer and (53) takes the form
e1 m x1 + e2 m x2 + . . . + e m x
2 x 1 d 1 F2 x 2 x

(63)
=C +
f2 x dx 1 x
where
f3 x
2 x = (log R + m log (R) + 2m log (2 R)
Rm
+ . . . + (n1)m log (n1 R)). (64)
Let us first suppose that all the coefficients in q0 , q1 , . . . , qn1 are indeterminate,
so that = hq0 + hq1 + . . . + hqn1 + n 1. In our case, h y = h y = . . . =

h y(n) = m . We have = 1 and n = n = k . Let us determine the minimum value
of . According to the relation (61),
m
hqm = hq1 + (1 m) Am (65)

with 0 Am < 1. Here the number is
hr = nhq1 + n m 1 (66)

and, according to (58), = = n n m2 1 n m 2+1 + 1 = n1 n+n
2 nh R 2 + 1.
But this value can be lowered by a more convenient choice of y. For 1 m
and 0 n 1, Abel puts m = E nm + E 2nm + . . . + E (n1) n
m
and
m m

m, = m E n n where E denotes the integral part of the following
98 C. Houzel
fraction and the m are natural integers. He takes the coefficients q of y of the

form q = v r11, r22, . . . r , where the v are polynomials in x. Then q R =

+ 1 + 2 + k k k
v r11 n r22 n . . . r n R() where R() = r11, r22, . . . r , , km, = mnm
( denotes the excess of the following fraction over its integral part), and

+ n1 2 + n2 + n
y(e+1) = (x, e)r11 r2 . . . r
where (x, e) = v0 R(0)+e v1 R(1)+2e v2 R(2)+. . .+(n1)e vn1 R(n1) , 0 e n 1.

n + n +
This gives F0 x =r1 1 1 r2 2 2 . . . rn+ and Fx = (x, 0) (x, 1) . . . (x, n 1).
f3 x ry f3 x Fx (x,e)
Now (54) takes the form Rx = ym y = F0 x y(e+1)m (x,e)
and yf3mx = sfxm
where
m m m m
E n 1 E n 2 E m
n
n1 n2 m
n

fx = f 3 x r1 r2 . . . r and sm = r1 r2 . . . r .
Thus

F0 x fx Fx Fx
Rx =
(x, 0) + m (x, 1)
sm (x, 0) (x, 1)
Fx
+ 2m (x, 2) + . . .
(x, 2)

Fx
+ (n1)m (x, n 1) .
(x, n 1)
Fx
Since the (x,e) (x, e) are polynomial in x, R(0) , R(1) , . . . , R(n1) , so linear com-
binations of the sm with coefficients polynomial in x, it results that F0 x divides Rx:
Rx = F0 x R1 x. Now one sees that F2 x = 1, 1 x = f 2 x and that (64) takes the form
fx
2 x = (log (x, 0) + m log (x, 1) + 2m log (x, 2)
sm
+ . . . + (n1)m log (x, n 1)).
Here
= hr h F0 x
= nhq1 + n m 1 ((n1 + 1 )hr1 + (n2 + 2 )hr2 + . . . + (n + )hr )
where n m = nh R = 1 hr 1 + 2 hr 2 + . . . + hr . Thus, putting for 1 ,
= nhq + (1 n1 1 )hr 1 + (2 n2 2 )hr 2

+ . . . + ( n )hr
= nhv + (n1, n1 + 1 1 )hr 1 + (nys n2 + 2 2 )hr 2
+ . . . + (n, n + )hr
1 1 2 2
= nhv + n hr1 + n hr2 + . . . + n hr .
n n n
On the other hand

= hv0 + hv1 + . . . + vn1 + n 1

= hq 0 + hq 1 + . . . + qn1 + n 1 (m,0 + m,1 + . . . + m,n1 )hrm
m

n(n 1) m

= n hv + m, hrm + n
m
2
n ( 1)
(m,0 + m,1 + . . . + m,n1 )hrm .
2 m
Abel computes m,0 + m,1 + . . . + m,n1 = m + (n 1)m and finally gets

1 1 n k1 2 2 n k2
= nhv + n hr1 + n hr2
n 2 n 2
n + n

n k
+ . . . + n hr 1 +
n 2 2
where km is the g.c.d. of m and n . This gives
n k1 n k2 n k n + n
= hr1 + hr2 + . . . + hr + 1 = (67)
2 2 2 2
independent of , 1 , 2 , . . . , and we have = nhv + nh R() (cf. (66)). The
degrees hvm are determined from hq m = 1,m hr 1 + 2,m hr 2 + . . . + ,m hr + vm
and (65) which give
m
hqm = hv + ( m) + (1, 1,m )hr1 + (2, 2,m )hr2

+ . . . + (, ,m )hr Am
= hv + E((k1, k1,m )hr 1 + (k2, k2,m )hr 2
+ . . . + (k, k,m )hr )
R()
= hv + Eh . (68)
R(m)
Abel adopts new notations: x+1 = z 1 , x+2 = z 2 , . . . , x = z ; e+1 = 1 , e+2 =
2 , . . . , e = ; e = and = and he rewrites (63) in the form
m m m m m m
1 x 1 + 2 x 2 + . . . + x + 1 z 1 + 2 z 2 + . . . + z (69)
fxx 1 d 1
fx x

=C +
sm (x) f 2 x dx 1 sm (x)x
where 1 (x) = f 2 x = A(x 1 )1 (x 2 )2 . . . , fx is an arbitrary polynomial,
x = log (x, 0) + m log (x, 1) + 2m log (x, 2)
+ . . . + (n1)m log (x, n 1)
100 C. Houzel

fxdx
and x = f 2 xsm (x) . Here, x1 , x2 , . . . , x are considered as independent variables

and z 1 , z 2 , . . . , z are the roots of the equation (zx(z,0) (z,1)... (z,n1)
1 )(zx 2 )...(zx )
= 0. The

coefficients a, a , a , . . . are determined by the equations (x1 , e1 ) = (x2 , e2 ) =
. . . = (x , e ) = 0 and the numbers 1 , 2 , . . . , by (z 1 , 1 ) = (z 2 , 2 ) =
. . . = (z , ) = 0.
Some particular cases are explicited by Abel, first the case in which f 2 x =
(x ) with, for instance, = 1 or 0. In this last case the right hand side of (69)
reduces to
fx x
C
sm (x) f 2 x
which is constant when h fx E(hs m (x)) 2.
When n = 1, there is only one sm = s0 = 1 and x = fxdx

(0)
f 2 x . Then R = 1,

(x, 0) = v0 and x = log v0 . The relation (68) takes the form
x1 + x2 + . . . + x + z 1 + z 2 + . . . + z
fx 1 d 1 fx
=C log v0 + log 0
f2 x dx 1 x
where v0 (x) = a(x x1 )(x x2 ) . . . (x x )(x z 1 )(x z 2 ) . . . (x z ), but it
is possible to make = 0 in (67). For = 1, one finds the known integration of
rational differential forms.
1 1 1
When n = 2 and R = r12 r22 , take 1 = 1 and 2 = 0. Then s0 = 1, s1 = (r1r2 ) 2 ,
1 1 1 1 1 1
R(0) = r12 , R(1) = r22 , (x, 0) = v0r12 + v1r22 , (x, 1) = v0r12 v1r22 and = 1.
1 1
v0 r12 +v1 r22
For m = 1, we find x = log 1 1 and, writing 0 x and 1 x respectively for
v0 r12 v1 r22
r1 and r2 , (69) takes the form

fx v0 0 x + v1 1 x
x + z = C log
f 2 x 0 x1 x v0 0 x v1 1 x
1 d 1
fx v0 0 x + v1 1 x
+ log
dx 1 x 0 x1 x v0 0 x v1 1 x

where x = f2 x fxdx

0 x1 x , v0 and v1 are determined by the equations v0 0 x 1 +

1 v1 1 x1 = v0 0 x2 + 2 v1 1 x2 = . . . = 0 and z 1 , z 2 , . . . , z by

(v0 (z))2 0 x(v1 (z))2 1 x v (z ) z
(zx1 )(zx2 )...(zx ) = 0. The signs k are given by k = v01 (zkk )01 zkk . We have

k1 = k2 = 1, = 12 hr1 + 12 hr2 n2 = 12 (h(r1r2 ) n ) where n is the g.c.d. of 2
and h(r1r2 ). Thus = m 1 for h(0 x1 x) = 2m 1 or 2m. Taking = 1, we
have by (68)
hv1 + 12 (h1 x h0 x) 1

1 2
hv0 = v1 + E (h1 x h0 x) =
2 hv1 + 12 (h1 x h0 x)
depending on whether h(0 x 1 x) is odd or even. When m = 1, = 0 and

x = f2fxdx

x R where R is a polynomial of degree 1 or 2. This integral is an algebraic

and logarithmic function and Abel explicits the computation, taking 0 x = 0 x + 0 ,

1 x = 1 x + 1 , f 2 x = (x ) , v1 = 1 and v0 = a. When m = 2, = 1 and
h(0 x 1 x) = 3 or 4 so that x is an elliptic integral. The relation (69) takes the form
1 x1 +2 x2 +. . .+ x = v1 z 1 where v is algebraic and logarithmic. The
product of the roots of the polynomial (v0 z)2 0 z (v1 z)2 1 z = A + . . . + Bz +1
+1
is x1 x2 . . . x z 1 , whence z 1 = BA x(1)x ...x , where BA is a rational function of x1 ,
1 2
x2 , . . . , x , 0 x1 , 0 x2 , . . . , 0 x , 1 x1 , 1 x2 , . . . , 1 x . When
0 x = 1, 1 x = 0 + 1 x + 2 x 2 + 3 x 3 , v1 = 1 and v0 = a0 + a1 x,

we must write v0 x1 = 1 1 x1 , v0 x2 = 2 1 x2 , whence

1 x2 1 x1 2 x1 1 x2 2 1 x2 1 1 x1
a0 = , a1 = .
x1 x2 x1 x2

x 2 x +x 2 x 2 x x x x
Then A = a02 0 , B = 3 and z 1 = 3 x11 x2 2 1 1 1 1 2(x x1 )22 1 2 1 1 1 2 0
1 2
which gives the addition theorem for elliptic integrals.
When m = 3, = 2 and h(0 x 1
x) = 5 or 6. Abel explains certain particular
cases, for instance that in which x = (A0 +A1 x)dx 6 , which gives
0 +1 x+...+6 x
x1 x2 . . . x = z 1 z 2 + C,
where z 1 , z 2are the

roots of aquadratic equation with coefficients rational in x1 ,
x2 , . . . , x , R1 , R2 , . . . , R (where Rk is the value of R corresponding to
x = xk ). As we have said in our 1, Abel explained this result, with = 3, in a letter
to Crelle.
The last example dealt with by Abel is not hyperelliptic for he takes n = 3,
1 2 1 2 2 1
R = r13 r23 , 1 = 2 = 0. Then s0 = 1, s1 = r13 r23 , s2 = r13 r23 , R(0) = s0 , R(1) = s1 ,
R(2) = s2 and
1 2 2 1 1 2 2 1
(x, 0) = v0 + v1r13 r23 + v2r13 r23 , (x, 1) = v0 + v1r13 r23 + 2 v2r13 r23 ,
1 2 2 1
(x, 2) = v0 + 2 v1r13 r23 + v2r13 r23 ,
which give Fx = (x, 0) (x, 1) (x, 2) = v03 + v13r1r22 + v23r12r2 3v0 v1 v2r1r2 .
Here
3 + n
= hr 1 + hr 2 + 1
2

where n is the g.c.d. of 3 and r1 + 2hr 2 . Thus = h(0 x1 x) 2 if hr 1 + 2hr 2 is
divisible by 3 and = h(0 x1 x) 1 in the contrary case.
Since the french Academy did not give any news of his memoir, Abel decided
to published his theorem for the particular case of hyperelliptic integrals in Crelles
102 C. Houzel
Journal (vol. 3, 1828, uvres, t. I, p. 444456). It is this publication which inspired

Jacobi for the formulation of the inversion problem (1832). In a letter to Legendre
(14 March 1829), Jacobi said of Abel theorem that it was perhaps the most impor-
tant discovery of what the century in which we live has made in mathematics . . .
though only a work to come, in a may be distant future, may throw light on its full
importance. The statement is the following: Let x be a polynomial in
x, decom-
fxdx
posed in two factors 1 x, 2 x and let fx be another polynomial and x = (x)
x
where is an any constant quantity. Let us designate by a0 , a1 , a2 , . . . , c0 , c1 , c2 , . . .
arbitrary quantities of which at least one is variable. Then if one puts
(a0 + a1 x + . . . + an x n )2 1 x (c0 + c1 x + . . . + cm x m )2 2 x
= A(x x1 )(x x2 ) . . . (x x )
where A does not depend on x, I say that
1 x1 + 2 x2 + . . . + x

f (a0 + a1 + . . . + an n ) 1 + (c0 + c1 + . . . + cm m ) 2
= log
(a0 + a1 + . . . + an n ) 1 (c0 + c1 + . . . + cm m ) 2
+r +C
1
where C is a constant quantity and r the coefficient of x in the expansion of

fx (a0 + a1 x + . . . + an x n ) 1 x + (c0 + c1 x + . . . + cm x m ) 2 x
log
(x ) x (a0 + a1 x + . . . + an x n ) 1 x (c0 + c1 x + . . . + cm x m ) 2 x
in decreasing powers of x. The quantities 1 , 2 , . . . , are equal to +1 or to 1

and their values depend on those of the quantities x1 , x2 , . . . , x .
Putting x = a0 + a1 x + . . . + an x n , 1 x = c0 + c1 x + . . . + cm x m and Fx =
(x)2 1 x (1 x)2 2 x, the quantities x1 , x2 , . . . , x are the roots of Fx = 0. We
have F xdx + Fx = 0 where
Fx = 2x 1 x x 21 x 2 x 1 x.

Now the equation Fx = 0 implies that x 1 x = 1 x x and 1 x 2 x = x x
fxdx
where = 1 . Thus F xdx = 2(x 1 x 1 x x) x and (x)
x =
2 fx(x1 x1 xx) x
(x)F x where x = (x )1 x + and 1 x are polynomials.
= (x)F x
This leads to
fx dx 1 x 1 x
=
+
= +
(x ) x Fx (x )F x F (x )Fx
(the sums are extended to x1 , x2 , . . . , x ) and then to the relation of the statement.

The values of the k are determined by the equations xk 1 xk = k 1 xk 2 xk .
In a second theorem, Abel explains that the same statement holds in the case in
which some of the roots of Fx are multiple, provided that x 1 x and 1 x 2 x
be relatively prime. The third theorem concerns the case in which f = 0, so that
x = fxdx fx

x where fx is a polynomial (written for x ). In this case, the right hand

side of the relation reduces to

fx x 1 x + 1 x 2 x
C+ log .
x x 1 x 1 x 2 x
On the contrary (theorem IV), when the degree of

( fx)2 is less than the degree of
f 1 +1 2
x, the right hand side reduces to C log 1 1 2 . Abel deals with the
case of the integrals x = (x)dxk x by successive differentiations starting from

k = 1 (theorem V).
The sixth theorem concerns the case in which deg( fx)2 < deg x, that is of

(0 +1 x+...+ x )dx
integrals of the form x =
0 +1 x+...+ x
where = 1 2 1 when is odd

and = 2 2 when is even or = m 2 for = 2m 1 or 2m. In this case,
the right hand side of the relation
rdxis a constant.
The general case of x = x where r is any rational function of x is reduced to
the preceding ones by decomposing r in simple elements (theorem VII). As there are
m + n + 2 indeterminate coefficients a0 , a1 , . . . , c0 , c1 , . . . , Abel arbitrarily chooses
= m + n + 1 quantities x1 , x2 , . . . , x and determines a0 , a1 , . . . , c0 , c1 , . . .

as rational functions of x1 , x2 , . . . , x , x1 , x2 , . . . , x by the equations

xk 1 xk = k 1 xk 2 xk , 1 k . Substituting these values in x and 1 x, Fx
takes the form (xx1 )(xx2 ) . . . (xx )R where R is a polynomial of degree
with the roots x +1 , x +2 , . . . , x . The coefficients of R are rational functions of

x1 , x2 , . . . , x , x1 , x2 , . . . , x . Putting 1 = 2 = . . . = 1 = 1,
1 +1 = 1 +2 = . . . = = 1, x1 +1 = x1 , x1 +2 = x2 , . . . , x = x 2 and
x +1 = y1 , x +2 = y2 , . . . , x = y , Abel rewrites the relation of the statement
in the form x1 + x2 + . . . + x1 x1 x2 . . . x 2 = v +1 y1
+2 y2 . . . y , where x1 , x2 , . . . , x1 , x1 , x2 , . . . , x 2 are independent
variables and y1 , y2 , . . . , y algebraic functions of these variables. He determines the
minimum value of = = m n 1, where = sup(2n + 1 , 2m + 2 ),
1 and 2 denoting the respective degrees of 1 x and 2 x. The mean value of
2n + 1 and 2m + 2 is m + n + 1 + 2
2 . Thus
1 +2
2 1 = 2 1 where
is the degree of , and this minimum value, which is the same as that of the
theorem VI is attained (theorem VIII). The signs j are determined by the equations

y j 1 y j = j 1 y j 2 y j , 1 j . Naturally, when some of the xk or of the xk

are equal, one must replace the corresponding equation xk 1 xk = k 1 xk 2 xk
by a certain number of derivatives of this equation.
The memoir X (uvres, t. II, p. 5566), unpublished by Abel, Sur la comparaison
des transcendantes, gives a testimony of an early form of Abel theorem. It begins
by the same demonstration as in the large memoir for the french Academy, to reach
a relation of the form
x1 + x2 + . . . + x = C + (z 1 + z 2 + . . . + z ) (70)
104 C. Houzel

where x = f(x, y)dx, y being an algebraic function of x and f(x, y) a rational

function of x and y. Here z 1 , z 2 , . . . , z are algebraic functions of x1 , x2 , . . . , x , C
is a constant and is a function algebraic and logarithmic of x1 , x2 , . . . , x . Abel
proposes a method to compute C in the hypothesis in which > ; we shall see
examples of it later.
In the following, he applies his theorem to the particular case in which y is
a rational function of x, defined by the equation + 1 y = 0 where , 1 are
polynomials in x. Then the auxiliary equation y = 0 is of degree 0 in y, of the form
da+xda1 +x n1 dan1
0 = q = a + a1 x + . . . + an1 x n1 + x n = s and ydx = ds 1 . The n
dx
quantities a, a1 , . . . , an1 are determined in function of the n independent variables
x1 , x2 , . . . , xn by writing that these variables are roots of the equation s = 0, and
m+1
= 0. When y = x m , x = xm+1 and the theorem states that
1

(x1m+1 + x2m+1 + . . . + xnm+1 ) = (Pm da + Pm+1 da1 + . . . + Pm+n1 dan1 )
m+1
x1k x2k xnk 1
where Pk = ds1 + ds2 +. . .+ dsn . Now the left hand side is a polynomial m+1 Q m+1
dx 1 dx 2 dx n
Q m+1
1
in a, a1 , . . . , an1 and we thus have Pm+k = m+1 ak . In particular Pk = Q
ak
1
where Q 1 = an1 and this gives

P0 = P1 = . . . = Pn2 = 0, Pn1 = 1, (71)
1
identities used several times by Abel. In the same manner, when y = (x)m ,
1 1
x =
m 1 (x )m1
and

1 1 1 1
+ + . . . +
m 1 (x1 )m1 (x2 )m1 (xn )m1

= (Pm(0) da + Pm(1) da1 + . . . + Pm(n1) dan1 )
x1k x2k xnk

where Pm(k) = ds + ds + ... + dsn . Thus we have
(x1 )m dx1 (x2 )m dx2 (xn )m dx
1 2 n
1 Q m1
Pm(k) =
m 1 ak
where
1 1 1
Q m1 = m1
+ m1
+ ... + .
(x1 ) (x2 ) (xn )m1
When m = 1, x = log(x ) and the left hand side of (70) is
log(x1 )(x2 ) . . . (xn ) = log(1)n (a + a1 + . . . + an1 n1 + n ).
k
Thus P1(k) = a+a n1 +n .
1 +...+an1
Now supposing that s = (a + a1 x + . . . + x1 x 1 + x )x fx where
x = 1 and fx = ( + 1 x + . . . + n1 x n1 ),
(da+xda1 +...+x 1 da1 )

we have ydx = ds and we see that = 0 in (70) if deg <
dx
deg 1 . The quantities x1 , x2 , . . . , xn are related by the equations a + a1 xk + . . . +
1 fxk
a1 xk +xk = x k
, 1 k n. Let x1 , x2 , . . . , xn be another set of quantitites and
suppose that deg < deg 1 ; we have x1 +x2 +. . .+xn = x1 +x2 +. . .+xn .
Now it is possible, by a convenient choice of , 1 , . . . , n1 , to impose xn =

xn1 = . . . = x+1 = 0. Thus the theorem is written
x1 + x2 + . . . + xn = x1 + x2 + . . . + x .
For instance, if = 1 and 1 = x, x = log x and s = + ax + a1 x 2 + . . . +

a1 x + x +1 . Thus = (1)+1 x1 x2 . . . x+1 = (1)+1 x1 x2 . . . x+1
and we

may impose x2 = x3 = . . . = x+1 = 1 to get x1 = x1 x2 . . . x+1 . In this case,
the theorem gives log x1 + log x2 + . . . + log x+1 = log(x1 x2 . . . x+1 ). A second
example is given by = 1, 1 = 1+x 2 , x = arctan x. Let x1 , x2 , x3 be solutions
of the equation 0 = + 1 x + (1 + x 2 )(a + x); we have arctan x1 + arctan x2 +
arctan x3 = C constant and x1 x2 x3 = a, x1 +x2 +x3 = a, x1 x2 +x1 x3 +x2 x3 =
1 + 1. Thus x1 + x2 + x3 x1 x2 x3 = and x1 x2 + x1 x3 + x2 x3 1 = 1 . Now
putting x3 = x2 , x2 = x2 and x1 = x1 , we get C = arctan x1 and x1 + x1 (x2 )2 = ,
x1 +x2 +x3 x1 x2 x3
1 + (x2 )2 = 1 , whence x1 = 1 = 1x 1 x 2 x 1 x 3 x 2 x 3
. Thus the theorem gives
x1 +x2 +x3 x1 x2 x3
arctan x1 + arctan x2 + arctan x3 = arctan 1x 1 x 2 x 1 x 3 x 2 x 3
.
At the end of this memoir, Abel generalises the relations (71). Considering the
integral

fx dx = x + A log(x )
where fx and x are rational functions and the auxiliary equation x = a + a1 x +

. . . + an x n = 0, with the roots x1 , x2 , . . . , xn . By the theorem

fx1 dx 1 + fx2 dx 2 + . . . + fxn dx n

= x1 + x2 + . . . + xn + A log(x1 )(x2 ) . . . (xn ) =

where d = da fx x11 + fx x22 + . . . + fx xnn +da1 x1 xfx11 + x2 xfx22 + . . . + xn xfxn n +
n
x n fx x n fx

x fx
. . . + da1 1 x11 + 2 x22 + . . . + n xnn .
Now x1 + x2 + . . . + xn is a rational function p of a, a1 , . . . , an and
(x1 )(x2 ) . . . (xn ) = (1)n an so that = p + A(log log an )
and
106 C. Houzel

p 1 1 an
= + A
am am am an am
m
x m fx2 x m fxn

x fx1
= 1 + 2 + ... + n .
x1 x2 xn
x m fx x m fx x m fx Am A 1 1
Abel deduces that 1 x11 + 2 x22 + . . . + n xnn = a pm + an 2 2
where the superior sign is taken when m = n and the inferior sign when m < n. For
a
fx = 1, x = x, p = x1 + x2 + . . . + xn = n1 an and A = 0. We find back (71)
and the relation
x1n xn xn an1

+ 2 + . . . + n = 2 .
x1 x2 xn an
1
For fx = x , p = 0 and A = 1; if Fx = + 1 x + . . . + n x n we have
Fx1 Fx2 Fxn n F

+ + ... + =
(x1 ) x1 (x2 ) x2 (xn ) xn an
and other relations by differentiating this one.
6 Elliptic functions
Abel is the founder of the theory of elliptic functions. He partook this glory with
Jacobi alone, for Gauss did not publish the important work he had done in this
field; the grand prix of the parisian Academy of sciences was awarded to Abel
and Jacobi for their work on elliptic functions in 1830, after Abels death. Abels
work on elliptic functions was published in the second and the third volumes of
Crelles Journal (18271828), in a large memoir titled Recherches sur les finctions
elliptiques (uvres, t. I, p. 263388).
Abel briefly recalls the main results of Euler, Lagrange and Legendre on elliptic
integrals and defines his elliptic function = x by the relation

dx
= (72)
(1 c x 2 )(1 + e2 x 2 )
2
0
where c and e are real numbers. This definition is equivalent to the differential
equation
= (1 c2 2 )(1 + e2 2 )

with (0) = 0. Abel puts f = 1 c2 2 and F = 1 + e2 2 and explains
that the principal aim of his memoir is the resolution of the algebraic equation of
degree m 2 which gives , f, F when one knows (m), f(m), F(m) (cf. our
3).
The first paragraph (p. 266278) of Abels memoir is devoted to the study of
the functions , f and F. According to (72), is a positive increasing function
of x for 0 x 1c . Thus is a positive increasing function of for 0 2

1/c
dx
and we have 2 = 1c . Since is an odd function of x,

=
(1c2 x 2 )(1+e2 x 2 )
0
() = (). Now Abel puts ix instead of x in (72) (where i = 1) and gets
x
a purely imaginary value = i, so that xi = (i) where = 2 dx

2 2 2
.
0 (1+c x )(1e x )
1
We see that is a positive increasing function of x for 0 x e and that x is
a positive increasing function of for
1/e
dx
0 =
2 (1 + c2 x 2 )(1 e2 x 2 )
0
and we have 2 = i 1e . Abel notes that the exchange of c and e transforms (i)
i
i
in , f(i) in F, F(i) in f and exchanges and .
The function is known for 2 2 and for = i with 2 2 .
Abel extends its definition to the entire complex domain by the addition theorem:
f F + f F
( + ) = ,
1 + e2 c2 2 2
f f c2 F F
f( + ) = , (73)
1 + e2 c2 2 2
F F + e2 f f
F( + ) = .
1 + e2 c2 2 2
This theorem is a consequence of Euler addition theorem for elliptic integrals,

but Abel directly proves it by differentiating with respect to and using =
f F, f = c2 F, F = e2 f. Thus, denoting by r the right hand
r r
side of the first formula, he finds that = which shows that r is a function of
+ . As r = when = 0, this gives r = ( + ). From (73), Abel deduces
2 f F 2 f F
(+)+() = , (+)() = , (74)
R R
2 f f 2c2 F F
f(+)+ f() = , f(+) f() = ,
R R
2F F 2e2 f f
F(+)+ F() = , F(+) F() =
R R
and
2 2 f 2 c2 2 F 2
( + )( ) = , f( + ) f( ) = (75)
R R
F 2 + e2 2 f 2
F( + )F( ) =
R
108 C. Houzel
where R = 1 + e2 c2 2 2 .
On the other hand f 2 = F 2 i = 0 give

1 f
= , f = e2 + c2 , (76)
2 c F 2 F

e2 + c2 1 i F
F = , i = ,
2 c F 2 e f

2 2
e2 + c2 1
F i = i e + c , f i = .
2 f 2 e f
These relations imply that

+ = , f + =f , (77)
2 2 2 2
F + = F , i + = i ,
2 2 2 2

F i + = F i , f i + = f i
2 2 2 2

2 2
and 2 + 2 i = cei , F 2 = F e c+c = f 2 i f. We

deduce that 2 + 2 i = f 2 + 2 i = F 2 + 2 i = 10 i.e. infinity. From (77)

we have
( + ) = = ( + i), f( + ) = f = f( + i), (78)

F( + ) = F = F( + i)
and
(2 + ) = = (2i + ) = ( + i + ), (79)
f(2 + ) = f = f(i + ), F( + ) = F = F(2i + ).
Thus the functions , f, F are periodic:
(m + ni ) = (1)m+n , f(m + ni ) = (1)m f, (80)

F(m + ni ) = (1)n F.
The equation ( + i) = 0 is equivalent to f(i)F(i)+(i) fF

1+e2 c2 2 2 (i)
= 0 (cf.
(73)) and, as , f(i), F(i) are real and (i) is purely imaginary, this signifies
f(i)F(i) = 0 and (i) f F = 0. These equations are satisfied by =
(i) = 0 or by f(i)F(i) = fF = 0. The first solution gives = m, = n
and it fits, for (m + ni) = 0. The second solution gives = m + 12 , =
n + 12 and it does not fit, for m + 12 + n + 12 i = 10 . In the same way,

Abel determines the roots of the equation fx = 0, which are x = m + 12 + ni

and those of the equation Fx = 0, which are x = m + n + 12 i. From these

results and the formulae

i 1 e2 + c2 1 e2 + c2 1
x =
, fx =
, Fx = ,
F x 2

ec x 2 2 i e f x 2i
c
(81)
he deduces the poles of the functions x, fx, Fx, which are x = m + 12 +

n + 12 i.

2( xa ) f ( x+a ) F ( x+a
2 )
From (74) x a = 1+e2 c22 2 x+a2 2 xa . Thus the equation x = a is
( 2 ) ( 2 )
equivalent to 2 = 0 or f 2 = 0 or F x+a = 10 or
xa x+a
= 0 or xa

x+a 2 2
m+n
2 = 0. Thus the solutions are x = (1) a + m + ni. In the same way,
the solutions of fx = fa are given by x = a + 2m + ni and those of Fx = Fa
by x = a + m + 2ni.
The second paragraph (p. 279281) of Abels memoir contains the proof by
complete induction that (n), f(n) and F(n) are rational functions of , f
P P
and F when n is an integer. Writing (n) = QPnn , f(n) = Qnn and F(n) = Qnn
where Pn , Pn , Pn and Q n are polynomials in , f and F, we have, by (74)
Pn
Pn+1 Pn1 2 f F Q n Pn1 (Q 2n + c2 e2 x 2 Pn2 ) + 2Pn Q n Q n1 yz
= + 2
=
Q n+1 Q n1 1 + e2 c2 2 QPn2 Q n1 Rn
n
where x = , y = f, z = F and Rn = Q 2n + e2 c2 x 2 Pn2 , and we conclude that
Q n+1 = Q n1 Rn , Pn+1 = Pn1 Rn + 2yz Pn qn Q n1 .

In the same way Pn+1 = Pn1 Rn + 2yPn Q n Q n1 and Pn+1 = Pn1 Rn +
2yPn Q n Q n1 . These recursion formulae, together with y = 1 c x and z 2 =
2 2 2

P2n+1
P2n+1
P2n P2n+1
1 + e2 x 2 , show that Q n ,
xyz , x , P2n ,

y , P2n and z are polynomials in x 2 .
P P
The equations (n) = QPnn , f(n) = Qnn and F(n) = Qnn are studied in
paragraph III (p. 282291). When n is even, here noted 2n, the first equation is
written
(2n) = xyz(x 2 ) = x(x 2 ) (1 c2 x 2 )(1 + e2 x 2 )
or 2 (2n) = x 2 (x 2 )2 (1 c2 x 2 )(1 + e2 x 2 ) = (x 2 ), where x = is one of
the roots. If x = is another root, (2n) = (2n) and, by the preceding
properties,
= ((1)m+ 2n + m + i)).
m
Thus the roots of our equation are = (1)m+ + 2n + 2n i , formula
in which we may replace m and by the remainders of their division by 2n, because
of (80). Abel remarks that, when 0 m, < 2n, all the values of so obtained
are different. It results that the total number of roots is equal to 8n 2 and this is
the degree of the equation, for it cannot have any multiple root. When n = 1,
2 2 4 2 2 2 2 2 2
the equation
is (1 + e c x ) (2) = 4x (1 c x )(1+e x ) and its roots are

, + 2 , + 2 i and + 2 + 2 i .
110 C. Houzel
P
When n is an odd number, here written 2n+1, the equation is (2n+1) = Q2n+1
2n+1
m
and its roots x = (1)m+ + 2n+1 + 2n+1 i where n m, n. The
number of these roots is (2n + 1)2 and it is the degree of the
equation.
For
example
n = 1 gives an equation of degree 9 with the roots , 3 , + 3 ,

3 i , + 3 i , 3 3 i , 3 + 3 i , + 3 3 i

and + 3 + 3 i .

P P
Abel studies in the same way the equations f(n) = Qnn and F(n) = Qnn of

2
which the roots are respectively y = f + 2m m

n + n i and z = F + n + n i ,
(0 m, < n ). Each of these equations is of degree n 2 .
2
m
There are particular cases: P2n = 0, with the roots x = 2n + 2n i
m
(0 m, 2n 1), P2n+1 = 0, with the roots x = 2n+1 + 2n+1 i (n
m, n), Pn = 0, with the roots y = f 2m + 12 n + n i , Pn = 0, with

the roots z = F mn + 2 + 12 i

(0 m, n 1) and Q 2n = 0, with the
1
n
1 i
roots x = m + 2 2n + + 2 2n (0 m, 2n 1), Q 2n+1 = 0 with the
roots x = (1)m+ m + 12 2n+1 + + 12 2n+1
i
(n m, n, (m, )
(n, n)).
P
The algebraic solution of the equations (n) = QPnn , f(n) = Qnn and F(n) =
Pn
Qn is given in paragraph IV (p. 291305). It is sufficient to deal with the case in
which n is a prime number. The case n = 2 is easy for if x = 2 , y = f 2 and
z = F 2 , we have
y2 c2 x 2 z 2 1 2c2 x 2 c2 e2 x 4
f = 2 2 4
= ,
1+e c x 1 + e2 c2 x 4
z 2 + e2 y2 x 2 1 + 2e2 x 2 e2 c2 x 4
F = = .
1 + e2 c2 x 4 1 + e2 c2 x 4
Hence F1
1+ = e2 x 2 , 1 f
= c2 x 2 and z 2 = F+ f 2 F+ f
1+ f , y = 1+F and we draw
f F+1
1 1 f F+ f F+ f
2 = c 1+F = 1e F1
f+1 , f 2 =

1+F , F 2 = 1+ f . From these formulae,

it is possible to express 2n , f 2n , F 2n with square roots in function of , f, F.
Taking = 2 as an example, Abel finds

1 c e2 + c2 c2
= = ,
4 c2 + c e2 + c2 ec

e2

1 2 2

2 2
4
f = e +c c e +c ,F = 1+ 2 = F .
4 e 4 c 2
The case n odd was explained in our 3. The essential point was that the auxiliary
functions such as 1 are rational functions of because of the addition theorem
(73). At the same place, we have dealt with the equation P2n+1= 0 (Vof Abels
memoir, p. 305314)) which determines the quantities x = m+i 2n+1 . We saw
that the equation in r = x 2 is of degree 2n(n + 1) and that it may be decomposed in

2n + 2 equations of degree n of which the coefficients are rational functions of the
roots of an equation of degree 2n + 2. The equations of degree n are all solvable by
radicals, but the equation of degree 2n + 2 is not solvable in general.
In paragraph VI (p. 315323), Abel gives explicit formulae for
((2n + 1)), f((2n + 1)) and F((2n + 1))

in function of the quantities + m+i
2n+1 , f + m+i
2n+1 ,F + m+i
2n+1 .
2 2
Let P2n+1 = Ax (2n+1) + . . . + Bx, P2n+1 = A y(2n+1) + . . . + B y, P2n+1 =
2 2 2
A z (2n+1) + . . . + B z and Q 2n+1 = Cx (2n+1) 1 + . . . + D = C y (2n+1) 1
+ ...
2
+D = C z (2n+1) 1 + . . . + D (an even function). From the equations
2 2 1
Ax (2n+1) + . . . + Bx = ((2n + 1) (Cx (2n+1) + . . . + D),
(2n+1)2 (2n+1)2 1
Ay + . . . + B y = f((2n + 1) (C y + . . . + D ),
2 2 1
A z (2n+1) + . . . + B z = F((2n + 1) (C z (2n+1) + . . . + D ),
considering the sum and the product of the roots, Abel deduces that
n n
A m+ m + i
((2n + 1)) = (1) +
C m=n =n 2n + 1
n n
A m + i
= + . (82)
D m=n =n 2n + 1
In the same way

n n
A

m m + i
f((2n + 1) = (1) f +
C m=n =n 2n + 1
n n
A

m + i
= f + (82 )
D m=n =n 2n + 1
and
n n
A

m + i
F((2n + 1)) = (1) F +
C m=n =n 2n + 1
n n
A

m + i
= F + . (82 )
D m=n =n 2n + 1

The coefficients CA , CA , CA , which do not depend on , are determined by letting
tend towards the pole 2 + 2 i, for they are the respective limit values of
((2n + 1)) f((2n + 1)) F((2n + 1))
, , .
f F
112 C. Houzel
Putting = 2 + 2 i + , where tends towards 0, and using (80) and (81), Abel
n ((2n+1))
1
determines CA = 2n+1 , CA = CA = (1)
2n+1 . Since the limit of when tends
towards 0 is 2n + 1 we find
n n
A 2 m i
2n + 1 = 2
D m=1 2n + 1 =1 2n + 1
n n
m + i 2 m i
2 .
m=1 =1
2n + 1 2n + 1
In the same way, letting tend respectively towards 2 and i 2 we get
n n
n A 2 m 2 i
(1) (2n + 1) = f + f +
D m=1 2 2n + 1 =1 2 2n + 1
n n
m + i m i
f2 + f2 +
m=1 =1
2 2n + 1 2 2n + 1
n n
A 2

m
2 i
= F i+ F i+
D m=1 2 2n + 1 =1 2 2n + 1
n n

2 m + i 2 m i
F i+ F i+
m=1 =1
2 2n + 1 2 2n + 1

from which it is possible to draw the values of DA , DA and DA . Abel further simplifies
the expressions of ((2n + 1)), f((2n + 1)) and F((2n + 1)) as products by the
formulae
2
( + )( ) 1 2
= 2
,
2 1 2 + + i
( 2 2 )
2
f( + ) f( ) 1 f 2 f +
(2 )
2
= 2
,
f 2 + 1 f 2 +f + i
( 2 2 )
2
F( + )F( ) 1 F 2 F i+
(2 )
= (cf. (75) and (81))
F 2 2 i + 2

1 F 2 +F + i
( 2 2 )
and he thus obtains
((2n + 1))
2 2
1 1
n m n i
2 2n+1 2 2n+1
= (2n + 1)
2 2
m=1 1
m

=1 1
i

2 2 +
2 i+ 2n+1 2 2 +
2 i+ 2n+1
2 2
1 1
n n

2 m+i
2n+1 2 mi
2n+1
, (83)
2 2
m=1 =1 1
m+i
1
mi

2 2 +
2 i+ 2n+1 2 2 +
2 i+ 2n+1
f((2n + 1))
2 2
1 f 1 f
n n

m i
f 2 2 + 2n+1 f 2 2 + 2n+1
=(1)n (2n + 1) f
f 2 f 2
m=1 1
m

=1 1
i

f 2 2 + 2 i+ 2n+1 f 2 2 + 2 i+ 2n+1
2 2
1 f 1 f
n
n

f 2 2 + m+i
2n+1 f 2 2 + mi
2n+1
(83 )
f 2 f 2
m=1 =1 1 1

m+i mi
f 2 2 +
2 i+ 2n+1 f 2 2 +
2 i+ 2n+1
F((2n + 1))
2 2
1 F 1 F
n n

m i
F2
2 i+ 2n+1
F2
2 i+ 2n+1
=(1)n (2n + 1)F
F2 F2
m=1 1
m

=1 1
i

F 2 2 +
2 i+ 2n+1 F 2 2 +
2 i+ 2n+1
F2 F2
n
n 1
m+i
1
mi

F2 2 i+ 2n+1 F2 2 i+ 2n+1
(83 )
F2 F2
m=1 =1 1
m+i
1
mi

F 2 2 +
2 i+ 2n+1 F 2 2 +
2 i+ 2n+1
expressions of ((2n + 1)), f((2n + 1)) and F((2n + 1)) in rational functions
of , f and F respectively. Abel also transforms the last two to have f((2n+1))
f
F((2n+1))
and F in rational functions of .
In his paragraph VII (p. 323351), Abel keeps = (2n + 1) fixed in the
formulae (82) and (83) and let n tend towards infinity in order to obtain expansions
of his elliptic functions in infinite series and infinite products. From (82) with the
help of (81), we have
1
=
2n + 1 2n + 1
n
1 m + m m
+ (1) +
2n + 1 m=1 2n + 1 2n + 1
n
1 + i i
+ (1) +
2n + 1 =1 2n + 1 2n + 1
n n
i
(1)m+ (n m, n )
ec m=1 =1
n n
i
+ (1)m+ 1 (n m, n )
ec m=1 =1
where
114 C. Houzel

1 1 1
(m, ) = +
2n + 1

1 1
+ m+ 2 + + 2 i 1 1
m+ 2 + 2 i
2n+1 2n+1
and

1 1 1
1 (m, ) = + .

2n + 1

1 1
+ m+ 2 + 2 i 1 1
m+ 2 + + 2 i
2n+1 2n+1
Now

+ m m
Am = (2n + 1) +
2n + 1 2n + 1
m m
2 2n+1 f 2n+1 F 2n+1
= (2n + 1) m
1 + e2 c2 2 2n+1 2 2n+1
and

+ i i
B = (2n + 1) +
2n + 1 2n + 1
i i
2 2n+1 f 2n+1 F 2n+1
= (2n + 1)
i

1 + e2 c2 2 2n+1 2 2n+1
n
1 1
(1)m (Am + Bm ) of

remain bounded and the first part 2n+1 2n+1 + (2n+1)2
m=1
has 0 for limit when n tends towards . Thus
n n
i
= lim (1)m+ (n m, n )
ec m=1 =1
n n
i
+ lim (1)m+ 1 (n m, n ).
ec m=1 =1
n1
n1
(1)m+ (m, ) for the second part

It remains to compute the limit of
m=0 =0
will be deduced from
the first by changing the sign of i. We have (m, ) =

2 2n+1 2n+1
1 where = fF and = m + 1 + + 1 i (cf.

2

2n+1 2 2 2
2n+1 2n+1
2
(74) and (75)) and this has for limit (m, ) = 2 when n
2 m+ 12 + + 21 i
n1 n1
(1) (m, ) (1) (m, )

tends towards . Abel tries to prove that
=1 =1
1
is negligible with respect to by estimating the difference (m, ) (m, ),
2n+1
n1
(1) (m, ) by the sum up to

but his reasoning is not clear. Then he replaces
=1

(1) (m, ), again negligible with

infinity using a sum formula to estimate
=n
1
respect to 2n+1 .
He finally obtains

i m

2
= (1) (1) 2
ec m=1 m + 2 + 12 i
1

2
=1

2
2
m + 12 + + 12 i

2

1 (2 + 1)
= (1)m (1) 2 2
ec m=1 m + 12 + + 12 2

=1

(2 + 1)
2 2 . (84)
+ m + 12 + + 12 2

By the same method, Abel obtains

2 + m + 12

1 m
f = (1) 2 2
e =0 m=0 + m + 12 + + 12 2

1

2 m + 2
(1)m 2 2 , (84 )
1
+ 12 2

m=0 m + 2 +

1 (2 + 1)
F = (1) 2 2
c m=0 =0 m + 12 + + 12 2

(2 + 1)
+ (1) 2 2 . (84 )
+ m + 12 + + 12 2

=0
He deals with the formulae (83)

in

the same way by taking the logarithms.

2 2n+1
1
m+i+k

1 2
2 (m+i+k)2
For any constants k and , 2n+1 has a limit equal to . Abel

2 2n+1 1 2
1 m+i+ (m+i+)2
2 2n+1
tries to proves that the difference of the logarithms (m, ) and (m, ) of these
1
expressions is dominated by (2n+1) 2 , with the difficulty that m and vary in the
n n

sum to be computed. He deduces that the difference (m, ) (m, ) is
=1 =1
116 C. Houzel
n
1
negligible with respect to 2n+1 and replaces (m, ) by (m, ). The proof
=1 =1

1
that (m, ) = (m, + n) is negligible with respect to 2n+1 is based on
=n+1 =1
the expansion of (m, + n) in powers of but it is not sufficient. Abel finally gets
n
n

lim (m, ) = (m, ). He deals in the same way with the simple
m=1 =1 m=1 =1
products in (83) and obtains

2 2

= 1 1 +
m=1
(m)2 =1 ()2

2 2
1 (m+i) 2 1 (mi) 2
,
2 2
1 1

m=1 =1 2 =1 2
m 12 + 12 i m 12 12 i

2
f = 1 2
m=1 m 12 2
2 2

1 2 1 2
1
m 2 +i 1
m 2 i
,
2 2
m=1 =1 1 2 1 2
m 12 + 12 i m 12 12 i

2
F = 1+
1 2

=1 2 2
2 2

1 2 1 2
m+ 12 i m 21 i
.
2 2
m=1 =1 1
2 1 2
m 2 + 12 i
1 m 2 12 i
1
Abel also writes these formulae in a real form.

The Eulerian products for sin y and cos y lead to
z2
2 1 2
1 2z 2

sin z 12 2 cos z
2
= and =
z cos y y2 cos y
=1 1 y 2 =1 1 2
1
2 2 12 2
and this permits to transform the double products of Abels formulae in simple
products:

sin i 2

= 1 2 2
i m=1
m
i i
cos2 m 12
i
sin( + m) sin( m)

cos + m 12 i 1
i i
2
m=1 cos m 2 sin m
i 2

m
2 2
( + m)( m) i2
i
sin2
sin i 1
i
2
sin m
= i
i sin2
m=1 1 2

1

i
cos m 2
2
h h
1
1 m
h h m

= h h
2
h h
m=1 1+
1
m 2 m 1
2
h h
4 sin2

1+ m m 2
h h
= sin (85)
m=1 1 4 sin2

(2m1) 2
2 (2m1)
2
h h
where h = 2.712818 . . . is the basis of natural logarithms. In the same way, he

obtains
4 sin2

1+
(2m+1) (2m+1) 2

h h
F = ,
4 sin2

m=1 1
(2m+1) (2m+1)
2
h h
4 sin2

1 m m 2
h +h
f = cos .
4 sin2

m=1 1 (2m1) 2
2 (2m1)
2
h +h
These expansions were known to Gauss and they were independently discovered by
Jacobi, who used a passage to the limit in the formulae of transformation for the
elliptic functions.
1
The expansion of chy in simple fractions gives

(2 + 1)
(1) 2 2
m + 12 + + 12 2

=1
2 1
=
m+ 12 m+ 21
h +h
which permits to transform the formulae (84) in simple series. Thus
118 C. Houzel

2 1
= (1)m
ec m=0 m+ 21 m+ 21
h +h

1

+ m+ 21 + m+ 21
h +h

m+ 12 m+ 12

h h h
h
2
= (1)m 2 2
ec m=0 h + h + h (2m+1) + h (2m+1)

m+ 12 m+ 12
sin
h
h
4
= (1)m (2m+1) (86)
ec m=0 h + 2 cos 2 + h (2m+1)

and
1
m+ 12

m+ 2
h + h h + h
2
F = ,
c m=0 h 2 2
+ h + h (2m+1) + h (2m+1)

m+ 21
m+ 12
cos h +h
4
f = .
e m=0 h (2m+1) + 2 cos 2 + h (2m+1)
In the lemniscatic case, where e = c = 1, one has = and these expansions take
a simpler form
3 3 5 5

h 2 h 2 h 2 h 2 h 2 h 2
=2 3 3
+ 5 5
...
2 h 2 + h 2 h 2 + h 2 h 2 + h 2
h 2 h 32 h 52

4
= sin sin 3 + sin 5 ... ,
2 1 + h 2 1 + h 3 2 1 + h 5
3 3 5 5

h 2 + h 2 h 2 + h 2 h 2 + h 2
F =2 3 3
+ 5 5
... ,
2 h 2 h 2 h 2 h 2 h 2 h 2
h 2 h 32 h 52

4
f = cos cos 3 +cos 5 . . .
2 2 h 1 2 h 3 1 2 h 5 1
3 5
1

h2 h 2 h 2 dx
and, taking = 0, 2 = 2 h 1 h 3 1
+ h 5 1
... = ,
0 1x 4
3 5

2 h2 h2 h2
= 2 3 3 + 5 5 ... .
4 h + 1 h +1 h +1
The second part of Abels memoir, beginning with paragraph VIII (p. 352362),
was published in 1828. This paragraph is devoted to the algebraic solution of the

equation Pn = 0 which gives n in the lemniscatic case and for n a prime number
of the form 4 + 1. Abel announces that there is an infinity of other cases where the
equation Pn = 0 is solvable by radicals.
Here, by the addition theorem (73), (m + i) = (m) f()F()+i() f(m)F(m)
12 (m)2 ()
= T where T is a rational function of ()2 because of the formulae of multipli-
cation. One says that there exists a complex multiplication. Putting = x, we have
(m + i) = x(x 2 ). Now (i) = i = ix and (m + i)i = i(m + i) =
ix(x 2 ) and this shows that (x 2 ) = (x 2 ). In other words, is an even function
and T is a rational function of x 4 . For instance
(2) f F + i f(2)F(2)
(2 + i) = ,
1 (2)2 2

2x 1x 4 12x 2 x 4
where (2) = 1+x 4
, f = 1 x 2 , F = 1 + x 2 , f(2) = 1+x 4
and
1+2x 2 x 4 12ix 4
F(2) = 1+x 4
.
Thus (2 + i) = 4.
Gauss had already discovered the
xi 1(12i)x
complex multiplication of lemniscatic functions and the fact that it made possible
the algebraic solution of the division of the periods. He made an allusion to this
fact in the introduction to the seventh section of his Disquisitiones arithmeticae, but
never publish anything on the subject. We have explained this algebraic solution in
our 3.
The ninth paragraph of Abels memoir (p. 363377) deals with the transformation
of elliptic functions. The transformation of order 2 was known since Landen (1775)
and Lagrange (1784) and Legendre made an extensive suty of it in his Exercices
de calcul integral. Later, in 1824, Legendre discovered another transformation, of
order 3, which Jacobi rediscovered in 1827 together with a new transformation, of
order 5. Then Jacobi announced the existence of transformations of any orders, but
he was able to prove this existence only in 1828, using the idea of inversion of the
elliptic integrals which came from Abel. Independently from Jacobi, Abel built the
theory of transformations. Here is his statement:
If one designates by the quantity (m+)+(m)i
2n+1 , where at least one of the
two integers m and is relatively prime with 2n + 1, one has

dy dx
= a (87)
(1 c21 y2 )(1 + e21 y2 ) (1 c2 x 2 )(1 + e2 x 2 )
2 2 2 2
x )( 2x )( nx ) 2 2
where y = f x (1+e2 c2 (
2 x 2 )(1+e2 c2 2 2x 2 )(1+e2 c2 2 nx 2 ) ,
1 f 2
= + + 2 + n ,
c1 c 2 2 2
2
1 f i i i
= + + 2 + n ,
e1 e 2 2 2
a = f( 2 3 n)2 . (88)
120 C. Houzel
Here f is indeterminate and e2 , c2 may be positive or negative. By (80) (periodicity),

we have ( + (2n + 1)) = or ( + (n + 1)) = ( n). Now if
1 = + ( + ) + . . . + ( + 2n),
we have 1 ( + ) = 1 and 1 admits the period . This function may be written
1 = + ( + ) + ( ) + ( + 2) + ( 2)
+ . . . + ( + n) + ( n)
2 f F 2 f 2 F2
= + +
1 + e2 c2 2 2 1 + e2 c2 2 2 2
2 fn Fn
+ ... + , (89)
1 + e2 c2 2 n 2
a rational function x of x = . Note that the auxiliary function 1 used to solve
the equation of division
in the first part was precisely of this type (see 3).
x
For any , R = 1 1 (1 + e2 c2 2 x 2 ) . . . (1 + e2 c2 2 nx 2 ) is a polynomial
of degree 2n + 1 in x. It is annihilated by x = and so by x = ( + m), m any
integer. Since , ( + ), ( + 2), . . . , ( + 2n) are all different, they are the
roots of R and

x x x
R = A 1 1 1 (90)
( + ) ( + 2n)
where A is found to be 1 by making x = 0. Multiplying by and then making
= 0, we obtain

x x x
1 1 2 1 2n
x = gx
(1 + e2 c2 2 x 2 ) (1 + e2 c2 2 n x 2 )
2
2
2
(91)
1 x2 1 2x2 1 2x2n
= gx
(1 + e2 c2 2 x 2 ) (1 + e2 c2 2 n x 2 )
1
where g = 1 + 2 f F + 2 f 2 F2 + . . . 2 fn Fn is the value of for = 0.
Doing = 2 in R , we have
x
1
1 2

x x x x 1
= 1 1 1 1
2 2 + 2 + 2 2 + 2n
2 2 2
x x x 1
= (1 cx) 1 1 1
2 + 2 + 2 2 + n
where = (1 + e2 c2 2 x 2 )(1 + e2 c2 2 2 x 2 ) (1 + e2 c2 2 n x 2 ). Changing

x in x, we have
2 2
x x x
1 + = (1 + cx) 1 + 1 +
1 2 2 + 2 + 2
2
x 1
1 + .
2 + n
1
Abel puts y = kx, c1 = k1 2
(k a constant),

x x x
t = 1 1 1 ,
2 + 2 + 2 2 + n

x x x
t1 = 1 + 1 + 1 +
2 + 2 + 2 2 + n
2 t2
in order to have 1 c1 y = (1 cx) t and 1 + c1 y = (1 + cx) 1 .
s2 2
In the same way, 1 e1 iy = (1 eix) 1 , 1 e1 iy = (1 + eix) s where
e1 = k i i and
1( 2 )

x x x
s = 1 1 1 ,
2i + 2 i + 2 2 i + n

x x x
s1 = 1 + 1 + 1 + .
2i + 2 i + 2 2 i + n

(1 c21 y2 )(1 + e21 y2 ) = tt1ss2 1 (1 c2 x 2 )(1 + e2 x 2 ). Now dy =
P
Thus 2
dx
t2
where P is a polynomial of degree 4n. Differentiating 1 c1 y = (1 cx) , we see
that
t dt d
P= ct (1 cx) 2 t
c1 dx dx
is divisible by t and, in the same manner, it is divisible by t1 , s and s1 . Since these
four polynomials of degree n cannot have any common factor, it results that tt1Pss1
dy dx
is a constant a and that = a . When x = 0,
(1c21 y2 )(1+e21 y2 ) (1c2 x 2 )(1+e2 x 2 )
t = t1 = s = s1 = 1 = and P = dy dx = k (0)

= kg.
2n+1 1
According to (90), the coefficient of x in R is (+)(+n) and, com-
n g
paring with (91) it is equal to (1)
1 (2n)2
. Thus
(1)n g
1 = ( + ) ( + n).
( 2 n)2
122 C. Houzel
x
Now (89) and (91) give two expressions for the limit of x for x infinite:
g(1)n
1 and .
(ce)2n ( 2 n)4
Thus, by comparison, g = (1)n (ec)2n (2 n)4 and
1 = (ec)2n (2 n)2 ( + ) ( + 2n).
In particular
1
1 = (ec)2n 2
= + + 2n ,
2 kc1 2 2 2
i
1 i = = (ec)2n 2 i i + i + 2n
2 ke1 2 2 2
where = 2 ng The values (88) of the statement for c11 and e11 result
if we put f = k(e2 c2 )n 2 and, as 2 + 2 i + = eci (cf. (76)), we obtain

n (ec)2n+1
c1 e1 = (1) f2
. On the other hand
e1 e 4
= (1)n (ec)2n + + 2 + n ,
c1 c 2 2 2
c1 c 4
= (1)n (ec)2n i + i + 2 i + n
e1 e 2 2 2
and a = kg = (1)n f 2 .
1 1c2 2 2 2
2 2
Using 2 + and 2 i + = e12 1+e

= c2 1+c2 2 1e2 2
(cf. (76)),
1 1
one transforms the expressions of c1 and e1 in rational symmetric functions of ,
2, . . . , n. Reasoning as in his V for the equation P2n+1 = 0, Abel deduces
that, when 2n + 1 is a prime number, c1 and e1 are determined by an equation of
degree 2n + 2 (the modular equation as it was called later). Now such an equation
has roots not necessarily real and Abel says that the theory must be extended to the
case of moduli c, e complex numbers.
2m
When c and e are real, the only values of giving c1 and e1 real are 2n+1 and
2i
2n+1 . The first value gives
2n 1 2

1 f 1 3
= ,
c1 c 2n + 1 2 2n + 1 2 2n + 1 2
2n 1 4

e1 ne 2n 1 3
= (1) (ec) .
c1 c 2n + 1 2 2n + 1 2 2n + 1 2
Abel explains in particular the case in which c = c1 = 1, (1)n = 1 and
0 < e < 1. Then e1 is very small when 2n + 1 is large. Abel carefully
studies
1x 2
the sign in (87). Since 2 is positive for x real, this sign is that of tt1 ss 1 1y2
.
Now ss1 is easily seento

be positive and
thesign we are looking
for is that of
2 2 2
tt 1 = 1 x 1 x 1 x for the radical is
1
2 2n+1 3
2 2n+1 2 2n1
2 2 2n+1 2
1 1
positive. For instance, when 2n+1 2 x 2n+1 2 the sign is + and we

1 dy
get (1)n a = 4n+2 1

by doing x = 2n+1 2 to which corresponds
0 (1y2 )(1+e21 y2 )
y = (1)n . If we neglect e21 , this gives approximately (1)n a = (2n + 1) and
x
dx (1)n
= arcsin y
(1 x 2 )(1 + e2 x 2 ) (2n + 1)
0

2 2
1 x 1 x

2 2n+1 n
2 2n+1
for y = (1)n (2n + 1) x

n x 2
. Abel also explains the
1+e2 2 2n+1 x 2 1+e2 2 2n+1
effect of the other real transformation = 2i 2n+1 . He states that every possible
transformation is obtained by combining the transformations of order 2k studied by
Legendre with his new transformations. He will publish a proof of this statement in
his Precis dune theorie des fonctions elliptiques (1829).
The last paragraph of the Recherches (p. 377388) is devoted to the study of the
dy dx
differential equation 2 2
= a 2 2
and, in particular to the cases
(1y )(1+y ) (1x )(1+x )
in which there is complex multiplication. Abel states two theorems: I. Supposing
a real and the equation algebraically integrable, it is necessary that a be a rational
number.
II. Supposing a imaginary and the equation
algebraically integrable, it is nec-
essary that a be of the form m 1 n where m and n are rational numbers.
In this case, the quantity is not arbitrary; it must satisfy an equation which has an
infinity of roots, real and imaginary. Each value of satisfy to the question.
Here Abel only considers a particular case, that in which e1 = 1e for the first real
transformation. Thus we have
dy dx ey
= a 1 (changing y in ), (92)
2 2
(1 y )(1 + e y ) 2 2 2
(1 x )(1 + e x ) 2 i

2 2n+1 n x 2
x 2 2 2n+1
where y = 1en x

,
n x 2
e being determined by
1+e2 2 2n+1 x 2 1+e2 2 2n+1
2
2n1 2 n
2n+1 2n+1
1
1 = en+1 2n+1 2 2n+1 2 and a by a = 1e 1 2n1
.
2n+1 2 2n+1 2
1

e
From (92), Abel deduces
2 = 2dz 2 2 = a 4n+2
(integration from
0 (1+z )(1e z )

x = 0 to x = 4n+2 , y = z 1) and 2 = a 2 . Thus a = 2n + 1 = .
dy dx
For instance, when n = 1, = 3 where
(1y2 )(1+e2 y2 ) (1x 2 )(1+e2 x 2 )
124 C. Houzel
2 ( )x 2
y= 1ex 1+e2 32 x 2 ,
(3)
1 2 e2 e2 2 3 2 3 1

2
1=e = and a = 2 = 3.
1 + e2 2 3

32 6 e

This gives 3 = e3 and e = 3+2, 3 = 2 33. Changing x in x 2 3

dy dx
and y in y 2 3 1, Abel obtains 2 4
= 3 2 4
where
12 3y y 1+2 3x x
2
y = x 1+3x
2.
3x
dy dx
2 ( )x 2
For n = 2, = 5 1e2 x 1+e2 52 x 2
where y =
(1y2 )(1+e2 y2 ) (1x 2 )(1+e2 x 2 ) (5)
2 25 x
2

, 1 = e2 2 10 2 3 5 = e2 2 5 2 2 2

10 , 5 . Using 10 =

1+e2 2 2
5 .x
2

f 2 2 f 2 ( 5 )
2 5
2
3
and 2 = 2

2 5 = 10 2 5 = F 2 ( 5 )
, Abel finally gets
F 2 2
5

1 1e 5
e1 e = e2 e 5
, which gives a cubic equation for e:

e3 1 (5 + 2 5)e(e 1) = 0.
2
5+1 5+1
This equation has only one solution larger than 1, as e must be, e = 2 + 2 .

5
It is then easy to compute = 5 and = 2 2 2

5 for = e2
and

e3 1 e(e 1) 5 = e2 (e + 1)(2 + 2 ).

x y 1
Changing x in e and y in e , Abel obtains the equation
dy dx
= 5
2

1 4 2 + 5y y4 1 + 4 2 + 5x 2 x 4

5 10+10 5x 2 +x 4
where y = x 2 4.
1+ 10+10 5x + 5x
For higher orders n Abel says that the equation giving the singular modu-
lus e is not necessarily algebraically solvable and he proposes
an expansion of
e in infinite series. He starts from (86) with = 2 , 2 = 1c = 1 and
3 5

gets e = 4 2+1 + 6+1 + 10 +1 + . . . where = h 2 . With = 2 i,
3 5

2 i = ei , Abel gets = 4 r 2r+1 + r 6r+1 + r 10r +1 + . . . where r = h 2

3 2n+1

2n+1
and, since = 2n + 1, = 4 2n + 1 h 2 2n+1 + h322n+1 + . . . and
h +1 h +1
1 3 1

4 h2 2n+1 h 2 2n+1
e= + 3
+ ... .
h 2n+1 +1 h 2n+1 +1
At the end of this memoir, Abel explains how his theory of transformation gives
the formulae published by Jacobi in 1827. Jacobi uses Legendres notations, with
a modulus k between 0 and 1 and the elliptic integral of the first kind F(k, ) =
d , so that if = F(k, ), = sin where is Abels elliptic function

2 2
1k sin
0
n
with c = 1 and e2 = k2 . Writing c1 = 1, e21 = 2 , = (1) n
a , x = (1) sin ,
y = sin and 2n + 1 = p, Abels formula for the first real transformation takes the
form d2 2 = d2 2 + C, where

1k sin 1 sin
2
sin sin sin (2n1)

2n+1 (2n1) 4
=k (sin sin . . . sin ) , =
sin sin sin (2n)
n+ 1 2 2 2 2 2 (2n) 2
k 2 (sin sin )(sin sin )...(sin sin )
and sin = sin (1k2 sin2 sin2 )(1k2 sin2 sin2 )...(1k2 sin2 (2n) sin2 ) , the an-

gles , , . . . , (2n) being defined by sin (m) = mp 2 or F(k, (m) ) = mp 2 . Since

1y 1sin 1(1)n sin sin sin sin +sin
c = c1 = 1, 1+y = tt1 1x1+x or 1+sin = 1+(1)n sin sin +sin sin sin . . .
sin (2n1) +(1)n sin
sin (2n1) (1)n sin
, relation which may be transformed in
tan 12 ( ) tan 12 ( + )

1
tan 45 = . ...
2 tan 12 ( + ) tan 12 ( )
tan 12 ( (2n1) + (1)n )

n1
tan 45 (1) .
tan 12 ( (2n1) (1)n ) 2
In 1828, Abel had begun the redaction of a second memoir to continue the
Recherches sur les fonctions elliptiques (uvres, t. II, p. 244253). Putting =
(m+)+(m)i
2n+1 where m, and n are integers such that m + , m and 2n + 1
have no common divisor, and
1 = ( + )( )(2 + )(2 ) (n + )(n )

2 2 2 2 2 2 n 2
= ,
1 + e2 c2 2 2 1 + e2 c2 2 2 2 1 + e2 c2 2 n 2
Abel remarks that this function is rational in and invariant by . It
results that the roots of the equation
0 = x(2 x 2 )(2 2 x 2 ) (2 n x 2 )
1 (1 + e2 c2 2 x 2 )(1 + e2 c2 2 2x 2 ) (1 + e2 c2 2 nx 2 )
are , ( + ), . . . , ( + 2n). Now let be a rational function of these roots

and suppose that it is invariant by + . Using the addition theorem (73), one
sees that
= 1 + 2 f F
126 C. Houzel
where 1 and 2 are rational in and 1 = 12 ( + ), 2 f F =

1
2 ( ) where is the function deduced from by changing into .
One has = 1 2 f F and is invariant by + . Thus
1 is invariant by + and it is thus a rational symmetric function of ,
( + ), . . . , ( + 2n), that is a rational function of 1 . In the same way,
one
shows that the square of 2 fF is a rational function of 1 , so that = p q
where p, q are rational functions of 1 . Let
= ()2 ( + ) + (( + ))2 ( + 2) + . . .
+(( + (2n 1)))2 ( + 2n) + (( + 2n))2
and let be the function deduced from by changing into . One has
= 1 + 2 f F, = 1 2 f F
where 1 and 2 are rational functions of , and 12 ( ) = 2 f F =

r where r is a rational function of 1 . Now 2 =
is a rational function of
2
by +, so a rational function q of 1 and 12 ( ) = q r,
invariant
= p q r where r does not depend of the function .
Abel proves that r is a polynomial in 1 , for if it had a pole 1 , we should
have = 10 which means that some ( ) would be infinite, but then
1 would also be infinite, which is absurd. The expansions of our functions in
decreasing powers of x = are 1 = ax + , = Ax 2 + where a, A are
constant and and contain powers of x respectively less than 1 and 2. If is the
degree of r, the equation r = 14 ( )2 is rewritten a x + . . . = 14 A2 x 4 + . . .
and it
shows that = 4. Since r must be annihilated by = 2 , 2 i, one has
2 2
r = C 1 1 1 1 i where C is a constant.
12 1 2
When is a polynomial in , ( + ), . . . , ( + 2n), the same rea-

soning shows that p and q are polynomials in 1 of respective degrees and
2 where is the degree of with respect to any one of the quantities ,
( + ), . . . , ( + 2n). If = 1, one has = A + B1 where A and B are
constants respectively determined by making = 0 and = 10 . For instance, let us
put = ( + 1 )( + 2 ) . . . ( + ) and
P = () + ( + ) + ( + 2) + . . . + ( + 2n)
where 1 , 2 , . . . , are distinct integers less than 2n + 1. One has A = () +
(2) + . . . + (2n) and B is the derivative of P for = 0. When is odd (resp.
even), B (resp. A) is equal to 0, for instance = 0 gives
+ ( + ) + ( + 2) + . . . + ( + 2n) = B1
and = 1 gives
( + ) + ( + )( + 2) + . . . + ( + 2n)
= 2 + 2 3 + . . . + (2n 1) 2n.
The second paragraph is not very explicit; Abel considers the functions
2n
2n

= k ( + k), 1 = k ( + k)
k=0 k=0
where is a primitive (2n+1)-th root of 1. Since ( +) = 2n and 1 ( +) =

2n 1 , the product 1 is invariant by + . It is an even polynomial
in the transformed
mi elliptic integral y = 1 (a), of the form A(y2 f 2 ) where
mi
f = 1 a 2n+1 . Thus this product is 0 when = 2n+1 and this gives a remarkable
identity

mi mi 2 mi
0= + + + + 2 + . . .
2n + 1 2n + 1 2n + 1

mi
+2n + 2n
2n + 1
for a convenient m. Abel has announced this type of identity in the introduction of
the Precis dune theorie des fonctions elliptiques, published in 1829 (see our 8);
Sylow and Kronecker have proposed proofs for them.
7 Development of the Theory of Transformation

of Elliptic Functions
The theory of transformation and of complex multiplication was developed by Abel in

the paper Solution dun probleme general concernant la transformation des fonctions
elliptiques (Astronomische Nachrichten (6) 138 and (7) 147, 1828; uvres, t. I, p.
403443), published in the Journal where Jacobi had announced the formulae for
transformation. Abel deals with the following problem: To find all the possible
cases in which the differential equation
dy dx
= a (93)
2 2 2 2
(1 c1 y )(1 e1 y ) (1 c x 2 )(1 e2 x 2 )
2
may be satisfied by putting for y an algebraic function of x, rational or irrational.

He explains that the problem may be reduced to the case in which y is a rational
function of x and he begins by solving this case. His notations are x = when
1
dx

c
, = (1 c x )(1 e x ), 2 = 2 dx

= 2 2 2 2 ,
2 2 2 2
(1c x )(1e x ) 2 2 2 (1c x )(1e x )
0 0
1
e
dx

2 = where e and c may be complex numbers. Abel recalls the

0 (1c2 x 2 )(1e2 x 2 )

addition theorem ( ) = 1c2 e2 2 2
and the solution of the equation = ,
m+m
which is = (1)
+m +m . Let y = (x) be the rational function we are
128 C. Houzel
looking for and x = , x1 = 1 two solutions of the equation y = (x), y being

given (it is supposed that this equation is not of the first degree). From the equation
dy
R
= ad = ad1 , we deduce d1 = d. Thus 1 = where is constant
and x1 = ( ), where we may choose the sign +, for ( ) = ( + ).
Now y = () = (( + )) = (( + 2)) = . . . = (( + k)) for any
integer k. As the equation y = (x) has only a finite number of roots, there exist
k and k distinct such that ( + k) = ( + k ) or ( + n) = where

n = k k (supposed to be positive). Then + n = (1)m+m + m + m and,

necessarily, (1)m+m = 1, n = m + m or = + where , are
rational numbers. If the equation y = (x) has roots other than ( + k), any one
of them has the form ( + 1 ) where 1 = 1 + 1 (1 , 1 rational) and all
the ( + k + k1 1 ) are roots of the equation. Continuing in this way, Abel finds
that the roots of y = (x) are of the form
x = ( + k1 1 + k2 2 + . . . + k )
where k1 , k2 , . . . , k are integers and 1 , 2 , . . . , of the form + (,

rational). The problem is to determine y in function of , the quantities 1 , 2 , . . . ,
being given.
Before the solution of this problem, Abel deals with the case in which y =
f + fx g c1 f +(gc1 f )x +(ge f )x
g +gx . In this case 1 c1 y =
g+gx , 1 e1 y = ge1 fg+gx 1
and
fg f g
dy = (g+gx)2
dx so that the differential equation (93) takes the form
fg f g

(g2 c21 f 2 )(g2 e21 f 2 )
dx

g+c1 f gc1 f g+e1 f ge1 f
1+ g+c1 f x 1+ gc1 f x 1+ g+e1 f x 1+ ge1 f x
dx
= a .
(1 c x 2 )(1 e2 x 2 )
2
c2 2 a 1 c2 e2
The solutions are y = ax, c21 = , e2 = ae2 ; y = ec
a2 1
2
x , c1 = a2
, e21 = a2
;

y = m 1x ec c e
, c = 1 , e1 =
1+x ec 1 m c+ e
1 c+e
m c e , a =
m 1
2 (c e).
In order to deal with the general case, in which the solutions of y = (x) are
, ( + 1 ), . . . , ( + m1 ),
Abel writes (x) = qp where p and q are polynomials of degree m in x, with

respective dominant coefficients f and g. The equation y = (x) is rewritten
p qy = ( f gy)(x )(x ( + 1 )) . . . (x ( + m1 )). (94)
If f and g are the respective coefficients of x m1 in p and q, we see that

f g y = ( f gy)( + ( + 1 ) + . . . + ( + m1 ))

and y = gf + f
+g where = + ( + 1 ) + . . . + ( + m1 ). It remains to
express rationally in function of x with the help of the addition theorem and to
determine f, f , g, g , e1 , c1 and a in order that (93) be satisfied. For some j , it is

possible that ( j ) = ( + j ) or ( + 2 j ) = . Then j = m2 + m2 with
m+ m even and the distinct values of (
+ j ) are
= x, ( + ) = = x,

+ 2 + 2 = ec1 1
= ec1 1x , + 3 2 + 2

= ec1 (+)1
= ec1 1x . For the
other ( j ) = ( + j ), so it is a root of the equation y = (x), of the form
2x
( + j ) and we have ( + j ) + ( j ) = 1e2 c2 2 j x 2 . Thus
j
3

= + k( + ) + k + + +k + +
2 2 2 2
+( + 1 ) + ( 1 ) + . . . + ( + n ) + ( n )
k k 1 2x j
= (1 k)x + +
ec x 1 e2 c2 2 j x 2
where k, k , k are equal to 0 or 1.

In the first case considered by Abel, k = k = k = 0. Let , , , be the
values of respectively corresponding to y = c11 , c11 , e11 , e11 . One has 1 c1 y =

g c1 f g +c1 f g e1 f
r 1 , 1 + c1 y = r 1 , 1 e1 y = r 1 and

1 + e1 y = g +er 1 f 1
where r = g + g . From the expression of , one
1+A x+A x 2 +...+A x 2n+1
gets 1 1 2 2n+1
= (1e2 c2 2 1 x 2 )(1e2 c2 2 2 x 2 )...(1e2 c2 2 n x 2 ) , which must be annihilated
by = , 1 , . . . , n ( arbitrary). Thus

2n+1
x x x
1 + A1 x + . . . + A2n+1 x = 1 1 1 ...
( + 1 ) ( 1 )

x x
1 1 .
( + n ) ( n )
The differential equation (93) is written
1 dy

(1 c21 y2 )(1 e21 y2 ) = (1 c2 x 2 )(1 e2 x 2 )
a dx

and it shows that when x = 1c , 1e or = 2 , 2 , the left hand side is 0. Thus,

for instance, = 2 , = 2 , = 2 , = 2 and g = c1 f 2 = e1 f 2 ,

f = cg1 2 = eg1 2 . A solution of this system is g = f = 0, gf = 1k ,

c1 = k
( 2 )
, e1 = where k is arbitrary. Then y = 1k and 1 =
k
2 (2)
2 2 2
1 x
(1 cx) 1 ( 2 1 ) 1 x . . . 1 x where
( 2 2) ( 2 n)
130 C. Houzel
= (1 e2 c2 2 1 x 2 )(1 e2 c2 2 2 x 2 ) . . . (1 e2 c2 2 n x 2 ).
We obtain similar expressions for 1+ and for 1 and, as in the Recherches,

(2) 2
t2 t2
1 c21 y2 = (1 c2 x 2 ) , 1 e21 y2 = (1 e2 x 2 )
2 2
where

x2 x2 x2
t = 1 2 1 2 . . . 1 ,
2 1 2 2 2 2 n

2 2 2
x x x
t = 1 1 . . . 1 .
2 2 1 2 2 2 2 2 n

Thus (1 c21 y2 )(1 e21 y2 ) = tt2 (1 c2 x 2 )(1 e2 x 2 ) and Abel shows, as in
2 dy
the Recherches, that ttdx is a constant a, so that the desired result is obtained.
The value of a is computed by comparing the limit values of dy for x infinite
dx
dy tt 1 ()
coming from dx = a 2 and from y = k x + 2x 1e2 c2 2 x 2
. Abel finds a =
(e2 c2 )n 1k 4 1 4 2 . . . 4 n . He gives some other forms for y, as

x2 x2 x2
x 1 2 1
1 2 2
... 1 2 n
y=a
(1 e2 c2 2 1 x 2 )(1 e2 c2 2 2 x 2 ) . . . (1 e2 c2 2 n x2)
1
= (ec)2n b (1 + )(1 ) . . . (n + )(n )
k

where b = 2 1 2 2 . . . 2 n . Doing = 2 and = 2, he obtains
1 b 2
= (1)n e2n c2n1 1 2 . . . n and
c1 k 2 2 2
2
1 b
= (1)n e2n1 c2n 1 2 . . . n .
e1 k 2 2 2
As Abel remarks, the transformation defined by Jacobi corresponds to the case

2
in which 1 = 2n+1 , c = c1 = 1 and the theory explained in the Recherches to the
case in which
m + m
1 =
2n + 1
with m+m even, m, m and 2n+1 having no common factor. In both cases, 2 = 21 ,
3 = 31 , . . . , n = nn . Jacobi independently found these transformations. The
f + f
more general transformation y = g +g is obtained by composing this particular

f + fx
one with a transformation of the type y = g+gx .

A second case considered by Abel is that in which k =0 and k or k is equal

to 1. It is impossible that k = k = 1 for if + 2 + 2 and + 3 2 + 2

+
are roots of y = (x), so is + 3 2 + 2 2 = ( + ) and k is not 0. As

in the first case, let 1 c1 y = 0 for x = 1c . Then 1 c1 y = g cr 1 f 1 ,
(2)
2 c2 f 2 2
g
1 c21 y2 = r2
1
1 and
(2)

1 x x x x
1 = 1 1 1 1
( + ) ( + 1 ) ( 1 )

x x
1 1 (95)
( + n ) ( n )
+ 3+
where = 2 (resp. 2 ) if k = 1 (resp. k = 1) and
= ecx(1 e2 c2 2 1 x 2 )(1 e2 c2 2 2 x 2 ) . . . (1 e2 c2 2 n x 2 ).
Abel takes = 2 in order to compute 1 c21 y2 and he finds

c21 f 2 g 2
1 c21 y2 = t (1 c2 x 2 )(1 e2 x 2 ).
2 r

( 2 )

r dy
Now it results from (93) that 1 e21 y2 = is a rational function
t dx
a c21 f 2 g 2

of x. If we impose that 1 e21 y2 be annihilated by x = 2 , we effectively
find

e21 f 2 g 2 x 2
x 2
1 e21 y2 = 1 1 . . .
2 r 2
2 2
2 1

x2
1
2
2 n
by doing = in the relation (95). Then we find that g = c1 f 2 =

2
( )

e1 f , f = cg1 2 = eg1 . A solution is f = g = 0, gf = c12 =

2 2

, which gives c1 = k 2 , e1 = k
2 1
, a = eck . Other

e1 2 , y = k
f + fx
solutions are obtained by composing with a transformation of the type y = g+gx .
132 C. Houzel

In the simplest case where
n = 0 and c1 = c = 1, = 3
2 + 2 , the formulae are
x 2 e
y = (1 + e) 1+ex 2 , e1 = 1+e and a = 1 + e.
In the third case, k = 1 and one finds that = ( + ) = so that
= 0. Let us return to (94), denoting by 12 f (resp. 12 g ) the coefficient of x m2 in
p (resp. q) and by F the function 2 + 2 ( + 1 ) + . . . + 2 ( + m1 ). We have
f F
f g y = ( f gy)F and y = gf +
+gF and we may proceed as in the preceding
cases with F in place of .
Abel states the general theorem for the first real transformation of arbitrary
order n:
dy adx
= (96)
(1 y2 )(1 e21 y2 ) (1 x )(1 e2 x 2 )
2
(n1) 1 2
where a = k n 2 3 k 3

n ... n , e1 = en 2n 2n . . . n 2 n ,1 =
2n 2n
(n1)
. . . n 12 n and y = k + n + 2

n ... + n .
With Legendres notation x = sin , y = sin and n very large, e1 be-

comes negligible and Abel writes, with an approximation = a d2 2 =

0 1e sin
n1 1 1
arctan tan 1 e2 2 m . For = 2 , = n 2 = a 2 , so that

n a = n and
m=0
passing to the limit for n infinite, Abel finds

d 1
= arctan tan 1 e2 2 x dx.
2 2
1 e sin
0 0
The order of the transformation, that is the degree m of the equation p qy = 0

is the number of distinct values of ( + k1 1 + . . . + k ) and Abel shows that
m = n 1 n 2 . . . n where, for each j, n j is the smallest strictly positive integer such
that ( + n j j ) = ( + m 1 1 + . . . + m j1 j1 ) for some m 1 , . . . , m j1 . Thus,
when m is a prime number, = 1 and m = n 1 . Abel states some theorems:
a) when the order of a transformation is a composite number mn, this trans-
formation may be obtained by the composition of a transformation of order m and
a transformation of order n;
b) the equation y = (x) is algebraically solvable and its roots x are rational
1 1 1
n n
functions of y and some radicals r1 1 , r2 2 , . . . , rn where n 1 , n 2 , . . . , n are prime
numbers,
n 1 n 2 n is the degree of the equation and r1 , r2 , . . . , r have the form
+ t (1 c21 y2 )(1 e21 y2 ) with and t rational in y;
dy dx
c) If the differential equation = a has a so-
(1c2 y2 )(1e2 y2 ) (1c2 x 2 )(1e2 x 2 )

lution algebraic in x and y, a = + where , are rational numbers and
0. There is an infinity of values for the moduli e, c, given by algebraic equations
solvable by radicals, for which > 0.
Recall that in the Recherches Abel doubted that these equations might be alge-
braically solvable. Kronecker (1857) gave a proof that they are solvable by radicals,
as Abel states here.
dy dx
d) If the differential equation = a where b2 =
(1y2 )(1b2 y2 ) (1x 2 )(1c2 x 2 )

1c2 has a solution algebraic in x and y, a = + 1 where , are rational
numbers and 0. In particular, when a is real, it is the square root of a positive
rational number. Thus, when e21 = 1 e2 in (96), a = n. Indeed the formula for k

gives y = 1 when = 2n , thus
1
dy a
= .
(1 y2 )(1 e21 y2 ) 2n
0

(n1)
One may write y = k n 2 because + m

n . . . n n
2 x 2 2 (n1) x 2

= (nm)n , so that y2 = k2 x 2 1e2n2 x 2 . . . 2 2
n
(n1) 2 . Now putting
n 1e x
n
x = p 1 and y = z 1 and letting p and z tend towards infinity, Abel ob-

1 dy
tains 2 = a 2
= a a
2n , whence a = n.
0 (1y2 )(1e1 y2 )
In the part of this memoir published in 1829, Abel gives another study of the
same transformation in the case in which 0 < c, c1 < 1 and e = e1 = 1. Then 2 =
1 1
1 dx

c dx

c dx
is real but 2 = = 2 1 =
0 (1x 2 )(1c2 x 2 ) 0 (1x 2 )(1c2 x 2 ) 1 (1x 2 )(1c2 x 2 )

1 dx

2 1 2 , where 2 = ,b= 1 c2 , is complex. Let us suppose
0 (1x 2 )(1b2 x 2 )
that the differential equation
dy dx
= a (97)
(1 y2 )(1 c21 y2 ) (1 x )(1 c2 x 2 )
2
has a solution f(y, x) = 0 algebraic in x and y and define the function y = 1 by
dy
= d and 1 (0) = 0.
(1 y2 )(1 c21 y2 )
The equation (97) takes the form d = ad, so that = + a where is constant
and y = 1 ( + a). Thus f(1 ( + a), ) = 0 identically in . This implies that
f(1 ( + 2ma + a), ) = f(1 ( + mai + a)) for any integer m. Then there
exists pairs of distinct integers (k, k ) and (, ) such that
1 ( + 2k a + a) = 1 ( + 2ka + a) and
1 ( + ai + a) = 1 ( + ai + a)
134 C. Houzel

or 2k a = 2ka + 2m1 + m 1, ai = ai + 21 + 1

1
1
where m, m , , are integers, 21 = dx
2
, 21 dx
,
0
2 2
(1x )(1c1 x ) 0 (1x 2 )(1b21 x 2 )

b1 = 1 c21 . From these relations we draw a = m 1 + m2 1 1 =
1 2 1 2 1 mm 2
1 and m 1 = 1 , m2 1 = 2 1
. Thus 2 = 4 2
21
2
and 12
= 14 mm . As 2
is a continuous function of c, these equations can
be satisfied for any c, c1 only if m = = 0 or if m = = 0. In the

first case we have a = m 1 = 1 , 1
1
=
m and in the second case

a = m2 1 1 = 2 1
1, 1
1
= 14 m . Abel states that if (97) has
1
a solution algebraic in x and y, then either 1
or
1
1
has a rational ratio to . In the
first case a = 1 and in the second case a = 1 1, with rational. Both ratios

k, k are rational for certain particular values of c, c1 , determined by = kk ,
1

k 1 1

1 = k and in these cases a = + 1 with , rational.
In order to prove that these conditions are sufficient for the existence of an
algebraic solution to the equation (97), Abel observes that = f b 2 b where

f = 1 x 2 is the function introduced in the Recherches. The expansion of f in
simple infinite product then gives
(1 t 2 )(1 t 2r 2 )(1 t 2r 2 )(1 t 2r 4 )(1 t 2r 4 ) . . .
= A
(1 + t 2 )(1 + t 2r 2 )(1 + t 2r 2 )(1 + t 2r 4 )(1 + t 2r 4 ) . . .

= A (+) () (2+) (2) . . . (98)

2x
where A is independent from , t = e , r = e and (x) = 1e
1+e2x
. From this
formula we draw

2 n1
+ + ... +
n n n

= An (1 + ) (1 ) (21 + ) ...
1 1 1 1
1 1
where = = n1 . On the other hand
and 1

1 = A1 (1 + ) (1 ) ...
1 1 1
and, by comparison, we conclude that 1 1 = AAn1 + n + 2

n
dy
. . . + n1

n = y, algebraic function of x = such that =

2 (1y2 )(1c1 y2 )
1 1 dx
d =
.
(1x 2 )(1c2 x 2 )
There are three cases
to consider: a real, a purely imaginary and a complex with

= kk and 1 = kk . In the first case, a = 1 and

1 1
1
= mn where , , m, n
are integers. The equation (97) is satisfied by x = (), y = 1 (1 ). Let c

be a modulus such that = n1 , where and are the periods corresponding
to c . We have
A

n1
( ) = n () + . . . +
A n n
1 1
and since = m 1 ,
A

1 m1
( ) = m 1 (1 )1 1 + 1 1 + 1 .
A1 m m
Finally

1 n1
() + +
An n n

1 1 m1
= m 1 (1 )1 1 + 1 1 + 1 (99)
A1 m m
where = and 1 = 1 . The left hand side is an algebraic function of
(), so an algebraic function of x = and, in the same way, the right hand side
is an algebraic function of y = 1 1 . Thus we have an algebraic integral of (97). One
sees that A = 1c and A1 = 1c1 . As an example, Abel explains the case in which

a = 1 and 1
= 23 ; the equation (99) takes the form c c + 3 + 2

3 =
1 2 x 2
1y 2
c1 1 1 1 1 + 21 or y 2 = cc1c x 1c232 x 2 .

1c1 y2 3

In the second case a = 1 1 and 1
= m
n with , , m, n integers. Let us
z

1 dx
1
put x = 2 so that 2 2 2
= 1 2dz 2 2 where b = 1 c2 .
1z (1x )(1c x ) (1z )(1b z )
dy 1 dz
The equation (97) takes the form =
and we are
(1y2 )(1c21 y2 ) (1z 2 )(1b2 z 2 )
reduced to the preceding case, with the algebraic integral (99) where z = =
1 1
x and replaced by . For instance if a = 1 and = 2 , (99) is
x 2 1 1

1y2
written b = c1 1 (1 )1 1 + 21 or y 2 = c1b x2 .

1c1 y2 x 1
1 1
In the third case a = + 1 where , , , are integers and
1 1
1 = k = k where k, k are rational numbers. The two equations
dz 1 dx
= and
(1 z 2 )(1 c21 z 2 ) (1 x )(1 c2 x 2 )
2
dv 1 dx
=
2
(1 v2 )(1 c1 v2 ) (1 x )(1 c2 x 2 )
2
have algebraic integrals and our equation (97) may be written

136 C. Houzel
dy dz dv
= +
2 2 2 2
2
(1 y )(1 c1 y ) 2
(1 z )(1 c1 z ) (1 v )(1 c21 v2 )
2

z (1v2 )(1c21 v2 )+v (1z 2 )(1c21 z 2 )
which is satisfied by y = 1c21 z 2 v2
, algebraic function of x.
Abel gives a translation of his general theorem in Legendres notations (which were
adopted by Jacobi).
The case where c1 = c corresponds to complex multiplication. Then a = m +
m
2 m m 2
2 1 = 1 and = , 2 = . The multiplicator a is

real if m = = 0 but otherwise we must impose = 12 m = k where k is

positive rational and we have a = + k 1 where , are rational numbers.
Doing = 2 in (98), Abel obtains

4 1 e k 1 e3 k 1 e5 k
c= and
1 + e k 1 + e3 k 1 + e5 k

3
5

4 1e k 1e k 1e k
b= 3 5

1 + e k 1 + e k 1 + e k
Abel continued the explanation of the theory of transformation in a memoir pub-

lished in the third volume of Crelles Journal (1828), Sur le nombre des transforma-
tions differentes quon peut faire subir a une fonction elliptique par la substitution
dune fonction rationnelle dont le degre est un nombre premier donne (uvres, t. I,
p. 456465). He puts
2 2
2 = (1 x 2 )(1 c2 x 2 ), = (1 y2 )(1 c x 2 )
dy
and supposes that the differential equation = a dx
is satisfied by
A0 + A1 x + . . . + A2n+1 x 2n+1
y=
B0 + B1 x + . . . + B2n+1 x 2n+1
where 2n + 1 is a prime number and one of the coefficients A2n+1 , B2n+1

is different from 0. He recalls that, according to the Solution dun probleme
general, when B2n+1 = 0, one has y = vp , c = 2 and a = where
2
2

p = x 1 x2 1 2x(n) , v = (1 c2 2 x 2 ) (1 c2 2 (n)x 2 ),
1 2 1
= cn+ 2 2 + . . . 2 + n , = cn+ 2 ( (2) (n))2 and

f + fy
= m+m2n+1

(m, m integers). Other solutions are given by composing with g+gy
f

where f , f, g, g are constants such that 1 + gg+ f
+ f x 1 + gg f
f x 1 + gg+c
+c f x

gc f 2 2 2
1 + gc f x = (1 x )(1 c x ). Thus, disregarding the signs, one finds 12
values for y and 6 values for c for each choice of :

I II III IV V VI
2 2 2 2
1 1 1+ 1i 1+i
c 2 2 1+ 1 1+i 1i
a
2
(1 + )2 i 2
(1 )2 i 2
(1 + i)2 i 2 (1 i)2 i

p v
v p 1+ v p 1 v p 1+i v pi 1i v pi
y 1 v vp 1 v p 1+ v p 1i v pi 1+i v pi .
p
This comes from the fact that the modulus c2 is a modular function of level
2 in Kleins sense: it is invariant by the group of 2 2 matrices congruent to the
identity modulo 2 operating on the ratio of the periods, and this group is of index 6
in SL(2,Z).
It remains to count the number of leading to different solutions. If and

lead to the same solution of type I, one finds that p = p, v = v and = .
From p = p it results that 2 = 2 () for an integer between 1 and n, thus
= k + k where k, k are integers. For such a value of , p = p, v = v,
m
= , and = and both solutions are effectively equal. Now when = 2n+1 , there

exist integers k, such that k(2n +1)m = 1 and one has k = 2n+1 . When

= m+m
2n+1 with m = 0, there exist integers k , such that k (2n + 1) m = 1

+
and one has k + k = 2n+1 where = k(2n + 1) m. Thus the different
+ +2 +2n
choices for are 2n+1 , 2n+1 , 2n+1 , 2n+1 , . . . , 2n+1

; their number is 2n + 2.
The values of y of types III, IV, V and
VI may bewritten in another way with
the help of theidentities v p = (1 x c)(12k1 x c+cx 2 )(12k2 x c+cx 2 )
(1 2kn x c + cx 2 ),

v p 1 = (1 x c)(1 2k1 x c cx 2 )(1 2k2 x c cx 2 )

(1 2kn x c cx 2 )
()
and similar expressions for v + p, v + p 1, where k = 1c 2 (a) , k =
()

1+c2 (a)
, () = (1 2 )(1 c2 2 ). When 0 < c < 1, Abel explains that

the only transformations for which c is real correspond to = 2n+1 or 2n+1 and
that they are of type I, II, III or IV.
As we saw above, when 0 < c < 1, isreal and = + 1 where is
real. Abel gives an expression of = f b 2 in infinite product

2 1 2q 2 cos 2 + q 4 1 2q 4 cos 2 + q 8
4
= q sin
1 2q cos 2 2 1 2q 3 cos 2 + q 6

c + q

where q = e and computes with the help of this formula. If = 2n+1 , he finds
2
1 + q 2(2n+1) 1 + q 4(2n+1)

4 2n+1
=2 q ... .
1 + q 2n+1 1 + q 3(2n+1)
138 C. Houzel
i+2
The other values of are of the form 2n+1 (0 2n) and give
1
2 1
4 2
2n+1 2n+1

4 1
1 + 1 q 1 + 1 q
= 2 1 q 2n+1 3 . . .

1
2n+1
1
1 + 1 q 1 + q 2n+1 1
2
2
where 1 = cos 2n+1 + 1 sin 2n+1 is a primitive (2n + 1)-th root of 1.
Thus the 2n + 2 values of are obtained by replacing q in the expression
2 1+q4 2
1+q 2m
2 4 q 1+q
1+q 1+q 3 . . . 1+q 2m1 . . . = c by
1 1 1 1
q 2n+1 , q 2n+1 , 1 q 2n+1 , 12 q 2n+1 , . . . , 12n q 2n+1 .
2 1q4 2
The same substitutions in the expression 2 4 q 1q
1q 1q 3 . . . give the 2n + 1
values of . Jacobi independently discovered similar rules for the transformations
(1829).
In a very short paper published in Crelles Journal(vol. 3, 1828; uvres,
t. I, p. 466), Abel states the rule for the transformation of elliptic integrals
of the third kind. Let f(y, x) = 0 be an algebraic integral of the differen-
dy dx 2
. Then A+Bx dx

tial equation 2 2
= a 2 2 2 x2

2 2 2
=
(1y )(1c y ) (1x )(1c x ) 1 2 (1x )(1c x )
n
A +B y2 dy
2
+ k log p where A , B , m and k are functions of A, B, n
1 y 2 (1y2 )(1c 2 y2 )
m
and p is an algebraic function of y and x. The transformed parameter m is determined
by the equation f(m, n) = 0. For n infinite, the integrals are of the second kind and
the rule for the transformation is

2 dx dy
(A + Bx ) = (A + B y2 ) +v
2 2 2
(1 x )(1 c x ) (1 y )(1 c 2 y2 )
2
where v is an algebraic function of x and y.
8 Further Development of the Theory of Elliptic Functions

and Abelian Integrals
In the fourth volume of Crelles Journal (1829; uvres, t. I, p. 467477), Abel

published a Note sur quelques formules elliptiques, devoted to the translation of the
formulae given in the the Recherches into Legendres notation in order to recover
results published by Jacobi. Supposing that c = 1, the problem is to pass from the case
in which e is real to the case in which e2 is negative. Abel puts = f 2 b
where b = 1 2 so that x = is equivalent to = dx

where
1+e 0 (1x 2 )(1c2 x 2 )

c = e 2 and b = 1 c2 . One has
1+e
1 1
dx dx
=b , =b .
2 (1 x 2 )(1 c2 x 2 ) 2 (1 x 2 )(1 b2 x 2 )
0 0

Abel auxiliary functions = 1 2 , = 1 c2 2 to play the
roles
of cosam,
am in Jacobis notation. One has = 2 b , =
bF 2 b . As in the preceding papers, Abel uses the expansions of f, and
F in simple infinite products to obtain expressions for , and (cf. (98)):
1 2 1 2r 2 1 2r 2 1 2r 4 1 2r 4
= A ,
1 + 2 1 + 2r 2 1 + 2r 2 1 + 2r 4 1 + 2r 4
2 1 2r 1 2r 1 2r 3 1 2r 3
= A ,
1 + 2 1 + 2r 2 1 + 2r 2 1 + 2r 4 1 + 2r 4
2 1 + 2r 1 + 2r 1 + 2r 3 1 + 2r 3
= A . ... (100)
1 + 2 1 + 2r 2 1 + 2r 2 1 + 2r 4 1 + 2r 4

2
2

where = e , r = e , 2 = d
, 2 = d
and
0 1c2 sin2 0 1b2 sin2
(1 + r)(1 + r 3 )
A= ,
(1 r)(1 r 3 )
(1 + r 2 )(1 + r 4 )(1 + r 6 )
A = ,
(1 r)(1 r 3 )(1 r 5 )
(1 + r 2 )(1 + r 4 )(1 + r 6 )
A = . (101)
(1 + r)(1 + r 3 )(1 + r 5 )

Doing = 2 + in (100), one obtains 2 = r and
2 i
1 + e2 1

(1 + r)(1 + r 3 )(1 + r 5 ) . . .
2
= f i = = =A = A2 ,
2 e c (1 r)(1 r 3 )(1 r 5 ) . . .
2
1 + r2 1 + r4

i i b
= = = i = 4A i r . . . .
2 e c 1 r 1 r3

Thus A = 1c , A = 2 1
4r
b
c . In a similar way, doing = 2 one obtains = r and
2
2 1+r 4
2
b
= b = 4A r 1+r1+r 1+r 3 . . . = 4A r A so that A = 2 4 r . These values
compared with (101) give

4 1 r 1 r3 1 r5
c= ,
1 + r 1 + r3 1 + r5

4 b 1 + r2 1 + r4 1 + r6
= 28r ,
c 1 r 1 r3 1 r5
4 1 + r2 1 + r4 1 + r6
b= 28r . (102)
1 + r 1 + r3 1 + r5
140 C. Houzel

The limit value of 12
for 0 is 2 and, comparing with (100), this gives

4 (1 r 2 )(1 r 4 )(1 r 6 )
c = and
(1 + r 2 )(1 + r 4 )(1 + r 6 )

(1 + r)(1 r 2 )(1 + r 3 )(1 r 4 )
=
(1 r)(1 + r 2 )(1 r 3 )(1 + r 4 )
= ((1 + r)(1 + r 3 )(1 + r 5 ) )2
(1 + r)(1 + r 2 )(1 + r 3 ) (1 r)(1 r 2 )(1 r 3 ) .
Abel puts P = (1 + r)(1 + r 3 )(1 + r 5 ) and P = (1 + r 2 )(1 + r 4 )(1 + r 6 )

so that
1
PP = (1 + r)(1 + r 2 )(1 + r 3 ) =
(1 r)(1 r 3 )(1 r 5 )

and 4
c= 1
P2 P
, 4b = 2 8 r PP . From these relations, he draws

6

6 24 r b 24 r 1
P= 2 , P = , (103)
b2 c2 3
2 12 c 8 r

12
b
PP = (1 + r)(1 + r 2 )(1 + r 3 )(1 + r 4 ) =
6
2 3
, (1 r)(1 r )(1 r ) =
24
2c r

12

b 3 c
, one of the formulae published by Jacobi.

6 24
2 r

Now putting q = e so that log r log q = 2 , = 2 + 2 1 + x 1
and exchanging b and c, Abel obtains from (100)
1 2q 2 cos 2x + q 4 1 2q 4 cos 2x + q 8

2
x = 4 q sin x ,
c 1 2q cos 2x + q 2 1 2q 3 cos 2x + q 6

1 + 2q 2 cos 2x + q 4 1 + 2q 4 cos 2x + q 8

b
x =2 4
q cos x ,
c 1 2q cos 2x + q 2 1 2q 3 cos 2x + q 6
1 + 2q cos 2x + q 2 1 + 2q 3 cos 2x + q 6

x = b . (104)
1 2q cos 2x + q 2 1 2q 3 cos 2x + q 6
By comparison with Jacobis formula for am, Abel finds
1 + 2q cos 2x + 2q 4 cos 4x + 2q 9 cos 6x + . . .

1 2q cos 2x + 2q 4 cos 4x 2q 9 cos 6x + . . .
(1 + 2q cos 2x + q 2 )(1 + 2q 3 cos 2x + q 6 )(1 + 2q 5 cos 2x + q 10 ) . . .
= .
(1 2q cos 2x + q 2 )(1 2q 3 cos 2x + q 6 )(1 2q 5 cos 2x + q 10 ) . . .
The logarithms of (104) are written


1 1
log x = log 2 log c + log sin x
2 4
q2 q3

q 1 1
+2 cos 2x + cos 4x + cos 6x + ... ,
1+q 2 1 + q2 3 1 + q3

1 1 1
log x = log 2 + log b log c + log cos x
2 2 4
q2 q3

q 1 1
+2 cos 2x + cos 4x + cos 6x + ... ,
1q 2 1 + q2 3 1 q3
q3

1 q 1
log x = log b + 4 cos 2x + cos 6x + . . . .
2 1 q2 3 1 q6

q 1 q3 1 q5
For x = 0, this last formula gives log 1b = 8 1q

2 + 3 1q 6 + 5 1q 10 + . . . and
the first one gives
1 1 q2 1 q3

1 q
log = 2 log 2 + 4 + . . .
c 2 1+q 2 1 + q2 3 1 + q3

r 1 r3 1 r5
which is also equal to 8 1r 2 + 3 1r 6 + 5 1r 10 + . . . according to (102). From
the expansions of 2 , f 2 and F 2 in simple series (Recherches, formulae

(86)), Abel deduces
q2

4 1 q
x = q sin x + sin 3x + sin 5x + . . .
c 1q 1 q3 1 q5
q2

4 1 q
x = q cos x + cos 3x + cos 5x + . . .
c 1+q 1 + q3 1 + q5
2 r r 1x r 3x r 33x r 5x r 55x
x
= + + ... ,
c 1+r 1 + r3 1 + r5
2 r + r 1x r 3x + r 33x r 5x + r 55x
x

x = + + ... .
1r 1 r3 1 r5
Let c be a modulus (between 0 and 1) such that there exists a transformation

from the elliptic functions of modulus c to those of modulus c , and let , , r , q
be associated to c as , , r, q are associated to c. The characterisation stated in
n
Solution dun probleme general is = mn where n, m are integers, or r = r m ,

m
q = q n . For instance, let us take c = 12 , so that = and r = e . Any
admissible value of c is given by

4 1 e 1 e3 1 e5
c =
1 + e 1 + e3 1 + e5
2 4 6
1 + e 1 + e 1 + e
= 2e 8 3 5

1 + e 1 + e 1 + e
142 C. Houzel
where is a rational number and such a c is expressible by radicals. Another

example is that in which b = c or c = b; then = , = . In this

case
n m 4 1e 1e3 1e5
= = m and = n = . Thus c = 1+e 1+e3 1+e5 . . .
3 5

4 1e 1e

1e

and b=
3
5

...
1+e 1+e
1+e
At the end of this paper, Abel deduces the functional equation for a theta-function
from (103). Exchanging c and b and r and q, he obtains (1 +q)(1+q 3 )(1 +q 5 ) =
6 24 q
2 12 and comparing with (103)
bc
1 1
(1 + r)(1 + r 3 )(1 + r 5 ) = 24
24 (1 + q)(1 + q 3 )(1 + q 5 )
r q
whenever r and q are between 0 and 1 and related by log r log q = 2 . He recalls
some other results due to Cauchy (1818) and to Jacobi (1829).
In a second paper of the fourth volume of Crelles Journal (1829), Theoremes
sur les fonctions elliptiques (uvres, t. I, p. 508514), Abel considers the equation
(2n + 1) = R of which the roots are x = ( + m + ) where is the elliptic
2 2i
function of the Recherches, = 2n+1 , = 2n+1 and m, are integers. He proves
that if is a polynomial in these roots which is invariant when is changed into
+ or into + , one has
= p + q f(2n + 1) F(2n + 1)
where p and q are polynomials in (2n + 1), of respective degrees and 2,

being the highest exponent of in . Indeed, by the addition theorem (73),
( + m + ) is a rational function of and f F. Since ( f F)2 =
(1 c2 2 )(1 + e2 2 ), one has
= 1 () + 2 () f F
where 1 () and 2 () are rational. They are respectively given by

1
1 () = ( + ( )) and
2
1
2 () f F = ( ( )). (105)
2
The invariance of by + or + implies that 1 (( + m + )) =
1 (), so that 1 () is a rational symmetric function of the roots of the considered
equation. Thus 1 () = p rational function of (2n + 1) = y. If y = (2n + 1)
is a pole of p, (105) shows that some + m + or some + m + is
a pole of , but then (2n + 1) is also a pole of , which is absurd. On the other
hand, f(2n + 1) = f u, F(2n + 1) = F v where u and v are rational functions
2 () fF
of . It results that f(2n+1)F(2n+1) = () rational function of also equal to
1 ()
2 f(2n+1)F(2n+1) according to (105). Thus () is invariant by + or +
and one proves as above that is it a polynomial q in (2n + 1). Abel computes the
degrees of p and q by considering the behaviour of and ( ) when is
infinite.
When = 1, p is of degree 1 and q = 0, so that = A + B(2n + 1) where
2n
2n
A and B are constants. This is the case for = ( + m + ) where
m=0 =0
is the product of some roots of the equation and one finds that A = 0 when the
number of factors of is odd whereas B = 0 when this number is even.
In the same way, Abel obtains that if is a polynomial in the quantities
f( + m + ) (resp. F( + m + )) such that () = ( + ) = ( + ),
then = p + q(2n + 1) F(2n + 1) (resp. p + q(2n + 1) f(2n + 1)) where
p and q are polynomials in f(2n + 1) (resp. F(2n + 1)) of respective degrees
, 2, beeing the highest exponent of f (resp. F) in .
As an application, Abel deduces a formula established by Jacobi (1828) for the
2 +4n
4n
1
2n+1
division of elliptic integrals: 2n+1 = 2n+1 pm + qm f F where pm
m=0
(resp. qm ) is an odd (resp. even) polynomial in of degree 2n + 1 (resp. 2n 2)
and p2m qm2 ( f)2 (F)2 = (2 am2 )2n+1 where am is a constant.
A third memoir of Abel in the volume 4 of Crelles Journal (1829) is
a small treatise on elliptic functions, titled Precis dune theorie des fonctions el-
liptiques
(uvres, t. I, p. 518617).

Hedx uses the following notations: (x, c) =
(1 x 2 )(1 c2 x 2 ), (x, c) = (x,c) (integral of the first kind), 0 (x, c) =
x 2 dx
(x,c) (integral of the second kind) and

dx
(x, c, a) = (integral of the third kind).
x2
1 a2
(x, c)
The general problem dealt with by Abel is the following: To find all the possible
cases in which one may satisfy an equation of the form
1 (x1 , c1 ) + 2 (x2 , c2 ) + . . . + n (xn , cn )

+1 0 (x1 , c1 ) + 2 0 (x2 , c2 ) + . . . + m

0 (xm , cm )
+1 (x1 , c1 , a1 ) + 2 (x2 , c2 , a2 ) + . . . + (x , c , a )

= u + A1 log v1 + A2 log v2 + . . . + A log v (106)
where 1 , 2 , . . . , n ; 1 , 2 , . . . , m
; 1 , 2 , . . . , ; A1 , A2 , . . . , A are con-
stant quantities, x1 , x2 , . . . , xn ; x1 , x2 , . . . , xm ; x1 , x2 , . . . , x variables related by

algebraic equations and u, v1 , v2 , . . . , v algebraic functions of these variables.

This problem is attacked by purely algebraic means, that is without the use of
elliptic functions and their double periodicity.
rdx
Here are some results announced in the introduction: If (x,c) , where r is an
arbitrary rational function of x, is expressible by algebraic and logarithmic functions
and by elliptic integrals , 1 , 2 , . . . , one may always suppose that
144 C. Houzel

rdx
= p(x, c) + (y) + 1 (y1 ) + 2 (y2 ) + . . .
(x, c)
q1 + q1 (x, c) q2 + q2 (x, c)
+A1 log + A 2 log + ...
q1 q1 (x, c) q2 q2 (x, c)
where all the quantities p, q1 , q2 , . . . , q1 , q2 , . . . , y, y1 , y2 , . . . are rational func-
tions of x. In this statement, (x, c) may be the square root of a polynomial of any
degree.
If any equation of the form (106) takes place and one designates by c any one
of the moduli which figure in it, among the other moduli there is at least one c
dy dx
such that the differential equation (y,c ) = (x,c) may be satisfied by putting for y
a rational function of x, and vice versa.
The second part of the memoir was not written by Abel and we have only the
statement of its principal results in the introduction. Abel supposes that 0 < c < 1
and introduces the elliptic function inverse of (x, c), with its main properties:

1 dx
double periodicity, with the fundamental periods 2, i given by 2 = (x,c) ,
0

1 dx
2 = (x,b) , determination of its zeros and poles, equation ( + )( ) =
0
2 2
1c 2
2 2 , expansion in infinite product. He proves that if the equation ()2n +
2n2
an1 () +. . .+a1 ()2 +a0 = (b0 +b1 ()3 +. . .+bn2 ()2n3 )(, c) is
satisfied by = 1 , 2 , . . . , 2n such that 2 1 , 2 2 , . . . , 2 2n be different between
them, then (1 + 2 + . . . + 2n ) = 0 and (2n ) = (1 + 2 + . . . + 2n1 ) =
a0
1 2 ...2n1 . This statement gives a general theorem for the addition and its proof
is given in the first part.
The roots of the equation of division of the periods are related by remarkable
2 2
linear relations, where = cos 2+1 + 1 sin 2+1 is a primitive (2 + 1)-th root
of 1:

2m 2m + i 2m + 2i
0= + k + 2k
2 + 1 2 + 1 2 + 1

2m + 3i 2m + 2i
+3k + . . . + 2k ,
2 + 1 2 + 1

mi 2 + mi 4 + mi
0= + k + 2k
2 + 1 2 + 1 2 + 1

6 + mi 4 + mi
+3k + . . . + 2k .
2 + 1 2 + 1
Sylow gave a demonstration of these relations in 1864 and he explains how to deduce
them from the theory of transformation in the final notes to Abels Works. He also
reproduces another proof communicated to him by Kronecker in a letter in 1876
(uvres, t. II, p. 314316).
dx
dy
If there is a transformation of (x,c) (with 0 < c < 1 ) into (y,c ) (with c

arbitrary) by putting for y an algebraic function of x, then c is given by one of the

4 (1+q2 )(1+q4 )(1+q6 ) 3
1q1 1q1 1q1
5
following formulae: c = 2 8 q1 (1+q1 )(1+q13 )(1+q15 ) , 4 c = 1+q1 1+q 3 1+q 5
1 1 1 1 1
+ i
where q1 = q = e( ) , , rational numbers.
As in the Solution dun probleme general, Abel obtains a statement concerning
the rational transformation of a real elliptic integral of modulus c into another of
modulus c , with 0 < c, c < 1. The periods , , , must be related by
n
= m where n , m are integers and this condition is sufficient, the multiplicator

being = m . Abel proposes to determine the rational function of x expressing y
by means of its zeros and poles.
When c may 2
be transformedinto its complement b = 1 c (singular mod-
m dy dx
ulus), = n and (y,b) = mn (x,c) . Abel says that c is determined by an
algebraic equation which seems to be solvable by radicals; he is thus doubtful
about this fact, later proved by Kronecker. In the final notes (uvres, t. II, p. 316
318), Sylow gives a proof of this fact by reduction to the solvability of the equation
of division of the periods. Abel gives an expression of 4 c by an infinite product.
He also state that two moduli c and c which may be transformed into one another
are related by an algebraic relation and that, in general, it does not seem possible to
draw the value of c by radicals. But it is possible when c may be transformed into its
complement. According to Abel, all the roots of a modular equation are rationally
expressible by two of them, but this statement is mistaken; they are expressible with
the help of radicals by one of them.
Abel gives an expression of as a quotient of two entire functions =
+ a 3 + a 5 + . . . and f = 1 + b 4 + b 6 + . . . related by the functional
equations
( + )( ) = ( f )2 ( f )2 ( f)2 ,
f( + ) f( ) = ( f f )2 c2 ( )2 .
These functions are similar to the al-functions of Weierstrass, later replaced by .
As we have said, Abel communicated the functional equations to Legendre, saying
that they characterise the functions and f (see 1).
Abel adds that most of these properties are still valid when the modulus c is
a complex number.
The first part of the memoir, the only one written and published, is divided in
five chapters. In the first one (p. 528545), Abel deals with the general properties
of elliptic integrals, beginning by Euler addition theorem proved as a particular case
of Abel theorem: Let fx and x be two arbitrary polynomials in x, one even and
the other odd, with indeterminate coefficients. Let us put ( fx)2 (x)2 (x)2 =
A(x 2 x12 )(x 2 x22 )(x 2 x32 ) . . . (x 2 x2 ) where A does not depend on x, I say
that one will have
a fa + a a
x1 + x2 + x3 + . . . + x = C log , (107)
2a fa a a
a denoting the parameter of the function x, such that x = dx

. The
x2
1 2 x
a
quantity C is the integration constant.
146 C. Houzel
When all the coefficients of fx and x are independent variables, x1 , x2 , . . . , x

are all distinct and one has x = ( fx)2 (x)2 (x)2 = 0 and fx + xx = 0
if x is any one of them. Let denote the differentiation with respect to the variable
coefficients of fx and x. One has
x dx + 2 fx fx 2x x (x)2 = x dx 2x(x fx fx x)
= 0,

2(x fx fxx)
thus x =
x2

and
1 2 x
a
x1 + x2 + x3 + . . . + x

x1 x2 x
(108)
= + + ... + 2
2
x1 2
x2 x x

1 a2 x 1
1 a2 x 2 1 a2

where x = 2(x fx fx x) is a polynomial in

x, of degree

less than that of
aa a fa faa
x. Therefore, the right hand side of (108) is equal to 2a = a ( fa) 2 (a)2 (a)2 =
C 2aa
log fa+aa
faaa . This proof is valid whenever x is the square root of an even
polynomial in x, as was seen in the publication of Abel theorem for hyperelliptic
functions in the third volume of Crelles Journal (see our 5). It is naturally extended
to the case in which the coefficients of fx and x are no more independent and some
of the x j may be equal. Taking a infinite, x is reduced to the integral of the first
kind x and the logarithmic part vanishes, so that
x1 + x2 + . . . + x = C.
An expansion of both members of (107) in ascending powers of a1 gives, by com-
parison of the coefficients of a12 , 0 x1 + 0 x2 + . . . + 0 x = C p where p is
an algebraic function of the variables.
As in the general case of Abel theorem, one may choose x1 , x2 , . . . , xm as
independent variables and determine the coefficients of fx and x in function of
2 2
them. The m quantities xm+1 , xm+2 , . . . , x2 are then the roots of an equation
of degree m and they are algebraic functions of x1 , x2 , . . . , xm . The minimum
value of m is 1. When = 2n is even, one may take
fx = a0 + a1 x 2 + a2 x 4 + . . . + an1 x 2n2 + x 2n ,
x = (b0 + b1 x 2 + b2 x 4 + . . . + bn2 x 2n4 )x,
( fx)2 (x)2 (1 x 2 )(1 c2 x 2 )
= (x 2 x12 )(x 2 x22 ) . . . (x 2 x2n1
2
)(x 2 y2 ), (109)
fx1 + x1 x1 = fx2 + x2 x2 = . . .
= fx2n1 + x2n1 x2n1 = 0. (110)
The linear equations of the last line determine the coefficients a0 , a1 , . . . , an1 ,
b0 , b1 , . . . , bn2 as rational functions of x1 , x2 , . . . , x2n1 , x1 , x2 , . . . , x2n1 .
For x = 0, (109) gives
a0
a02 = x12 x22 . . . x2n1
2
y2 whence y = .
x1 x2 . . . x2n1
fy
Since fx2n + x2n x2n = 0, putting x2n = y one has y = y a rational
function of x1 , x2 , . . . , x1 , x2 , . . . as is y. If, in (109), we put x1 = x2 =
. . . = x2n1 = 0, the right hand side becomes divisible by x 4n2 and we must have
a0 = a1 = . . . = an1 = b0 = b1 = . . . = bn2 = 0. Thus we obtain x 4n =
x 4n2 (x 2 y2 ) and y = 0. Abel shows that if x1 = x2 = . . . = x2n1 = 1 for
x1 = x2 = . . . = x2n1 = 0, then y = 1. Indeed, for x1 , x2 , . . . , x2n1 infinitely
small, the equations (110) reduce to
x 2n + an1 x 2n2 + bn2 x 2n3 + . . . + b0 x + a0 = 0, (111)
with 2n roots x1 , x2 , . . . , x2n1 and z such that a0 = zx1 x2 x2n1 . Thus z = y

and consequently
y2n + an1 y2n2 + . . . + a1 y2 + a0 = (b0 + b1 y2 + . . . + bn2 y2n4 )y,
relation equivalent to y = 1. Since the sum of the roots of (111) is 0, we have
y = x1 + x2 + . . . + x2n1
for x1 , x2 , . . . , x2n1 infinitely small.

It is also possible to take fx odd of degree 2n 1 and x even of degree 2n 2.
1 a0
In this case, one finds that cy = x1 x2 ...x 2n1
.
When = 2n + 1 is odd, let us take
fx = (a0 + a1 x 2 + a2 x 4 + . . . + an1 x 2n2 + x 2n )x and

x = b0 + b1 x 2 + b2 x 4 + . . . + bn1 x 2n2 ,
( fx)2 (x)2 (1 x 2 )(1 c2 x 2 ) = (x 2 x12 )(x 2 x22 ) . . . (x 2 x2n
2
)(x 2 y2 ),
fx1 + x1 x1 = fx2 + x2 x2 = . . . = fx2n + x2n x2n = 0.
As in the preceding case, one obtains y = x1 x2b...x 0

2n
and y = y fy
. For
x1 , x2 , . . . , x2n infinitesimal, x1 , x2 , . . . , x2n being 1, one has y = x1 + x2 +
. . . + x2n and y = 1. One may also suppose fx even and x odd, and then
1 b0
cy = x1 x2 ...x2n .
When n = 1, fx = a0 x + x 3 , x = b0 where a0 and b0 are determined by the
equations
a0 x1 + x13 + b0 x1 = a0 x2 + x23 + b0 x2 = 0
x23 x1 x13 x2 x2 x13 x1 x23

which give a0 = x1 x2 x2 x1 , b0 = x1 x2 x2 x1 . Then
b0 x12 x22 x1 x2 + x2 x1
y= = = .
x1 x2 x1 x2 x2 x1 1 c2 x12 x22
148 C. Houzel
One may verify that (a0 x +x 3 )2 b20 (1x 2 )(1c2 x 2 ) = (x 2 x12 )(x 2 x22 )(x 2 y2 ).
The addition theorem takes the form x1 + x2 = y + C, 0 x0 + 0 x2 =
0 y x1 x2 y + C,
a a0 a + a3 + x1 x2 ya
x1 + x2 = y log + C,
2a a0 a + a3 x1 x2 ya
a0 y+y3 a0 +y2
and y = b0 = x1 x2 .
When x1 , x2 , . . . , x = x ( = 2n 1 or 2n), the coefficients a0 , a1 , . . . ,
b0 , b1 , . . . are determined by the equation fx + x x = 0 and its first 1
derivatives. Let x = xa0 for = 2n 1, xb0 for = 2n be the corresponding value
of y, such that x = C + x. One has (x+m ) = C + x + xm = C + y
x x +x x ye+ey
if y = m 1c2 x 2x 2 m and this equation gives x+m = 1c2 e2 y2 where e is a constant.
m
Letting x tend towards 0, one sees that x+m is equivalent to (m + )x as is y, so
that e = 0, e = 1 and
xm x + x xm
x+m = . (112)
1 c2 xm2 x2
xm x x xm
In the same way, xm = 2 x2
1c2 xm
. For m = 1, this gives x+1 = x1 +

2x x x2+1 x2
1c2 x 2 x
2 and it is easy to deduce by induction that x2+1 , x , x and x2 are
2 x 2
x m
rational functions of x. One has x+m xm = 2 x2
1c2 x
; for m = 1, this gives
m
2 2
1 x x1 2x x
x21 = 2 x2
x 1c2 x . On the other hand, (112) with m = gives x2 = 4 .
1c2 x
1
p r
Let us write x = q , x = 2
q
where p2 and q are polynomials in x
p2 2 p q r
without any common divisor. We have q2 = 4 c2 p4
q
whence p2 = 2 p qr ,

q2 = q4 c2 p4 , for these expressions are relatively prime. On the other hand,
2
p2 q1 2 p2
q
x p21 1
q21 =
2 q2
q 2 2 2 which is an irreducible fraction. Indeed the simultaneous
1 c p p1
equations p q1 q2 p21
2 2 2
= q2 q1 c2 p2 p21 = 0 would give x2 = x1
2
2 x 2
x
2 x x1 +x1 x
and 1 c2 x2 x1 = 0. Since x21 = 2 x2
1c2 x
= 1
x x1 x1 x , we
1
1
should have x x1 = x1 x = 0 and this is absurd for = Thus x2 c.
1 2 2 2 2 2 2 2 2 2
p21 = x ( p q1 q p1 ), q21 = q q1 c p p1 and, from these
relations, Abel recursively deduces that p21 is an odd polynomial in x of degree
(2 1)2 , p2 = p x where p is an odd polynomial of degree (2)2 3, q is
an even polynomial of degree 2 1 (resp. 2 ) when is odd (resp. even). More
precisely,
2 1
p21 = x(2 1 + A2 x 2 + . . . + A(21)2 1 x (21) ),
2 (2)2 4
p2 = xx(2 + B2 x + . . . + B42 4 x ),
(21)2 1
q21 = 1 + A14 x 4 + . . . + A1(21)2 1 x ,
(2)2
q2 = 1 + B41 x 4 + . . . + B4
1
2x .
In his second chapter (p. 545557), Abel considers an exact differential form
y1 dx1 + y2 dx2 + . . . + y dx (113)
where the variables x1 , x2 , . . . , x are related by some algebraic relations in number
less than and y1 , y2 , . . . , y are algebraic functions of them. He supposes that its
primitive is of the form
u + A1 log v1 + A2 log v2 + . . . + A log v
+ 1 t1 + 2 t2 + . . . + n tn (114)
where A1 , A2 , . . . , A , 1 , 2 , . . . , n are constants,
u, v1 , v2 , . . . , v , t1 , t2 , . . . , tn

algebraic functions of x1 , x2 , . . . , x and m x = mdxx is an elliptic integral of

modulus cm (1 m n), with m x = (1 x 2 )(1 c2m x 2 ) and = 1, x 2
or 1x 2 . Let us suppose that x1 , x2 , . . . , xm are independent variables and that
1 2
a
xm+1 , xm+2 , . . . , x are algebraic functions of them. Abel introduces an algebraic
function such that
u, v1 , v2 , . . . , v , t1 , t2 , . . . , tn , 1 (t1 ), 2 (t2 ), . . . , n (tn ) (115)
are rationally expressible in , x1 , x2 , . . . , x , y1 , y2 , . . . , y . He says that a con-
venient linear combination of the functions (115) has this property. In other words,
Abel uses what is now called a Galois resolvant, which is most remarkable. Let
V = 0 the minimal algebraic equation satisfied by , with coefficients rational with
respect to x1 , x2 , . . . , x , y1 , y2 , . . . , y , and let be its degree. Writing that (113)
is the differential of (114), one obtains a relation
p1 dx1 + p2 dx2 + . . . + pm dxm = 0
where p1 , p2 , . . . , pm are rational functions of , x1 , x2 , . . . , x , y1 , y2 , . . . , y and
this implies that p1 = p2 = . . . = pm = 0. These last relations are still verified
when is replaced by any one of the roots 1 , 2 , . . . , of V = 0 because it is an
irreducible equation. It results that
(y1 dx1 + y2 dx2 + . . . + y dx ) = U + A1 log V1 + . . . + A log V
+ 1 (1 t1 + 1 t1 + . . . + 1 t1() ) + . . .
+ n (n tn + n tn + . . . + n tn ())
where U = u + u + . . . + u () is the sum of the values taken by u when is succes-
()
sively replaced by 1 , 2 , . . . , , log Vm = log vm +log vm +. . .+log vm is the anal-
()
ogous sum associated to log vm and tm , tm , . . . , tm are the values taken by tm . Now,
150 C. Houzel
by the addition theorem for elliptic integrals, m tm +m tm +. . .+m tm () = m Tm +

p+ B1 log q1 + B2 log q2 +. . .+ B log q where Tm , m Tm , p, q1 , q2 , . . . , q are ra-
tional functions of tm , tm , . . . , tm() , m (tm ), m (tm ), . . . , m (tm() ) and consequently
of 1 , 2 , . . . , , x1 , x2 , . . . , x , y1 , y2 , . . . , y . But since they are symmetric with
respect to 1 , 2 , . . . , , they are rational functions of x1 , x2 , . . . , x , y1 , y2 , . . . , y
as are U and the Vm . We finally obtain a relation of the form
(y1 dx 1 + y2 dx 2 + . . . + y dx
= r + A log + A log + . . . + A(k) log (k)
+ 1 1 1 + 2 2 2 + . . . + n n n
where is an integer, 1 , 2 , . . . , n are the same as in (114), A , A , . . . are

constants and 1 , 1 (1 ), 2 , 2 (2 ), . . . , n , n (n ), r, , , . . . , (k) are ratio-
nal functions of x1 , x2 , . . . , x , y1 , y2 , . . . , y .
A particular case concerns the differential forms (113) of which the primitive is
of the form u + A1 log v1 + A2 log v2 +. . .+ A log v where u, v1 , v2 , . . . , v are al-
gebraic functions of x1 , x2 , . . . , x . Then one may suppose that u, v1 , v2 , . . . , v are
rational functions of x1 , x2 , . . . , x , y1 , y2 , . . . , y . In a footnote, Abel announces
a general theory, based on this result, for the reduction of integrals of algebraic
differential forms by algebraic and logarithmic functions.
Applied to elliptic integrals, the preceding theorem takes the form: if

1r1 2r2 r
dx1 + dx2 + . . . + dx
1 x1 2 x2 x
= u + A1 log v1 + A2 log v2 + . . . + A log v (116)
where r1 , r2 , . . . , r are rational functions and u, v1 , v2 , . . . , v are algebraic func-

tions of x1 , x2 , . . . , x , one may suppose that u, v1 , v2 , . . . , v are rational functions
of x1 , x2 , . . . , x , 1 x1 , 2 x2 , . . . , x . From (116), we may also conclude that
there exists an integer such that
1 1 x1 + 2 2 x2 + . . . + m m xm + m+1 m+1 1 + . . . + m
= r + A log + A log + . . . + A(k) log (k)
where j x = r1j x dx and 1 , m+1 1 , 2 , m+2 2 , . . . , m , m , r, , ,

. . . , (k) are rational functions of x1 , x2 , . . . , xm , 1 x1 , 2 x2 , . . . , m xm . When

only one elliptic integral m x is isolated, this gives
m m x = 1 1 1 2 2 2 . . . m1 m1 m1 m+1 m+1 m+1 . . .

+ r + A log + A log + . . . + A(k) log (k) (117)
where 1 , m+1 1 , 2 , m+2 2 , . . . , r, , , . . . are rational functions of x and

m x, that is of the form p + qm x with p, q rational in x.
When x1 = x2 = . . . = x = x and c1 = c2 = . . . = c = c, one obtains the
following theorem: if there is a relation
x + 0 0 x + 1 1 x + 2 2 x + . . . + x
= u + A1 log v1 + A2 log v2 + . . . + A log v
where u, v1 , v2 , . . . , v are algebraic functions of x, one may suppose that they are
of the form p + qx with p, q rational in x.
Differentiating (117) we obtain a relation of the form P + Qm x = 0 which
implies P = Q = 0 and therefore P Qm x = 0. When the sign of m x is
changed into the opposite, the j take new values j and we have m m x =
+ v where v designates the algebraic and logarithmic part. It results that

2m m x = ( ) + v v where, by the addition theorem,
= y v

if y = , v denoting an algebraic and logarithmic function. Now =
1c2 2 2
p + qm x and = r + m x where p, q, r, are rational functions of x and it
results that = pqm x, = r m x and that y = tm x where t is a rational
function of x. Then it is easy to see that y is a rational function of x. One may replace
ye+ey
y by z = 1c 2 e2 y2 where e is a constant because y and z differ by an algebraic
y
and logarithmic function. For e = 1, z = is a rational function of x and
1c2 y2
c2 1
z = 1c2 y2
y has a rational ratio to m x. We have 2m m x = z + V where
V is an algebraic and logarithmic function. Then V = u + A1 log v1 + A2 log v2 +. . .
where u, v1 , v2 , . . . are of the form p + qm x with p and q rational in x.
Taking m = , we obtain 2 x = 1 1 z 1 +2 2 z 2 +. . .+1 1 z 1
+ V and we may eliminate x between this relation and (116), getting
1 (21 x1 1 z 1 ) + . . . + 1 (21 x1 1 z 1 ) = V .
Since is an integer, there exist x1 , x2 , . . . , x1

such that 2 j x j j z j =
j x j + V j (1 j 1) and we have 1 1 x1 + 2 2 x2 + . . . + 1 1 x1

=

u + A1 log v1 + A2 log v2 + . . . + A log v of the same form as (116) with one
elliptic integral less. We may iterate until we arrive at a relation with only algebraic
and logarithmic functions.
The general problem (106) has been reduced to the following one: To satisfy in
the most general manner the equation
x = 1 1 y1 + 2 2 y2 + . . . + n n yn
+u + A1 log v1 + A2 log v2 + . . . + A log v (118)
where , 1 , 2 , . . . , n designate elliptic integrals of the three kinds, supposing

that y1 , y2 , . . . , yn are rational functions of x and that 1 y1 , 2 y2 , . . . , n yn are
of the form px where p is rational in x and x designate
the radical which
appears in the function x. If m ym = pm x and m x = mxdx mx
where m x is
rational, we have m ym = mpmym dydxm x dx
where mpmym dydxm is a rational function of x.

Thus m ym = r + Ax + A0 0 x + A (x, a ) + A (x, a ) + . . . where r is an

algebraic and logarithmic expression. Equation (118) finally takes the form
152 C. Houzel
x + 0 0 x + 1 (x, a1 ) + 2 (x, a2 ) + . . . + (x, a )

= u + A1 log v1 + A2 log v2 + . . . (119)
The problem (118) is thus reduced to the three following ones:

A) To find all the possible cases in which
2
(1 y2 )(1 c y2 ) = p2 (1 x 2 )(1 c2 x 2 )
with y and p rational functions of x (c, c are constants).

B) To reduce (y, c ), 0 (y, c ) and (y, c , a), where y and c are as in A), to
the form
r + Ax + A0 0 x + A (x, a ) + A (x, a ) + . . .
C) To find the necessary and sufficient conditions for (119) to be satisfied.

The third chapter (p. 557565) is devoted to the solution of problem C), where
one may suppose that u, v1 , v2 , . . . , v are of the form p + qx, p and q rational
in x. Abel takes the problems dealt with in the second chapter of his unpublished
memoir Theorie des transcendantes elliptiques (see our 4) in a more general setting.
Equation (119) is rewritten

x = u + A log v,
where dxx = x + 0 0 x + 1 1 + 2 2 + . . . + n n and m =

; it is supposed that it is impossible to find any similar relation
2
1 x2 x
m
not containing all the m and that all the m are different from 1 and
1c . Changing thesign of x, we obtain x = u + A log v and then

2x = u u + A log vv . Changing the sign of x without changing that of

x, we obtain 2x = u u + A log vv and

1 1 vv
x = (u u u + u ) + A log .
4 4 vv
If v = p + qx + ( p + q x)x where p, q, p , q are even functions, v = ( p + qx)
( p + q x)x, v = ( p qx) + ( p q x)x and v = p qx( p q x)x.

Thus vvv v = fx+xx
fxxx where fx and x are polynomials, one even and the other one
odd. The algebraic part 14 (u u u + u ) is of the form rx where r is an odd
rational function of x and we may rewrite our equation in the form
fx + x x
x = rx + A log (120)
fx x x
with A in place of 14 A. We may suppose that there is no linear relation with integer
coefficients between the Am , otherwise it would be possible to reduce the number
of the terms in the sum.
fx+xx vdx
Differentiating one term = log fxxx , we obtain d = xx where
x = ( fx)2 (x)2 (x)2 and vx = 2 f xx fx x,
so that v is an even polynomial. If the roots of x are a1 , a2 , . . . , a , the

v 1 2
decomposition of x in simple elements is of the form k + a12 x 2
+ a22 x 2
+ ... +

fa
2 x 2
a
where k is a constant and j = 2m j a j a jj = 2m j a j a j where m j is the
multiplicity of a j as a root of x. Thus the differentiation of (120) gives, after
multiplication by x:
12 1 2 2 2 n
+ 0 x 2 + + 2 2 2 + ... + 2 n 2
a12x 2 a2 x an x
dr
= (x)2 r((1 + c2 )x 2c2 x 3 )
dx
2m 1 a1 a1 2m 2 a2 a2
+ A1 k1 2 2 ...
a1 x 2 a2 x 2
2m 1 a1 a1 2m 2 a2 a2

+ A2 k2 ... + ... .
a1 2 x 2 a2 2 x 2
From this relation, Abel deduces that r = 0 and that only one of the coefficients Am
may be different from 0. He takes A1 = 1, A2 = A3 = . . . = A = 0 and finds
= k1 , 0 = 0, 1 = a1 , 2 = a2 , . . . , 1 = 2m a1 a
1
1
, 2 = 2m 2aa
2
2
, . . . . The
most general relation between elliptic integrals with the same modulus is thus of the
form
2m 1 1 2m n n
x 1 . . . n
1 n
fx + x x
= log +C (121)
fx x x
where the parameters 1 , 2 , . . . , n are related by the equation
( fx)2 (x)2 (1x 2 )(1c2 x 2 ) = (x 2 12 )m 1 (x 2 22 )m 2 . . . (x 2 n2 )m n . (122)
Abel remarks that this implies
m 1 1 + m 2 2 + . . . + m n n = C and
a fa + a a
m 1 1 + m 2 2 + . . . + m n n = C log
2a fa a a
d

if = 2 .
1 2
a
fx+xx
When n = 1, 1 = and m 1 = m we have = 2m x 2m log fxxx
if the parameter verifies
154 C. Houzel
( fx)2 (x)2 (1 x 2 )(1 c2 x 2 ) = (x 2 2 )m . (123)

For m = 2, fx = ax is of degree 1, x = 1c 1 is constant and (x 2 2 )2 =
2
x 4 1+c
c2
a2 x 2 + c12 and this gives a = 1 1c , = 1c . For m = 3 and fx
odd, fx = x 3 + ax, x = b and
(x 3 + ax)2 b2 (1 x 2 )(1 c2 x 2 ) = (x 2 2 )3 ,
equation which leads to 3 = b, 3 + a + b = 0, 2a c2 b2 = 32 , a2 +

(1 + c2 )b2 = 34 . This gives a = 12 (c2 6 32 ) and determined by the equation
= 12 (1 c2 4 ).
Generally, as Abel states, must be a zero or a pole of the function xm defined
dxm dx
in the first chapter and such that x m
= m x (xm = 0 for x = 0). Indeed we have
p2 q 2 (x)2 = (x 2 2 )m (x 2 m )
where m is the value of xm for x = (chapter I) and, multiplying by (123), we

obtain
( p fx qx(x)2 )2 ( px q fx)2 (x)2 = p2 ( fx)2 q 2 (x)2 (x)4

= (x 2 2 )2m (x 2 m ).
It results that p fx + qx(x)2 or p fx qx(x)2 is divisible by (x 2 2 )m and we

have a relation r 2 2 (x)2 = x 2 m 2
where r, are polynomials, one even and
the other odd. But this relation implies that = 0 and that m = 0 or 10 . Conversely,
it is easy to see that such an satisfies an equation (123). Abel remarks that, in these
cases, the coefficient in (121) is always different from 0, so that there is no elliptic
differential of the third kind integrable by algebraic and logarithmic functions.
When n = 3 and m 1 = m 2 = m 3 = 1, (121) takes the form
1 2 1 fx + x x
1 + 2 = + x log
1 2 2 fx x x
where fx = x 3 + ax, x = b and (x 3 + ax)2 b2 (1 x 2 )(1 c2 x 2 ) = (x 2 2 )

2 +2 1
(x 2 12 )(x 2 22 ). This gives = 1
1c2 2 2
, b = 1 2 ,
1 2
1 2 2 2 2 2 + a
a= (c 1 2 2 12 22 ), = , = c2 1 2
2 1 2
(cf. chapter I). In particular, for 2 infinite, = c11 and + c

1
= x +
1 x+x
2 log xx . Other relations between two elliptic integrals of the third kind are
obtained by (121) with n = 2.
In the fourth chapter (p. 565606), Abel solves the problem A) of the second
chapter, that is to satisfy the equation (1 y2 )(1 c 2 y2 ) = r 2 (1 x 2 )(1 c2 x 2 ),
y and r being rational functions of x. Since 1 y2 and 1 c 2 y2 have no common
factor (c = 1), this equation implies 1 y2 = r12 , 1 c 2 y2 = r22 where r1 , r2 are

rational functions of x, r1r2 = r and = (1 x 2 )(1 c2 x 2 ). Differentiating, we
obtain 2ydy = r1 (r1 d + 2dr1 ), 2c 2 ydy = r2 (r2 d + 2 dr 2 ) which show that
the numerator of dy dy
dx is divisible by r1 and r2 , and so by their product r: dx = rv where
v is a rational function without any pole among the zeros of r. Let y = qp , irreducible
fraction where p, q are polynomials of respective degrees m, n. One has r = q2
where is a polynomial and v = q 2 dy
dx =
qd p pdq
dx , whence v is a polynomial. If
m > n, the equation
2
(q 2 p2 )(q 2 c p2 ) = 2 (1 x 2 )(1 c2 x 2 )
shows that 4m = 2 + 4 where is the degree of . If is the degree of v, we

then see that + = m + n 1 and = m + n 1 < 2m 1 = 1.
Thus = 0 and v is constant. In the same way, if n > m, we have 4n = 2 + 4,
< 2n 1 = 1 and = 0. In the remaining case, in which m = n, it is for
instance possible that q p = be of degree m k < m. Then 4m k = 2 + 4
and + = 2m k 1 for v = pdd dx
p
and = 2m k 1 = 1 12 k is
again 0. In any case v is a constant and
dy dx
= . (124)
2 2 (1 x )(1 c2 x 2 )
2
(1 y )(1 c y2 )
The second result announced in the introduction is thus demonstrated.
It remains to determine the rational function y and the transformed modulus c .
Abel begins by considering the case in which y = +x + x and he explains the
6 cases already met in Sur le nombre de transformations differentes . . . (our 7).
A +A x+A x 2 +...+A x
He then considers the case in which y = x = B0 +B1 x+B2 x 2 +...+B x (irreducible
0 1 2
fraction, one of the coefficients A , B different from 0). The treatment uses only the
addition theorem for elliptic integrals of chapter I and not the elliptic function and
its double periodicity as in the preceding memoirs; but the lines are similar. If x, x
are two roots of the equation y = x, one has x dx
= 1 dy y = x
dx
and consequently
xe+ex
x = 1c2 e2 x 2 = x where e is a constant. Thus (x) = x and we see that the
equation y = x has the roots x, x, 2 x, . . . , n x, . . . where it is easy to see that
xen + en x
n x = ,
1 c2 e2n x 2
den de
en being the rational function of e defined by en
= n e and en = 0 for e = 0 (see
chapter I). Since the equation has only roots, there exists an n such that n x = x
en
that is en = 0 and en = 1. These equations are equivalent to 1c 2 e2 = 1, which is
n
of degree n 2 in e. The number n must be minimal and we must eliminate the roots e
which would lead to e = 0, = 1 for a < n. If, for instance, n is a prime
number, the root e = 0 is to be eliminated and it remains n 2 1 solutions e.
Let us suppose that two rational functions z = qp , z = qp where p, q, p , q
are polynomials of degree and the two fractions are irreducible. If the equations
156 C. Houzel
pqy
y = (x) and y = (x) have the same roots x, x , x , . . . , x (1) we have p q y =
aby
a b y where a, b, a , b are the respective coefficients of z in p, q, p , q and z has
any value. We draw y = +y
+ y ; if moreover y and y correspond to the same
1
modulus c , we have y = cy .
When n = , the roots of y = x are x, x, . . . , n1 x and
p qy = (a by)(z x)(z x) (z n1 x). (125)
We can draw y from this equation, giving to z a particular value. If n is odd,

+axx 2 x 2 x
noted 2 + 1, putting z = 0, we obtain y = ab +bxx
2 x 2 x where a , b are the
respective constant terms of p, q. Since enm = em and enm = em , we see
x 2 e2
that nm x = xe m em x
1c2 e2m x 2
and m x nm x = 1c2 e2mx 2 . It results that the value found
m
for y is rational in x. Moreover, it is invariant by the substitution x x because
2+1 x = x, and it results that (125) is verified for any value of z. For x = 1 or
1c , x = 0 and m x = 2+1m x, so that
p q = (a b)(1 z)2 , p q = (a b)(1 + z)2 ,

p q = (a b)(1 cz)2 , p q = (a b)(1 + cz)2
where , , , are the values of y corresponding to x = 1, 1, 1c , 1c and

, , , are polynomials of degree in z. Now we want that
2
(q 2 p2 )(q 2 c p2 ) = r 2 (1 z 2 )(1 c2 z 2 )
and this implies that {, , , } = 1, 1, c1 , c1 ; conversely, this condition will

+ax
be sufficient. Let us take = 1, = 1, = c1 , = c1 . Since y = ab +bx where
x(x 2 e2 )(x 2 e22 ) (x 2 e2 )

x = x x 2 x 2 x =
(1 c2 e2 x 2 )(1 c2 e22 x 2 ) (1 c2 e2 x 2 )

a +a(1) a a(1) a +a 1c a a 1c
is an odd function, we have = = b +b(1) ,
= =b b(1) ,
,
b +b 1c b b 1c
b
or a b (a b)(1) = 0, a a cb 1c = 0. These equations

c
are compatible only if a or b is 0 (c = 1). Let us suppose that a = 0 =
(1) a x 1e2 1e2
2 1e2
b; we have c = , y = b x = (1) where (1) = 1c2 e2 1c2 e22
1c2e2 ,
1c
1 2 2
1 1c2 e2 1c e2 1c2 e2 1
c = c2+1 1e2 1e22
1e2
= c2+1 (1)
. Then c = c2+1 ((1))2 . In order
dy
to determine the multiplicator , Abel uses the value of dx = xy for x = 0, which
+ 1
1
is e2 e22 e2 (1) ; thus = e2 e22 e2 cc2 . He has reconstituted the formulae for
the transformations of odd order 2 + 1: if e is a root of the equation e2+1 = 0
which does not satisfy any other equation e2m+1 = 0 where 2m + 1 is a divisor of
2 + 1, let us put
1
c+ 2 x(e2 x 2 )(e22 x 2 ) (e2 x 2 )
y= ,
c (1 c2 e2 x 2 )(1 c2 e22 x 2 ) (1 c2 e2 x 2 )
2
2+1
(1 e2 )(1 e22 ) (1 e2 )
c =c ,
(1 c2 e2 )(1 c2 e22 ) (1 c2 e2 )
1
c+ 2
= e2 e22 . . . e2 . (126)
c
dy dx
Then we have = . Five other systems (y, c , )
(1y2 )(1c2 y2 ) (1x 2 )(1c2 x 2 )
corresponding to the same value of e are obtained by composing with a transforma-
tion of order 1.
For instance, when = 1, 2 + 1 = 3 is prime and we may take for e any
root different from 0 of the equation e3 = 0, that is 0 = 3 4(1 + c2 )e2 + 6c2 e4
2
1e2 c 2 c c x(e2 x 2 )
c4 e8 of degree 4 in e2 and the c = c3 1c 2 e2 , = c c e , y = c 1c2 e2 x 2 .
Eliminating e, we obtain the modular equation in the form

(c c)2 = 4 cc (1 cc )2 .
+ 1
The roots of the equation 0 = cc2 z(z e2 )(z e22 ) (z e2 ) + y(1
c e z )(1 c2 e22 z 2 ) (1 c2 e2 z 2 ) are x, x, . . . , 2 x, thus x + x + . . . +
2 2 2
(1)+1 c2 e2 e2 e2 2em x
2 x = + 1 1
2
y. Since m x + 2+1m x = 1c 2 e2 x 2 , this gives y =
c 2 c 2 m

2e x +1
2ex 2e2 x c (1)
x + 1c2 e2 x 2 + 1c2 e2 x 2 + . . . + 1c2 e2 x 2 c c e2 e2 e .

2 2
xe +e x
xe e x
If n is even, noted 2, we have x = 1c2 e2x 2 = 1c2 e2x 2 , which imposes

e = 0 or 10 . In the last case, x = cx1 and +m x = c1m x . Thus the roots of
y = x are x, cx1 , x, . . . , 1 x, +1 x, . . . , 21 x and we have

1
p qy = (a by)(z x) z (z x)(z 21 x)
cx
(z 1 x)(z +1 x). (127)
We deduce from this equation that

1 2e x 2e2 x 2e1 x
a b y = (by a) x + + + . . . +
cx 1 c2 e2 x 2 1 c2 e22 x 2 1 c2 e21 x 2
where a and b are the coefficients of z 21 in p and q. It results for y a rational
expression in x, invariant by x x. Choosing a = b = 0, we obtain
a 1
y= 2e1 x
b x 1
+ 2ex
+ ... +
cx 1c2 e2 x 2 1c2 e21 x 2
x(1 c2 e2 x 2 )(1 c2 e22 x 2 ) (1 c2 e21 x 2 )

=A = Ax.
1 + a1 x 2 + a2 x 2 + . . . + a x 2
158 C. Houzel
1
If, for instance, y = 1 when x = 1, we have A = (1) and, from (125), q
2
p = (1 z)(1 cz) where is a polynomial in z. Since q is even and p odd,
q + p = (1 + z)(1 cz)2 and
q 2 p2 = (1 z 2 )(1 c2 z 2 )( )2 .
It results that q 2 c 2 p2 must be a square and c = 1 , where is the value of y

corresponding to x = 1c , satisfies to this condition. Indeed +m x = cx1 =

1c = x for x = 1c , so that p q is a square and the same may be said of
p+q. Thus p2 2 q 2 = t 2 where t is a polynomial in z and (q 2 p2 )(q 2 c 2 p2 ) =
(1 z 2 )(1 c2 z 2 )r 2 for c = 1 . Sylow observes that is never 0 nor , but it is
equal to 1 when is even and this value does not work for c . He explains how to
find a correct value in this case (uvres, t. II, p. 520521). Then dy y = x
dx
where
is the value of dy 1
dx for x = 0, that is = A = (1) . Abel gives an expression of the
denominator q of x as a product b(z )(z ) (z 21 ) where is a pole
of y. It is easy to see that = 1c is such a pole. Thus, if e is a pole of e such
the equations em = 0 and em = 1 cannot be satisfied for any divisor m of 2, the
formulae
1 1 2ex 2e2 x 2e1 x
=x + + 2
+ ... + ,
cy cx 2 2
1c e x 2 2
1 c e2 x 2 1 c2 e21 x 2

1 2e 2e2 2e1
= c 1 + + + ... +
c 1 c2 e2 1 c2 e22 1 c2 e21
dy dx
lead to = . For instance, when = 1, = 1 c,
(1y2 )(1c 2y2 ) (1x 2 )(1c2 x 2 )

x 2 c
y = (1 c) 1cx 2 and c
= 1c .
Another possible value for e is a root of e = 0 such that e = 1 (for
e = 1 would lead to x = x). Here x = x, +m x = m x and equation
(127) is replaced by
p qy = (a by)(z 2 x 2 )(z 2 (x)2 ) (z 2 ( 1 x)2 )
which gives a b y = (a by)(xx 1 x)2 for z = 0, a and b denoting

the constant terms of p and q. Thus y is a rational function of degree 2 of x and
it remains to determine a, b, a , b and c , . For instance, when = 1, Abel finds
2
y = 1+cx
1cx 2 , c
= 1c
1+c , = (1 + c) 1 and he also gives the 5 other possible values

for c .
When the equation y = x has other roots than x, x, . . . , n1 x, Abel shows
that the degree of this equation is a multiple mn of n and that its roots may be
distributed in m cycles x ( j) , x ( j) , . . . , n1 x ( j) , 0 j m1. The proof is identical
with that used for the second theorem of the Memoire sur une classe particuliere
deqations . . . published in the same volume of Crelles Journal. According to the
precedind results, there exists a rational function y1 = 1 x such that the roots of the
equation y1 = 1 x are x, x, . . . , n1 x and that, for convenient c1 , 1
dy1 dx
= 1 . (128)
(1 y12 )(1 c21 y12 ) (1 x )(1 c2 x 2 )
2
p
Let 1 z = q , so that p q y = (a b y)(z x)(z x) (z n1 x). If
pqy
( j)
y j+1 = 1 x (0 j m 1), we see that aby = ap b
q y1 p q y2 p q ym
y a b y a b y . Now
1 2 m
let be a zero and a pole of z and let 1 , 2 , . . . , m , 1 , 2 , . . . , m be the
corresponding values of y1 , y2 , . . . , ym ; from the preceding relation we deduce that
p = A ( p 1 q )( p 2 q ) ( p m q ) and
q = A ( p 1 q )( p 2 q ) ( p m q )
1 1 )(y1 2 )(y1 m )
where A and A are constants, and this gives y = A (y
(y1 1 )(y1 2 )(y1 m ) , rational
A
function of degree m of y1 where A = A . The combination of (124) and (128) gives
the equation
dy dy1
=
2 2 2
(1 y )(1 c y ) 1 (1 y2 )(1 c2 y2 ) 1 1 1
and we see that the transformation of order = mn is obtained by composing

a transformation 1 of degree n and a transformation of order m. This result permits
to reduce the theory of transformations to the case in which the order is a prime
number.
(x)(x )(x(1) )
In the general case, by the above reasoning y = A (x)(x )(x (1) ) where
, , . . . , (1) are the zeros and , , . . . , (1) the poles of x. Abel considers
in particular the cases in which b or a is 0. When b = 0, the equation
p qy = a(z x)(z x ) (z x (1) ) (129)
implies that a b y = a(x + x + . . . + x (1) ) where a and b are the respective

coefficients of z 1 in p and q. If xem +em x
1c2 e2m x 2
= xe m em x
1c2 e2m x 2
for all m, = 2n + 1 is

odd, a = 0 and

2e1 2en
y = Ax 1 + + . . . + .
1 c2 e21 x 2 1 c2 e2n x 2
Therefore q = (1 c2 e21 x 2 ) (1 c2 e2n x 2 ) and p is obtained by making x = 0 in

(129):
p = az(z 2 e21 ) (z 2 e2n ) and

x(e21 x 2 )(e22 x 2 ) (e2n x 2 )
y=a .
(1 c2 e21 x 2 )(1 c2 e22 x 2 ) (1 c2 e2n x 2 )
160 C. Houzel
On the contrary, if xe+ex

1c2 e2 x 2
= xeex
1c2 e2 x 2
, e = 0 or 10 . When e = 1
0,x = cx1 ,

1 2xe1 2xe
= 2n is even, a = 0 and y = A x cx + 1c2 e21 x 2
+ . . . + 1c2 e2n1x 2 =
n1
a(112 x 2 )(122 x 2 )(1n2 x 2 )
x(1c2 e21 x 2 )(1c2 e22 x 2 )(1c2 e2n1 x 2 )
. When e = 0, x = x and one finds that p and q
have the same degree, contrary to the hypothesis.
When a = 0, p qy = by(z x)(z x ) (z x (1) ) and it results that
(1 c2 e21 x 2 )(1 c2 e22 x 2 ) (1 c2 e2n x 2 )

y=a or
x(e21 x 2 )(e22 x 2 ) (e2n x 2 )
x(1 c2 e21 x 2 )(1 c2 e22 x 2 ) (1 c2 e2n1 x 2 )
a
(1 12 x 2 )(1 22 x 2 ) (1 n2 x 2 )
according to the parity of .
In particular
x(e21 x 2 )(e22 x 2 ) (e2n x 2 )

x2+1 = a
(1 c2 e21 x 2 )(1 c2 e22 x 2 ) (1 c2 e2n x 2 )

2e1 x 2e2 x 2en x
= A x+ + + ... +
1 c2 e21 x 2 1 c2 e22 x 2 1 c2 e2n x 2
where 2n = (2 + 1)2 1. Doing x = 10 and 0, one finds Ac2n e21 e22 e2n = a,
22 +2
1
A = 2+1 and ae21 e22 e2n = 2+1. Thus e21 e22 e2n = 2+1 n
cn and a = c = c .
xe1 e1 x xe2 e2 x xen en x
The roots of the equation x2+1 = y are x, 1c2 e21 x 2
, 1c2 e2 x 2 , ... , 1c2 e2n x 2
.
2

Let x = xe+ex
1c2 e2 x 2
and 1 x = xe +e x
1c2 e2 x 2
be two of these roots such that neither e

nor e is a root of x2m+1 = 0 for a divisor 2m + 1 of 2 + 1 and such that 1 x
2
is different from x, x, . . . , 2 x. Then x, x, . . . , 2 x, 1 x, . . . , 1 x are 4 + 1
distinct roots of x2+1 = x = y. Thus, for any m and k, ( x) = (1k x) and m
it results that (1k m x) = ( 2m x) = x2+1 , so that 1k m x is also a root. Now it

is easy to prove that, for 0 m, k 2, all these roots are different when 2 + 1
is a prime number. We have thus written the (2 + 1)2 roots of our equation. Their
expression is
xem,k + em,k x em ek + ek em
1k m x = where em,k = .
1 c2 e2m,k x 2 1 c2 e2m e2
k
The roots of the equation x2+1 = 0 are the em,k , where e0,0 = 0. The non-zero roots
are given by an equation of degree 42 + 4 which may be decomposed in 2 + 2
equations of degree 2 with the help equations of degree 2 + 2. It is the result of
the Recherches of 1827 (see our 3), demonstrated here by a purely algebraic way.
Indeed, if p is a rational symmetric function of e1 , e2 , . . . , e2 , it may be expressed
as a rational function e1 of e1 such that e1 = e2 = . . . = e2 . Replacing
e1 by em,1 , we see that em,1 = em,2 = . . . = e2m,2 . It results that the sums
k = (e1 )k + (e0,1 )k + . . . + (e2,1 )k are rational symmetric in the 42 + 4
quantities em,k different from 0 and so rational functions of c. Thus p is the root
of an algebraic equation of degree 2 + 2 with coefficients rational in c. We may
apply this result to the coefficients of the algebraic equation of which the roots are
e1 , e2 , . . . , e2 .
According to the formula (126), the modulus c obtained from c by a transfor-
mation of order 2 + 1 is a rational symmetric function of e1 , e2 , . . . , e2 . It is thus
a root of an equation of degree 2 + 2 (the modular equation). Abel once more
says that this equation seems not to be solvable by radicals. He adds that, since
dx2+1 2+1 dy 2
x2+1 = y , the multiplication by 2 + 1 (which is of degree (2 + 1) )
may be decomposed in the transformation of order 2 + 1 from x to y and another
transformation of the same order from y to x2+1 . Jacobi also used such a decom-
position. The expressions of x2+1 and c in y and c are given by (126) with a root e
determined from c as e was from c. Thus the modular equation is symmetric in
(c, c ).
Abel recalls the total number of transformed moduli for a given order : 6
for = 1, 18 for = 2 and 6( + 1) for an odd prime number. Then he
explains the algebraic solution of the equation y = x where x is a rational
function defining a transformation. It is sufficient to consider the case in which the
order is an odd prime number 2 + 1 and we know that, in this case, the roots are
x, x, . . . , 2 x where m x = xe m +em x
1c2 e2 x 2
and 2+1 x = x. Let be a root of 1 and
m
v = x +x +2 2 x +. . .+2 2 x, v = x + 2 x +2 21 x +. . .+2 x. They
are of the form v = p + qx, v = p qx where p and q are rational functions of
x and vv = s, v2+1 +v 2+1 = t are rational functions of x. Since theyare invariant
2+1 t t4 2+1 .
by x x, they are rational functions of y and we have v = 2 + 4 s
If v0 , v1 , . . . , v2 are the values of v corresponding to the 2 + 1 roots of 1, we
1 1
obtain x = 2+1 (v0 + v1 + . . . + v2 ), m x = 2+1 (v0 + m v1 + . . . + 1m v2 ).
The last chapter of this first part deals with the following problem: Given an
elliptic integral of arbitrary modulus, to express this function by means of other
elliptic integrals in the most general way. According to the results of the second
chapter, this problem is expressed by the equation rdx

x = k1 1 y1 + k2 2 y2 + . . . +
km m ym + V where x = rdx

x is the given integral, 1 , 2 , . . . , m are elliptic

1 y1 2 y2 m ym
integrals of respective moduli c1 , c2 , . . . , cm , y1 , y2 , . . . , ym , x , x , . . . , x
are rational functions of x and V is an algebraic and logarithmic function. One
may suppose that the number m is minimal and, according to a theorem of the
fourth chapter, one has dy1 y11 = 1 x dx
, dy2 y22 = 2 x
dx
, . . . , dy m
m ym
= m xdx
where
1 , 2 , . . . , m are constant. Now, for 1 j m, there exists a rational function
x j of x such that (x j , c) = ( j) (x, c j ) and it results that there exits a rational
function y of x such that y be expressed as an elliptic integral of modulus c j where
x is the variable.
The part of the memoir published in Crelles Journal stops here and Sylow
completed it with a manuscript written by Abel and discovered in 1874. Here the
transformation of elliptic integrals of the second and third kinds is explained. For the
second kind, Abel proposes two methods. The first one is based on the differentiation
162 C. Houzel
with respect to the modulus c of the equation (y, c ) = (x, c), which gives

y2 dy 1

dc dy
c +
dc (1 c2 y2 )(y, c ) dc (y, c )
x 2 dx

d dx
= + c . (130)
dc (x, c) (1 c2 x 2 )(x, c)
x 2 dx 1 x(1x 2 ) 1 (1x 2 )dx

Now one can verify that (1c2 x 2 )(x,c)

= c2 1 (x,c)
+ 1c2 (x,c) and there

y2 dy
is a similar identity for (1c2 y2 )(y,c )
. Thus (130) is rewritten
c dc y(1 y2 )

dy 1
2
(y, c ) 0 (y, c )
+
1 c dc (y, c ) dc (y, c )
x(1 x 2 )

d c
= (x, c) + (x, c) 0 (x, c)
dc 1 c2 (x, c)
2 )

d(1c2 )
or 0 (y, c ) = A(x, c) + B0 (x, c) + p where A = 1 ccdc(1c
dc (1c2 ) c dc ,
c(1c2 )dc )dc dy 2 x(1x ) y(1y ) 2 2
B= and p = (1c
c (1c2 )dc c dc
1
dc (y,c ) + B (x,c) (y,c ) .
The second method is based on the decomposition of y2 in partial fractions:
A B
y2 = + +S
(x a)2 xa

where a is a pole of y and A, B are constant. If y = x 1
, A = (1a)2 and B = ( a)a3
and we have
(1 x 2 )(1 c2 x 2 )( x)2 = 2 (x)2 1 (x)2 c2 .

(131)
For x = a, this gives (1 a2 )(1 c2 a2 )( a)2 = 2c2 . Let us differentiate

(131) and
make x = a; we obtain 2(1a2 )(1c2 a2 ) a a 2(1 + c2 )a 4c2 a3 ( a)2 = 0
and we conclude that
1 (1 a2 )(1 c2 a2 ) a (1 + c2 )a + 2c2 a3
A= = , B= = .
( a)2 2 c2
( a) 3
2 c2
Thus
y2 dy (1 a2 )(1 c2 a2 ) 2c2 a3 (1 + c2 )a

1

dx
= 2 +
(y, c ) c (x a)2 x a (x, c)

Sdx
+ . (132)
(x, c)

(1a2 )(1c2 a2 ) 2c2 a3 (1+c2 )a
Now d (x,c)
xa = (xa)2
+ xa + c2 2
a c2 2
x dx
(x,c) and (132)
takes the form:
y2 dy

1

(x, c) Sdx
= 2 c2 a2 (x, c) + c2 0 (x, c) + .
(y, c ) c ax (x, c)
If the poles of y are a1 , a2 , . . . , a , we finally obtain
c2 0 (y, c ) = 0 (x, c) (c2 (a12 + a22 + . . . + a2 ) 2 c2 k2 )(x, c)

1 1 1
+(x, c) + + ... +
a1 x a2 x a x
where k is the value of y for x infinite. Abel separately considers the cases in which
k = 0 or k = 10 . This last case is reduced to the first one by putting x = cz1 . For
example, when

2 c x
c = , y = (1 + c) and = 1 + c ,
1+c 1 + cx 2
0 (y, c ) = c(1+c) 1+c 1+c x(x,c)

2 0 (x, c) + 2 (x, c) 2 1+cx 2 .
For the integral of the third kind, Abel uses the equation
1

dy ydy
= (y, c , a ) +

(a x)(y, c ) a (a y2 )(y, c )
2
and the decomposition in partial fractions

1 1 (a1 , c) (a2 , c) (a , c)
=k + + + ... +
a y (a , c ) a1 x a2 x a x
which lead to
(a , c )

ydy
(y, c , a ) + (a , c )
a (a y2 )(y, c )
2
(a, c)
= k1 (x, c) + (x, c, a) + v
a
where k1 is a constant and v is an algebraic and logarithmic function. Now the sum
of integrals in the right hand side may be reduced to a single integral with the help
of the result of the third chapter: if is determined by
( fx)2 (x)2 ((x, c))2 = (x 2 a12 )(x 2 a22 ) (x 2 a2 )(x 2 2 )
where fxand x are polynomials, one even and the other odd, according to (121)
(a,c) (,c) 1 fx+x(x,c)
we have a (x, c, a) = k2 (x, c) + (x, c, ) 2 log fxx(x,c) . The
coefficients of fx and x are determined by the equations fam + am (am , c) =
0(1 m ) and the sign of (, c) by f + (, c) = 0. Another way to
do this reduction consists in observing that if a is any one of a1 , a2 , . . . , a , that is
a root of a = (x), any other has the form am = a(em1c ,c)+em (a,c)
2 e2 a2 where em does
m
not depend of a. The same formula (121) with n = 3 and m 1 = m 2 = m 3 = 1 gives
164 C. Houzel
(am , c) (a, c)
(x, c, am ) = (x, c, a) + m (x, c)
am a
(em , c)
+ (x, c, em ) + log Sm
em
(em ,c)
and Abel shows that em (x, c, e) = 0.
A posthumous paper, Memoire sur les fonctions transcendantes de la forme
ydx, ou y est une fonctions algebrique de x (uvres, t. II, p. 206216) contains
extensions of the preceding

results to more general Abelian integrals. Abel first con-
siders integrals r j = y j dx(1 j ) where y j is an algebraic function of x and
he supposes that they are related by an algebraic relation R = (r1 , r2 , . . . , r ) = 0
where is a polynomial with coefficients algebraic with respect to x and is
minimal. He proves that in that case there is a linear relation
c1r1 + c2r2 + . . . + cr = P (133)
where c1 , c2 , . . . , c are constant and P is a rational function of x, y1 , y2 , . . . , y .

Indeed, one may suppose that R = rk + Prk1 + P1rk2 + . . . is irreducible with re-
spect to r (the coefficients P, P1 , . . . being rational with respect to r1 , r2 , . . . , r1 ).
By differentiation, one obtains
rk1 (ky + P ) + (k 1)Py + P1 rk2 + . . . = 0 ,

hence ky + P = 0 and kr + P = constant. This gives k = 1 and R = r + P = 0.

Now the decomposition of P in partial fractions with respect to r1 has the form
Sk
k
P= k
+ vk r1 ,
(r1 + tk )
where tk and vk are rational with respect to r1 , r2 , . . . , r2 ; by differentiation,

kSk (y1 + t ) Sk

k
+
(r1 + tk )k+1 (r1 + tk )k

k k1
+ (vk r1 + kvk r1 y1 ) = y
and this relation implies that Sk = 0 and vk = 0. Moreover, if k is not equal to 1, we

must have kvk y1 + vk1 = 0, but this would imply kvk r1 + vk1 = constant,
which is impossible. So k = 1 and P = v1r1 + P1 where v1 is a constant and
P1 is rational with respect to r1 , r2 , . . . , r2 . In the same way, we obtain, with
a slight change of notation, P j = v1 j r1 j + P j+1 (0 j 2) where
v1 , v2 , . . . , v1 are constant and P j is rational with respect to r1 , r2 , . . . , r1 j .
Finally, we have r + v1r1 + v2r2 + . . . + v1r1 + v0 = 0 where v0 is
an algebraic function of x and this gives a relation of the form (133) where P is
algebraic in x. Let P k + R1 P k1 + . . . = 0 be the minimal equation of P with coef-
ficients rational in x, y1 , y2 , . . . , y . Differentiating, we get (kdP + dR1 )P k1 +
((k 1)R1 dP + dR2 ) P k2 + . . . = 0 with dP dx = c1 y1 + c2 y2 + . . . , so that
kdP + dR1 = 0 and P = Rk1 + constant. This gives k = 1 and P = R1 , ra-

tional with respect to x, y1 , y2 , . . . , y .
In his next theorem, Abel considers a relation
c1r1 + c2r2 + . . . + cr = P + a1 log v1 + a2 log v2 + . . . + am log vm (134)
where v1 , v2 , . . . , vm are algebraic functions of x and P is a rational function of

x, y1 , y2 , . . . , y , v1 , v2 , . . . , vm . If vm is root of an equation of degree n with
(n)
coefficients rational in x, y1 , y2 , . . . , y , v1 , v2 , . . . , vm1 , let vm , vm , . . . , vm be
its n values. One has
1
c1r1 + c2r2 + . . . + cr = (P + P + . . . + P (n) )
n
+a1 log v1 + a2 log v2 + . . . + am1 log vm1
1 (n)
+ am log(vm vm vm )
n
where P + P + . . . + P (n) and vm (n)
vm vm are rational in x, y1 , y2 , . . . , y ,
v1 , v2 , . . . , vm1 . Iterating we finally obtain c1r1 + c2r2 + . . . + cr = P +
1 log t1 + 2 log t2 + . . . + m log tm where P, t1 , . . . , tm are rational functions of
x, y1 , y2 , . . . , y .
In particular, if y is an algebraic function of x and (x, y) a rational func-
tion such that the integral (x, y)dx is algebraic in x, y, log v1 , log v2 , . . . ,
log vm , then this integral may be expressed in the form P + 1 log t1 + 2 log t2 +
. . . + m log tm where P, t1 , . . . , tm are as above. If there is a relation (x, y)dx +
1 (x, y1 )dx = R where R is of the form of the right hand side of (134), and if
the minimal equation for y1 remains irreducible after adjunction of y, then one has
separately (x, y)dx = R1 and x (x, y1 )dx = R2 . For if y1 , y1 , . . . , y1(n) are the
values of the algebraic function x,
n(x, y)dx + (1 (x, y1 ) + 1 (x, y1 ) + . . . + 1 (x, y1(n) ))dx

= d(R + R + . . . + R(n) ) ,
hence a relation (x, y)dx = n1 (R + R + . . . + R(n) ) f(x)dx = R1 and

then x (x, y1 )dx = R R1 = R2 . If there is a relation ydx = R where y =
1 2 n1
p0 + p1 s n + p2 s n +. . .+ pn1 s n , p0 , p1 , . . . , pn1 , s algebraic functions such
1
that s n is not rational in p0 , p1 , . . . , pn1 , s, then one has separately pmmdx = R j

1 1 sn
(0 m n 1). Indeed dR = d f s n = s n dx and the same relation is true
1 1
for any value k s n of s n ( primitive n-th root of 1). It is easy to deduce that
1

pm dx
m = ( f( n s) + m f( n s) + . . . + (n1)m f(n1 n s)) .
s n n
The rest of the paper is not finished. Abel studies the cases in which an integral

1 1 1
y= f (x, (x a1 ) m 1 , (x a2 ) m 2 , . . . , (x an ) m n )dx
166 C. Houzel
( f is rational) is an algebraic function, and the corresponding reductions of Abelian

integrals. According to the preceding results, he is reduced to
k k
1 2 kn
dx p (x a1 ) m 1 (x a2 ) m 2 (x an ) m n
k k
1 m1 1 m2 kn
= P = v(x a1 ) 1 (x a2 ) 2 (x an )1 m n
k1 k2
where p and v are rational and m1 , m2 , . . . , mknn are between 0 and 1. This gives
p = v(A0 + A1 x + . . . + An1 x n1 )
dv dv
+ (x a1 )(x a2 ) (x an ) = vx + fx
dx dx
where
A0 + A1 x + . . . + An1 x n1

k1
= 1 (x a2 )(x a3 ) (x an )
m1

k2
+ 1 (x a1 )(x a3 ) (x an ) + . . .
m2

kn
+ 1 (x a1 )(x a2 ) (x an1 ) .
mn
1
Abel explains the cases in which v = x m or (x)m . In the first case
p = x m (A0 + A1 x + . . . + An1 x n1 )
+ mx m1 (B0 + B1 x + . . . + Bn1 x n1 + x n )
= m B0 x m1 + (A0 + m B1 )x m + (A1 + m B2 )x m+1 + . . .
+ (An1 + m)x n+m1 .
k k
m1 m2 kn
(x an ) m n = R , he gets

Putting x dx(x a1 ) 1 (x a2 ) 2
1 k
1 1
k
1 2 kn
Rm+n1 = x m (x a1 ) m 1 (x a2 ) m 2 (x an )1 m n
m + An1
m B0 An2 + m Bn1
Rm1 . . . Rm+n2
m + An1 m + An1
a recursion formula which permits to express Rm+n1 by R0 , R1 , . . . , Rn2 .
In the second case
x m fx
p= m

(x ) (x )m+1
mf
m f m f 2
= + + + ...
(x )m+1 (x )m (x )m1
(n1) f (n)
12(n1) m 12n
+ .
(x )mn+1
k k
m1 m2 kn
dx
(x an ) m n = S , he gets

Putting (x) (x a1 ) 1 (x a2 ) 2
k k
1 m1 1 2 1 kn
(xa1 ) 1 (xa2 ) m 2 (xan ) m n
(x)m
n1 m f (n)
= m fSm+1 + ( m f )Sm + . . . + 12(n1) 12n Smn+1 .
If f = 0, this permits to express Sm+1 in S1 , R0 , R1 , . . . , Rn2 . If f = 0 but

m f = 0, Sm is a linear combination of R0 , R1 , . . . , Rn2 . Now

k1
a1 m f a1 = 1 m (a1 a2 ) (a1 an ) = 0,
m1
so that the Sm with parameter a1 are linear combinations of R0 , R1 , . . . , Rn2 .
By the same method, Abel proves that a linear relation
c0 R0 + c1 R1 + . . . + cn2 Rn2 + 1 t1 + 2 t2 + . . . + t
k k
1 m1 1 m2 kn
= v(x a1 ) 1 (x a2 ) 2 (x an )1 m n ,
k k

dx m1 2 kn
where tk = (x k)
(x a1 ) 1 (x a2 ) m 2 (x an ) m n , is not possible. He
finally proves that, in a relation c0 R0 + c1 R1 + . . . + cn2 Rn2 + 1 t1 + 2 t2 +

. . . + t = P + 1 log v1 + 2 log v2 + . . . + m log vm , the right hand side may be

reduced to the form r1 1 + k log( (sk k k k )) where is the g.c.d. of

1 2 n
m 1 , m 2 , . . . , m n , for each k [0, 1], k = (x a1 ) m 1 (x a2 ) m 2 (x an ) m n ,
j being the remainder of the division of kk j by m j , is a primitive -th root of 1
and r1 , s0 , s1 , . . . , s1 are polynomials. First of all, the right hand side has the
form

r0 + r1 1 + . . . + r1 1 + log(s0 + s1 1 + . . . + s1 1 )
k kk
and when 1 is replaced by another value
fxdx1 , k becomes k . We thus get
expressions for the considered integral 1 and the terms rk k with k < 1
disappear from the sum of these expressions. It is then possible to prove that r1 = 0
and that the relations of the considered
type are combinations of those in which only
one is different from 0. In this case fxdx
1 = (x, 1 ) = log (1 )+ log (1 )+
2 2 1 1
log ( 1 ) + . . . + log ( 1 ) where (1 ) = s0 + s1 1 + . . . + s1 1
and Abel attacks the determination of the possible forms for fx, but the paper is left
incomplete (see Sylows note, uvres, t. II, p. 327329).
9 Series
We saw above (3) that in his first papers, Abel did not hesitate to use infinite series
in the 18th century manner, that is without any regard to questions of convergence.
On the contrary, when dealing with expansions of elliptic functions (6), he tried to
168 C. Houzel
treat the problem much more rigourously. In the meantime, he had read Cauchys
lectures at the Ecole Polytechnique and he was impressed by this work. In a letter to
Holmboe (16 January 1826), he writes On the whole, divergent series are the work
of the Devil and it is a shame that one dares base any demonstration on them. You
can get whatever result you want when you use them, and they have given rise to so
many disasters and so many paradoxes. Abel then explains that even the binomial
formula and Taylor theorem are not well based, but that he has found a proof for the
binomial formula and Cauchys lectures contain a proof for Taylor theorem.
The memoir Recherches sur la serie 1 + m1 x + m(m1)
12 x +
2 m(m1)(m2) 3
123 x +...,
published in the first volume of Crelles Journal (1826; uvres, t. I, p. 219250) is
devoted to a rigourous and general proof of the binomial formula. We have already
explained the formal part of this memoir (1) and we shall now analyse the part
where Abel studies questions of convergence. Abel defines a convergent series as
a series v0 + v1 + v2 + . . . + vm + . . . such that the partial sum v0 + v1 + v2 +
. . . + vm gets indefinitely nearer to a certain limit, which is called the sum of the
series, for increasing m, and he states Cauchys criterium for convergence. The first
theorem says that a series 0 0 + 1 1 + 2 2 + . . . + m m + . . . is divergent

when 0 , 1 , 2 , . . . are positive numbers such that m+1
m
has a limit > 1 and
the m do not tend towards 0. On the contrary (theorem II), if the limit is < 1
and the m remain 1, the series is convergent. The proof uses the comparison of
0 + 1 + . . . + m + . . . with a convergent geometric series and Cauchys criterium.
In the third theorem, Abel considers a series
t0 + t1 + . . . + tm + . . .
of which the partial sums pm = t0 + t1 + . . . + tm remain bounded by some

quantity and a decreasing sequence of positive numbers 0 , 1 , . . . , m , . . . The
theorem states that
r = 0 t0 + 1 t1 + 2 t2 + . . . + m tm
remains bounded by 0 . Abel uses what is now called Abel transformation, putting
t0 = p0 , t1 = p1 p0 , t2 = p2 p1 , . . . so that r = p0 (0 1 ) + p1 (1 2 ) +
. . . + pm1 (m1 m ) + pm m 0 .
Theorem IV concerns a power series f = v0 + v1 + v2 2 + . . . + vm m + . . .
and it says that if the series is convergent for a (positive) value of , it remains
convergent for the (positive) values and, for such an , the limit of f( )
for 0 is f. Abel puts = v0 + v1 + v2 2 + . . . + vm1 m1 and =
m m+1 m
vm m + vm+1 m+1 + . . . = vm m + vm+1 m+1 + . . . p where
p vm m , vm m + vm+1 m+1 , vm m + vm+1 m+1 + vm+2 m+2 , . . . (theorem III),
and this bound is arbitrarily small for m sufficiently large. Now f f( ) =
( ) + ( ) and,
since
m is a polynomial, it is sufficient to
m
bound ( ) by + p, which is easy to do.
In the following theorem, the coefficients v0 , v1 , . . . are continuous functions of
x in an interval [a, b] and Abel says that if the series is convergent for a value of ,
its sum fx for < is a continuous function in [a, b]. Unfortunately, this theorem is
not quite correct. Abels proof consists in writing fx = x + x where x is the sum
of
the
m terms up to m1 and x is the corresponding remainder, which is bounded by
m
x where x vm , vm m + vm+1 m+1 , vm m + vm+1 m+1 + vm+2 m+2 , . . .
(theorem III). For each x, this bound tends towards 0 as m but the convergence
is not necessarily uniform in x and Abels reasoning implicitly uses this uniformity.
Recall that Cauchy stated more generally that the sum of a convergent series of
continuous functions is continuous. In a footnote, Abel criticises this statement,
giving the series sin x 12 sin 2x + 13 sin 3x . . . as a counterexample: the series is
everywhere convergent but its sum is discontinuous for x = (2m+1) (where it is 0).
Theorem VI correctly states the formula for the product of two absolutely con-
vergent series v0 + v1 + v2 + . . . = p and v0 + v1 + v2 + . . . = p . Let (resp. m )

be the absolute value of vm (resp. vm ). The hypothesis is that 0 + 1 + 2 + . . . = u

and 0 + 1 + 2 + . . . = u are convergent and the conclusion that the series of

general term rm = v0 vm
+ v1 vm1 + v2 vm2 + . . . + vm v0 is convergent and that its
sum is equal to p p . Indeed r0 + r1 + r2 + . . . + r2m = pm pm + t + t where

pm = v0 + v1 + . . . + vm , pm = v0 + v1 + . . . + vm

,

t = p0 v2m + p1 v2m1 + . . . + pm1 vm+1 ,
t = p0 v2m + p1 v2m1 + . . . + pm1 vm+1 .

Now |t| u(2m + 2m1 + . . . + m+1 ), |t | u (2m + 2m1 + . . . + m+1 ) so that
t and t tend towards 0. This result had been given by Cauchy in the sixth chapter of
his Analyse algebrique (1821).
As an application, Abel considers two convergent series t0 + t1 + t2 + . . . , t0 +
t1 + t2 + . . . with real terms and such that the series t0 t0 + (t1 t0 + t0 t1 ) + (t2 t0 +

t1 t1 + t0 t2 ) + . . . is also convergent. Then the sum of this last series is equal to the
product of the sums of the two given series. Indeed, by theorem IV, it is the limit of
t0 t0 + (t1 t0 + t0 t1 ) + (t2 t0 + t1 t1 + t0 t2 )2 + . . . for 1 ( < 1). Since both
series t0 + t1 + t2 2 + . . . and t0 + t1 + t2 2 + . . . are absolutely convergent for
< 1 according to theorem II, the product of their sums is equal to
t0 t0 + (t1 t0 + t0 t1 ) + (t2 t0 + t1 t1 + t0 t2 )2 + . . .
and the conclusion is clear.
In the third volume of Crelles Journal, Abel published a Note sur un memmoire
de M.L. Olivier, ayant pour titre Remarques sur les series infinies et leur
convergence (1828; uvres, t. I, p. 399402). In his memoir, Olivier stated
a wrong criterium for the convergence of a series an : that nan must tend to-
wards 0. As a counterexample, Abel gives the divergent series of general term
1 1
an = n log n for which nan = log n tends towards 0. He proves the diver-
gence using the inequality log(1 + x) < x, which gives log 1 + n < n1 or 1

1 1
log log(1 + n) < log log n + log 1 + n log n < log log n + n log n . It results that
1 1 1
log log(1 + n) < log log 2 + 2 log 2 + 3 log 3 + . . . + n log n and the divergence follows
from lim (log log(1 + n)) = .
170 C. Houzel
More generally, Abel proves that there is no function n such that lim(nan ) = 0
be a criterium for the convergence of an . Indeed, when an is divergent, the
a3
same is true for the series aa10 + a0a+a
2
1
+ a0 +a1 +a2 + . . . + an
a0 +a1 +...+an1 + . . . for
log(a0 + a1 + . . . + an ) log(a0 + a1 + . . . + an1 )

an an
= log 1 + <
a0 + a1 + . . . + an1 a0 + a1 + . . . + an1
and log(a0 + a1 + . . . + an ) log a0 < aa10 + a0a+a

2
1
+ a0 +aa31 +a2 + . . . + a0 +a1 +...+a
an
n1
.
Now if n is a function such that n an 0 is a criterium of convergence, the
series
1 1 1 1 1
+ + + + ... + + ...
(1) (2) (3) (4) n
is divergent but
1 1 1
1
+ + + ...
(2). (1) 1
(3) (1) + 1
(4) 1
+ 1
+ 1
(2) (1) (2) (3)
1
+ + ...
1 1 1 1
n (1) + (2) + (3) + ... + (n1)
is convergent, which is contradictory.

Abel left unpublished a memoir Sur les series (uvres, t. II, p. 197205),
probably written at the end of 1827. He begins by giving the definition of convergence
and recalling Cauchys criterium. Then the first part deals with series of positive
terms and the second part with series of functions. The first theorem states that if
a series u 0 + u 1 + u 2 + . . . + u n + . . . with u n 0 is divergent, then the same
is true of us1 + us2 + us3 + . . . + su n + . . . , where sn = u 0 + u 1 + u 2 + . . . + u n
0 1 2 n1
and 1. It is an immediate extension of the preceding lemma, where was
taken
u n equal to 1. The following theorem says that, under the same hypotheses,

1+ is convergent when > 0. Indeed sn1 sn = (sn u n ) sn >
sn
un
1+ . For example, if u n = 1, the first theorem gives the divergence of the series
sn
1 + + 13 + 14 + . . . + n1 + . . . and the second theorem gives the convergence of the
1
2
1 1 1 1
series 1 + 2+1 + 3+1 + 4+1 + . . . + n +1 + . . . for >
0. When a series n is
divergent, a necessary condition for the convergence of u n is that
un
lim inf =0.
n
un
Indeed, if it is not the case, there exists > 0 such that pn = n for n

large enough and un n is divergent. Thus u n is convergent only if
lim inf nu n = 0 but this condition is not sufficient and Abel recalls the final result of
the preceding memoir. Abel next considers a function n increasing without limit,
implicitly supposed to be differentiable and concave, so that (n + 1) n n

and (0) + (1) + . . . + (n) > (n + 1) (0) and this implies the divergence
of (0) + (1) + . . . + (n) + . . . This applies to the iterated logarithm m n =
1
logm (n + a): n n = (n+a) log(n+a) log2 (n+a) logm1 (n+a)
and the series
1
2
n log n log n logm1 n
1 1
is divergent. On the contrary, when n = C 1 (logm n)1
where > 1,

(n + 1) n > (n + 1) and n < 1 (logm (n1))1 (logm1n)1 . It
1 1
1 1
results that (a) + (a + 1) + . . . + n < 1 (logm (a1))1
and the series
1
n log n log2 n logm1 n(logm n)1+
is convergent for > 0. Abel derives from this

statement

a rule for the
convergence of a series u n : the series is convergent if
log 1
u n n log n logm1 n
lim log m+1 n
> 1 and it is divergent if this limit is < 1. For instance, in
the first case, there exists an > 0 such that u n < n log n logm11 n(logm n)1+ for n
large enough.
Then first result stated by Abel on the series of functions is that when a power series
an x converges in ] , [, it may be differentiated term by term in this interval.
Abel returns to theorem V of his memoir on the binomial formula, which shows that
he was not satisfied with its proof. He considers 0 (y) + 1 (y)x + 2 (y)x 2 + . . . +
n (y)x n + . . . = f(y) and he supposes that it is convergent for 0 x < and y
near a value . Let An be the limit of n (y) when y tends towards and suppose
that A0 + A1 x + . . . + An x n + . . . is convergent. Then the sum R of this series is
the limit of f(y). Abel writes
f( ) R = (0 ( ) A0 ) + (x1 1 ( ) A1 x1 ) x2 + . . .
+ n ( )x1n A1 x1n x2n + . . .

where x = x1 x2 , x1 < , x2 < 1 and tends towards 0 and he chooses m such that
m ( )x1m A1 x1m n ( )x1n A1 x1n
k
for all n, so that f( ) = R + 1x 2
(m ( )x1m A1 x1m ) where 1 k 1.
Unfortunately, the value of m may depend on and the proof is still insufficient. As
Lie remarks in the final notes (uvres, t. II, p. 326), it is sufficient to suppose that
there exists M such that
(n ( ) An ) 0n M
for all n, for x1 < 0 < and for small enough in order to restaure a correct
proof. Abel applies his theorem to the series
172 C. Houzel
1y x + 2y x 2 + 3y x 3 + . . . + n y x n + . . . ,
1 1
sin y x + sin 2y x 2 + sin 3y x 3 + . . . ,
2 3
y y y
+ x+ x2 + . . .
1 + y2 4 + y2 9 + y2
continuous functions of y R when 0 x < 1; the second one is still convergent
when x = 1, but its sum has discontinuities as a function of y. The third one has 0
for limit when y tends towards , if x < 1, but the limit is 2 if x = 1. Abel adds
two remarks:
sin ay sin a2 y sin an+1 y n
I. the series y + y x + ... + y x + . . . is convergent for 0 x < 1
sin an+1 y
and y > 0, but when y tends towards 0, the limit An of y is an+1 , so that
the series
A0 + A1 x + . . . + An x n + . . .
is divergent when ax > 1.
II. the sum of
(y + 1)(y + 2) y 2
1+a+. . .+a y (1+2a+. . .+(y+1)a y )x+(1+3a+. . .+ a )x . . .
2
1 a ay
is equal to 1+x + (1+x)2
+ ... + (1+x) y+1
= fy for 0 x < 1 and y integer.
1
When y , this sum has for limit if a < 1 + x, but, if a 1, the
1+xa

y+n y
limit of n (y) = 1 + (n + 1)a + . . . + a is infinite and for a < 1, it
n
is (1a)1 n+1 = An . The series A0 + A1 x + . . . + An x n + . . . does not converge
when 1 x a < 1.
Abel gives an extension of his theorem IV of the memoir on the binomial series
to the case in which a0 + a1 + a2 2 + . . . is divergent. In this case, if an x n is
positive for n large, the limit of a0 + a1 x + a2 x 2 + . . . for x < tending towards
is infinite. The end of the paper contains a proof of Taylor theorem for a function
fx = a0 + a1 x + a2 x 2 + . . . defined by a power series convergent for 0 x < 1.
A lemma states that if
fx = (a0(0) + a1(0) x + a2(0) x 2 + . . . ) + (a0(1) + a1(1) x + a2(1) x 2 + . . . ) + . . .
+(a0(n) + a1(n) x + a2(n) x 2 + . . . ) + . . .
is convergent for 0 x < 1 and if A0 = a0(0) + a0(1) + . . . + a0(n) + . . . , A1 =

a1(0) + a1(1) + . . . + a1(n) + . . . , . . . then fx = A0 + A1 x + A2 x 2 + . . . + Am x m + . . .
whenever this series is convergent. Then Abel writes
f(x + ) = a0 + a1 (x + ) + a2 (x + )2 + . . .
= a0 + a1 x + a2 x 2 + . . . + (a1 + 2a2 x + . . . ) + . . .
f x f x 2
= fx + + + ...
1 12
if this series is convergent. It remains to prove the convergence under the condition
x + < 1. Abel writes x + = x1 and x = x1 x2 , so that x2 < 1 and
f nx
x1n = x1n an + (n + 1)an+1 x1n+1 x2
1 2n
(n + 1)(n + 2)
+ an+2 x1n+2 x22 + . . .
12
1
vn
(1 x2 )n+1
where vn is the least upper bound of an+k x1n+k for k 0. This gives
n
f nx

1 vn
n vn =
1 2n x1 x1 x2 1 x2 1 x2
where vn tends towards 0.
10 Conclusion
Two main subjects constitute the core of Abels work: algebraic equations and
elliptic functions, with an extension to the most general abelian integrals. As we
saw, they are intimately connected. Within our modern terminology, these subjects
may be symbolised by the terms Abelian group, which refers to a class of solvable
equations discovered by Abel, that is equations with a commutative Galois group,
and by the theorem of Abel on Abelian integrals and the term Abelian variety.
The theory of algebraic equations was one of the earliest fields of activity of
Abel. He proved the impossibility to solve by radicals the general quintic equation.
But later on he discovered that the so called Abelian equations are algebraically
solvable and he attacked the general problem to characterise solvable equations. He
obtained important results on the form of the solutions of solvable equations, and
this part of the theory was the point of departure of Kroneckers work in algebra.
Galois attacked the same problem from a different point of view, introducing the
Galois group which measures the indiscernability between the roots.
Abel studied elliptic integrals in Legendres Exercices de Calcul Integral, fol-
lowing Degens advice, and he immediately found fundamental new results. At the
same time, Jacobi began to investigate this subject and Abel was stimulated by the
competition with Jacobi. His theory contains all Jacobis results up to the year 1829,
but also some results of his own, as the study of the equation of division of an elliptic
integral or of a period of such an integral. Particularly important is his discovery of
complex multiplication which became a favourite subject for Kronecker and one of
the sources of class field theory.
Abels extension of the addition theorem for elliptic integrals to the general case
of Abelian integrals is rightly considered as one of the most important discoveries
in the first half of 19th century. It led Jacobi to formulate the inversion problem for
174 C. Houzel
hyperelliptic integrals. Through the works of Riemann and Clebsch, it became the
base of a new method to study the geometry of algebraic curves. Abels method to
prove this theorem contains in germ the notions of divisors and of linear families of
divisors on an algebraic curve and Riemanns interpretation of Abels result leads to
the notion of Jacobian of an algebraic curve.
With Gauss, Bolzano, Cauchy and Dirichlet, Abel is one of the reformators of
rigour in the first half of 19th century. Abels transformation of series gave him a way
to prove the continuity of the sum of a power series up to the end of the interval of
convergence in the case in which the series converges in this point. This theorem is
the base of a method of summation for divergent series.
Abel always tried to attack problems in the most general way instead of studying
particular cases and particular objects. In the theory of algebraic equations, he studied
the structure of a general expression built with radicals and he asked under which
conditions such an expression was the root of an algebraic equation of given degree.
In the theory of Abelian integrals, he investigated the most general algebraic relation
between given integrals and he proved that it is reducible to a linear relation. In the
case of elliptic functions, a further reduction led to complex multiplication. This part
of Abels work announces Liouvilles investigations on integration in finite terms
and his classification of transcendental functions. We saw the same concern with
generality in Abels treatment of functional equations. This general method of Abel
is well ahead of his time and close to the modern conception of axiomatic method.
References
[1] N.-H. Abel, uvres completes, ed. by L. Sylow and S. Lie, Christiania, 1881, 2 vol.
VIII+621+338 pages
[2] N.-H. Abel, Il manoscritto parigino di Abel, ed. A. Del Centina, Florence, 2002
[3] A.-L. Cauchy, Sur le nombre de valeurs quune fonction peut acquerir lorsquon y
permute de toutes les manieres possibles les quantites quelle renferme (1812), Journ.
Ec. Polytech., 17e cahier, t. X (1815), p. 128 = uvres, 2e serie, t. 1, p. 64 sq.
[4] A.-L. Cauchy, Sur les fonctions qui ne peuvent obtenir que deux valeurs egales et
de signes contraires par suite des transpositions operees entre les variables quelles
renferment (1812), ibid. = uvres, 2e serie, t. 1, p. 90169
[5] A.-L. Cauchy, Sur une loi de reciprocite qui existe entre certaines fonctions, Bull. Soc.
Philom., 1817, p. 121124 = uvres, 2e serie, t. 2, p. 223227
[6] A.-L. Cauchy, Seconde note sur les fonctions reciproques, ibid., 1818, p. 188191 =
uvres, 2e serie, t. 2, p. 228232
[7] A.-L. Cauchy, Cours dAnalyse de lEcole Royale Polytechnique; premiere partie:
Analyse algebrique, Paris, 1821 = uvres, 2e serie, t. 3; modern ed. by U. Bottazzini,
Bologne, 1992
[8] A.-L. Cauchy, Resume des Lecons donnees a lEcole Royale Polytechnique sur le Calcul
infinitesimal, Paris, 1823
[9] A.-L. Cauchy, Memoire sur lanalogie des puissances et des differences, et sur
lintegration des equations lineaires, Paris, 1825 = uvres, 2e serie, t. 15, p. 2340
[10] A.-L. Cauchy, Sur lanalogie des puissances et des differences, Exerc. de Math. 2, 1827
= uvres, 2e serie, t. 7, p. 198235
[11] A.-L. Cauchy, Sur les developpements des fonctions en series periodiques, Mem. Ac.
Sc. Paris, t. 6, 1827, p. 603612 = uvres, 1e serie, t. 2, p. 1219
[12] A. Clebsch, Ueber die Anwendung der Abelschen Functionen in der Geometrie, Journ.
fur die reine und angew. Math., t. 63, p. 189243
[13] A. Clebsch, Ueber diejenigen Curven, deren Coordinaten sich als elliptische Functionen
eines Parameters darstellen lassen, ibid., t. 64, p. 210270
[14] L. Euler, De progressionibus transcendentibus, seu quarum termini generales algebraice
dari nequeunt (1730), Comm. Ac. Sc. Imper. Petrop., 1st ser., n 5, 1738, p. 3857 =
Opera omnia, I14 , p. 124
[15] L. Euler, De formis radicum aequationum cuiusque ordinis coniectatio, ibid., n 6, 1738,
p. 216231 = Opera omnia, I6 , p. 119
[16] L. Euler, Introductio in Analysin Infinitorum, Lausanne, 1748, 2 vol. = Opera omnia,
I89
[17] L. Euler, Institutiones Calculi differentialis, Saint-Petersbourg, 1755 = Opera omnia,
I10
[18] L. Euler, De resolutione aequationum cuiusvis gradus, Nov. Comm. Petrop., 9, 1764,
p. 7098 = Opera omnia, I6 , p. 170196
[19] L. Euler, Institutiones Calculi integralis, Saint-Petersbourg, vol. II, 1769 = Opera
omnia, I12
[20] L. Euler, Demonstratio theorematis Newtoniani de evolutione potestatum binomii pro
casibus, quibus exponentes non sunt numeri integri, 1773, Opera omnia, I16 , p. 207216
[21] J.-B. Fourier, Theorie analytique de la chaleur, Paris, 1822 = uvres, t. I, p. 1563
[22] J.-B. Fourier, Theorie du mouvement de la chaleur dans les corps solides (1811), Mem.
Acad. Sciences, t. 4, 1824 = uvres, t. I
[23] E. Galois, Memoire sur les conditions de resolubilite des equations par radicaux (1831),
Ecrits et Memoires mathematiques, Paris, 1976, p. 43101
[24] C.-F. Gauss, Disquisitiones arithmetic,
Braunschweig, 1801 = Werke, t. I
[25] C.-F. Gauss, Varia imprimis de integrali du 2
, Werke, t. III, p. 433435
1+ sin u
[26] C.-F. Gauss, Theoria sinus lemniscati universalissime accepti, Werke, t. X1 , p. 194206
[27] C.-F. Gauss, Zur Theorie der Transcendenten Functionen gehorig, Werke, t. III,
p. 436445
[28] C.-F. Gauss, Zur Theorie der neuen Transcendenten, Werke, t. 3, p. 446460
[29] C.-F. Gauss, Hundert Theoreme uber die neuen Transcendenten, Werke, t. 3, p. 461469
[30] A. Genocchi, Sulla formula sommatoria di Eulero e sulla teorica dei residui quadratici,
Ann. Sc. mat. fis., t. 3, 1852, p. 406436
[31] A. Genocchi, Intorno ad alcune formule sommatorie, ibid., t. 6, 1855, p. 70114
[32] C. Huygens, Horologium oscillatorium, Paris, 1673
[33] C.G. Jacobi, Letter to Schumacher, Astron. Nachr., 6, 1827 = Ges. Werke, t. 1, p. 3133
[34] C.G. Jacobi, Fundamenta nova theori functionum ellipticarum, Konigsberg, 1829 =
Ges. Werke, t. 1, p. 55236
[35] C.G. Jacobi, Considerationes generales de Transcendentibus abelianis, Journ. Fur die
reine und angew. Math., 9, 1832, p. 394403 = Ges. Werke, t. 2, p. 716
[36] C.G. Jacobi, De functionibus duarum variabilium quadrupliciter periodicis, quibus
theoria transcendentium abelianarum innititur, ibid., 13, 1834, p. 5578 = Ges. Werke,
t. 2, p. 2550
[37] L. Kronecker, Ueber die algebraisch auflosbaren Gleichungen I, Monatsb. K. Preuss.
Ak. Wissensch. Berlin, 1853, p. 365374 and 1856, p. 203215 = Math. Werke, t. 4,
p. 111 and 2537
176 C. Houzel
[38] L. Kronecker, Ueber die elliptischen Functionen, fur welche complexe Multiplication
stattfindet,, 1857, p. 455460 = Math. Werke, t. 4, p. 179183
[39] L. Kronecker, Ueber die complexe Multiplication der elliptischen Functionen, ibid.,
1862, p. 363372
[40] L. Kronecker, Bemerkungen uber die Darstellung von Reihen durch Integrale, Journ.
Fur die reine und angew. Math., t. 105, p. 157159 and 345354 = Mat. Werke, t. 5,
p. 329342
[41] J.-L. Lagrange, Reflexions sur la resolution algebrique des equations, Nouv. Mem. Ac.
Roy. de Berlin, (1771) 1773, p. 138253 = uvres, t. 3, p. 305421
[42] J.-L. Lagrange, Sur une nouvelle espece de calcul relatif a la differentitation et a
lintegration des quantites variables, ibid., 1772 = uvres, t. 3, p. 441476
[43] J.-L. Lagrange, Theorie des fonctions analytiques, Paris, 1797
[44] J.-L. Lagrange, Lecons sur le calcul des fonctions, Paris, 1806 = uvres, t. 10
[45] P.S. Laplace, Memoire sur les approximations des formules qui sont fonctions de tres
grands nombres (1782), Mem. Ac. Royale des Sc., 1785 = uvres, t. 10, p. 209291
[46] A.-M. Legendre, Exercices de Calcul Integral, 2 vol., Paris 1811
[47] G.W. Leibniz, Mathematische Schriften, ed. C.I. Gerhardt, vol. 2, Berlin, 1849
[48] J. Liouville, Memoire sur quelques questions de geometrie et de mecanique, et sur un
nouveau genre de calcul pour resoudre ces questions, Journal de lEc. Polytech., 13,
sect. 21, 1832, p. 169
[49] J. Liouville, Memoire sur le calcul des differentielles a indices quelconques, ibid.,
p. 71162
2
[50] J. Liouville, Memoire sur lintegration de lequation (mx 2 + nx + p) ddx 2y + (qx + r) dydx
+ sy = 0 a laide des differentielles a indice quelconque, ibid., p. 163186
[51] J. Liouville, Memoire sur le theoreme des fonctions complementaires, Journ. fur die
reine und angew. Math., t. 11, 1834, p. 119
[52] J. Liouville, Sur la classification des transcendantes et sur limpossibilite dexprimer
les racines de certaines equations en fonction finie explicite des coefficients, Journ. de
Math. pures et appl., t. 2, 1837, p. 56105, t. 3, 1838, p. 523547 and t. 4, 1839,
p. 423456
[53] J. Liouville, Sur la determination des integrales dont la valeur est algebrique, ibid., t. 3,
1838, p. 2024
[54] A. de Moivre, quationum quarundam Potestatis terti, quint, septim, non, et
superiorum, ad infinitum usque pergendo, in terminis finitis, ad instar Regularum pro
cubicis, qu vocantur Cardani, Resolutio Analytica, Phil. Trans., 25, n 209, 1707,
p. 2368
[55] G.A.A. Plana, Sur une nouvelle expression analytique des nombres bernoulliens propre
a exprimer en termes finis la formule generale pour la sommation des suites, Mem. Ac.
Torino, t. 25, 1820, p. 403418
[56] S.-D. Poisson, Note sur le developpement des puissances des sinus et des cosinus en
series de sinus ou de cosinus darcs multiples, Correspondance sur lEcole Polytech-
nique, vol. 2, n 3 (1811), p. 212217
[57] B. Riemann, Versuch einer allgemeinen Auffassung der Integration und Differentiation
(1847), Coll. Works, New York, 1953, p. 353366
[58] B. Riemann, Theorie der Abelschen Functionen, Journ. fur die reine und angew. Math.,
54, 1857 = Coll. Works, p. 88144
[59] P. Ruffini, Teoria generale delle equazioni in cui si dimostra impossibile la soluzione
algebraica delle equazioni generali di grado superiore al quarto, Bologne, 1799
[60] P. Ruffini, Della soluzione delle equazioni algebraiche determinate particolari di grado
superiore al quarto, Mem. Mat. Fis. Soc. Ital., 9, 1802 = Op. Mat., t. 1, p. 343406
[61] P. Ruffini, Riflessioni intorno alla soluzione delle equazioni algebraiche generali,
Modene, 1813 = Op. Mat., t. 2, p. 155268
[62] M. Schaar, Memoire sur les integrales euleriennes et sur la convergence dune certaine
classe de series, Mem. Couronnes et Mem. des savants etrangers publ. Par lAc. De
Belgique, t. 22, 1848, p. 325
[63] A. Stubhaug, Et foranskutt lyn. Niels Henrik Abel og hans tid, Oslo, 1996, 598 pages;
english transl. Niels Henrik Abel and his times, Springer-Verlag, Berlin, 2000, X+580
pages
[64] G. Zolotareff, Sur la methode dintegration de M. Tchebychef, Math. Ann., 5, 1872,
p. 560580

Abels Complete Work Houzel PDF

Uploaded by

Copyright:

Available Formats

Abels Complete Work Houzel PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Abels Complete Work Houzel PDF

Uploaded by

Copyright:

Available Formats

The Legacy of Niels Henrik Abel The Abel Bicentennial, Oslo 2002

Springer-Verlag 2004. (Editors O. A. Laudal and R. Piene)

The Work of Niels Henrik Abel

f(x, y) = (x) + (y) (5)

x + y = (x f y + y fx) = (r), (6)

where r = x f y + y fx; this equation includes, as particular cases, thelaws of

where a = (0), = f(0) and = f (0), a differential equation which deter-

f x ( fx + x) + (mx fx) = 0, (9)

which determines f ; as this differential equation is homogenenous, it is easily

for instance x = logloglogn x if = 0 and a = e.

the series (for |x| 1) in the form of an integral

for (x, y) in the interior domain of the figure

The constant of integration is determined by taking y = 0 and is found to be a.

so that (m) = Am = (1 + x)m for m rational as was observed by Euler. Lagrange

f(k + , k + ) = f(k, k ) f(, ); (14)

where m is an integer, which must be constant because of the continuity of . In

with a = 2m + (0, k ) + (0, ), whence

where c is a function of k , . Indeed, taking = k, 2k, . . . , k in (15) and adding

(k, k + ) = ck + 2m + (0, k ) + (0, ), (17)

where c = (k , ), a function of k and . For k = 0, this gives

(0, k + ) = 2m + (0, k ) + (0, ),

with an arbitrary constant ; then (17) becomes

also equal to 2m + (k, k ) + (0, ) = (k, k ) + by (14), so that (k, k ) =

(k, k ) = k + k 2m. (18)

and it remains to determine the constants , , and .

f = ek cos ks cos k ek sin ks sin k,

and cos k = ek ( fcos ks+sin ks), sin k = ek (cos ks fsin ks),

this is Cauchys result for f + i = |1 + x|k (cos ks + i sin ks) = (1 + x)k .

so that (m) = (1 + x)m .

by making tend toward 1, 12 log(22 cos ) = cos 12 cos 2 13 cos 3. . .

for 4 4 (for = 4 , m must be > 1).

(1 + x)m (cos i sin ) :

where is an integer such that | 2| (with the restriction m > 0 in

with P(x) = a + a1 x + a2 x 2 + a3 x 3 + a4 x 4 + a5 x 5 + x 6 and c + c1 x j + c2 x 2j +

solution becomes y = x m+1 tan(log k x). Another easy case of integration

(y + s)dy + ( p + qy + ry2 )dx = 0,

which is reduced to the form zdz +

dx e rdx . If P = 0, this equation has the solution z =

(y + s)dy + (qs rs 2 + qy + ry2 )dx = 0

and = 2c qdx qdx

A function r = leads to the conditions d

 Another form tried by Abel is r = a log(  + y); he finds

2 Integral Transforms and Definite Integrals

equation a = (ax)n where is a given function, s an unknown function of x

(k) ak ( is implicitly supposed to be analytic); by identification,

Therefore, in the original problem a = (ax)n , one has

In this paper, there is no mention of derivatives of non-integral order.

giving as a particular case cos y =

formula + . . . and Taylor series for

gets (1+t 2 )(etdtt et ) = 12 log 2 14 .

As an example, for x = eax and n = 2, this formula gives:

(x, y, z, . . . ) = exu+yv+z p+... f(u, v, p, . . . )dudvd p . . . ,

and he calls the generating function of f and f the determinant function of ,

Dn x = (ev 1)n fv, fg((ev 1)n fv) = n x, Dn (x) = (eva 1)n fv

(ev 1)n = exp(n log(ev 1)), vn = en log v , and (1 + v)n = en log(1+v) .

x = fg(log(ev 1) fv), log v fv and (1 + v) fv;

these operators are respectively x = x + d x, 1 x 12 12 x + 13 13 x

which is obtained in the following manner: the equality

In an addition to this paper (Sur quelques integrales definies, uvres, t. II,

Another form tried by Abel is r = a log( + y); he finds