The first Memoir of Galois.

13th April 2015

The prehistory of the concept of extension of a field.

I It is important to note that Galois, by help of a quantity

considered as known, wants to obtain a new structure with
the same properties of the original one.
 Let
ϕ(x) = 0 (1)
be an equation without multiple roots having its coefficients
in a field K of characteristic 0. Let x1 , x2 , . . . , xn be the
roots of the equation and denote by P the set of all the
permutations of the roots. Galois states, leaving the proof
to his reader, that
Lemma
It is possible to find a rational function

V :P→K

which assumes different values over all the permutations. In

particular, we can choose n elements of the field K,
a1 , a2 , . . . , an , such that the function defined by

V : (xi1 xi2 . . . xin ) 7→ a1 xi1 + a2 xi2 + · · · + an xin

has this property.

Two simple examples

I Consider a second degree equations whose roots are x1 , x2

I V : (xi , xj ) 7→ xi − xj is a possible choice.
I Consider a third degree equation

x3 + px2 + qx + r = (x − x1 )(x − x2 )(x − x3 ) = 0.

I V : (xi , xj , xk ) 7→ xi − xj is again a possible choice if the

equation (polynomial) is irreducible.
I If the degree is greater than 3 there is not an obvious
choice.
Dedekind’s proof of the Lemma.
Let ai = hi−1 and consider the polynomial
Y
p(t) = [t − (xi1 + xi2 h + · · · + xin hn−1 )]
i1 ,i2 ,...,in

The polynomial p(t) has two coincident roots if and only if

there is an equality of the form

xi1 + xi2 h + · · · + xin hn−1 = xj1 + xj2 h + · · · + xjn hn−1 .

where (xi1 xi2 . . . xin ) and (xj1 xj2 . . . xjn ) are two different
permutations. If we consider the discriminant of this
polynomial1 we have a polynomial in h that has only a finite
number of roots. Hence it is possible to choose a value of h
which is not a root. 

1
Which is different from 0 as it is easily proved.
Example
Consider a second degree equation having the two different
roots x1 , x2 . Let it be

ϕ(x) = x2 + px + q = (x − x1 )(x − x2 ).

We have to consider
p(t) = [t − (x1 + hx2 )][t − (x2 + hx1 )] =
(2)
= t2 − (x1 + x2 )(1 + h)t + x1 x2 + (x21 + x22 )h + x1 x2 h2

The discriminant of the polynomial p(t) is

(h − 1)2 (x1 − x2 )2

and it follows that it is sufficient to take h different from 1.

 A function V which has the peculiarities indicated in
Lemma 1 sometimes is called a Galois Resolvent. Let us
choose now a value of V over a fixed permutation, for
example the identity (which we denote by 1),

V1 = V (x1 , x2 , . . . , xn )

and let F (t) be its minimum polynomial.2

Theorem
All the roots of F (t) have the form

V (xi1 , xi2 , . . . , xin ).

2
This denomination is modern, but the concept is surely within Galois’s
reach.
Proof.
Denote by Vσ , Vτ , . . . the new values that V1 assumes when we
apply all the permutations σ, τ, . . . to (x1 x2 . . . xn ) and consider

f (t) = (t − V1 )(t − Vσ )(t − Vτ ) · · · (3)

The Theorem of Symmetric Polynomials grants that the

polynomial f (t) has its coefficients in K and it has the root V1 .
From the properties of the minimum polynomial it follows that
F (t)|f (t). 
 We prove now that K(V1 ) is the splitting field of the
polynomial ϕ(x) at the left side of the equation (1). We
give to this theorem the form chosen by Galois.
Theorem
Let
V :P→K
be an injective function and set

V1 = V (x1 , x2 , . . . , xn ).

Then all the roots of ϕ(x) = 0 are rational functions of V1 .

Proof.
Let us introduce a new variable t and consider the polynomial

p(t) = [t − V (x1 , x2 , x3 , . . . )][t − V (x1 , x3 , x2 , . . . )] · · · (4)

which is obtained by considering all the permutations that fix

one root, in this case x1 . This polynomial has for its coefficients
symmetric functions of the remaining roots x2 , . . . , xn .
Denote by u1 , u2 , . . . , un the elementary symmetric functions of
all the variables x1 , x2 , . . . , xn and by v1 , v2 , . . . , vn−1 the
elementary symmetric functions of the x2 , . . . , xn alone. We
have the relations3

v1 = u1 − x1 , v2 = u2 − u1 x1 + x21 , . . . (5)

that allow us to write the polynomial (4) simply as a

polynomial, with known coefficients, in the variables t, x1 . We
denote this polynomial by p(t, x1 ).
3
Since
x − u1 xn−1 + · · · + (−1)n un = (x − x1 )(xn−1 − v1 xn−2 + · · · + (−1)n−1 vn−1 ).
n
Proof (continued).
We have identically p(V1 , x1 ) = 0, and from this identity we
may deduce that the equation, which is obtained by
substituting the given root x1 by a variable x,

p(V1 , x) = 0

has the root x1 .

Since we have also ϕ(x1 ) = 0, the two equations
p(V1 , x) = 0, ϕ(x) = 0 have at least the common root x1 .
Hence, if d(x) denotes their greatest common divisor, it is
d(x1 ) = 0.
By the very definition of greatest common divisor, the only
other possible roots are x2 , x3 , . . .
We proof that x1 is the only root that the two polynomials
p(V1 , x), ϕ(x) have in common.4

4
This explanation is given in [2, p. 256]. Galois gives only a brief
indication.
Proof (continued).
Suppose by contradiction that, for example, d(x2 ) = 0.
Let us introduce a new variables y and consider the equation

p(y, x2 ) = 0.

This equation is satisfied by posing y = V1 . But this equation

may be deduced from the equation (4) by the variable
substitutions x1  x2 .
No root of this new equation is a root of (4) since the chosen
substitutions of variable fix x2 in the first position. It follows
that p(V1 , x2 ) cannot be 0.
Hence d(x) is a first degree polynomial which has the root x1 ,
and it may be obtained by successive divisions starting from the
polynomials ϕ(x) and p(V1 , x). It has its coefficients in K(V1 )
and, since it is a first degree polynomial, also its root x1 is in
this field.
Of course to express xi as a function of V1 , for i = 2, 3, . . . , n we
have to fix the roots xi in the function V and to construct the
polynomials corresponding to the one given by (4). 
Example
Consider the third degree equation ϕ(x) = x3 + px + q = 0, and
let x1 , x2 , x3 be its roots. Supposing that ϕ(x) is irreducible, V
may be chosen in the form5

V = x1 − x2 .

We have

p(t) = [t−(x1 −x2 )][t−(x1 −x3 )] = t2 −[2x1 −(x2 +x3 )]t+x21 −(x2 +x3 )x1 +

Observing that

x2 + x3 = −x1 , x1 (x2 + x3 ) + x2 x3 = p, x2 x3 = p + x21 ,

we have
p(t, x1 ) = t2 − 3x1 t + p + 3x21 .
5
We have to consider two similar cases: if x1 − x2 = x2 − x1 then
x1 = x2 . If x1 − x2 = x2 − x3 then 2x2 = x1 + x3 , hence
3x2 = x1 + x2 + x3 = 0. In both cases ϕ(x) is reducible.
Example (continued)
We set, for sake of simplicity, V1 = V . We have to consider the
polynomials (x1 ← x, t ← V )

x3 + px + q, 3x2 − 3V x + p + V 2 .

The remainder of the division of the first by the second is

2 1 1
V 2 + p x − V 3 − V p + q.


3 3 3
It follows that
1 3 q
x1 = V − .
2 2p+V2
The Galois Group of an equation

I Let ϕ(x) be again the polynomial having the simple roots

x1 , x2 , . . . , xn and let V (x1 , . . . , xn ) be a Galois Resolvent.
I Let F (t) be the minimum polynomial of V1 and let
V2 , . . . , Vr be the other roots of this polynomial (r ≤ n!).
I We want to prove that the substitutions corresponding to
the roots V1 , V2 , . . . , Vr form a group.
I By definition this group is the Galois group.
Proof.
As a consequence of Theorem 4, we have

x1 = f1 (V1 ), x2 = f2 (V1 ), . . . , xn = fn (V1 ). (6)

Now
∀k, ϕ(xk ) = 0 ⇒ ϕ(fk (V1 )) = 0
so that V1 is a root of ϕ(fk (t)). It follows that every root of F (t)
is a root of ϕ(fk (t))6 and consequently fk (Vi ) is a root of ϕ(x).
That means that by substituting Vi for V1 into (6) we obtain n
roots of ϕ(x). We prove that they are all distinct. Indeed, from

fk (Vi ) = fj (Vi )

it follows that Vi is a root of fk (t) − fj (t) = 0; then also

fk (V1 ) − fj (V1 ) = 0, which contradicts the hypothesis
xk 6= xj .
6
If χ(t) is an irreducible polynomial which has a root in common with a
rational function Φ(t) then all the roots of χ(t) are roots of Φ(t). Galois
states this property in his Lemma I [1, p. 47].
Proof (continued).
Consider now the permutations

1, σ, τ, . . .

corresponding to the roots of F (t) = 0. Let Vi be one of these

roots: it has the form

Vi = V (xi1 , xi2 , . . . , xin ).

Since all the roots are rational functions of V1 , we have

Vi = θ(V1 ).

Hence V1 is a root of F (θ(t)) so that F (θ(Vj )) = 0, for

j = 2, . . . , r. An index k exists such that θ(Vj ) = Vk . Denote by
σi the permutation which carries the array of the roots
corresponding to V1 into the one corresponding to Vi , by σk the
permutation which carries the array of the roots corresponding
to V1 into the one corresponding to Vk and lastly by σj the
permutation which carries the array of the roots corresponding
 Proof(continued).
We have the situation described by the following diagram:

V1 fα (V1 ) fβ (V1 )
σi
 
Vi fα (Vi ) = fα (θ(V1 ))
σj
 
Vj fβ (Vj ) = fα (θ(Vj ))

σk

Vk fα (Vk )

That means that σj σi = σk . The permutations corresponding to

the roots of F (t) constitute a group. 
The third degree equation
I Consider the simple example given by an irreducible third
degree equation which can be assumed to be of the form7
x3 + px + q = 0. (7)

I Denote by x1 , x2 , x3 its roots. The function

V : (x, y, z) 7→ x − y takes, over the six arrangements of the
roots distinct values.
I It is easy to calculate the Galois Group. Actually we have
R(t) = t6 + 6 pt4 + 9 p2 t2 + 4 p3 + 27 q 2 = 0. (8)

I The polynomial R(t) may be written in the following way:

2
t3 + 3pt + 4 p3 + 27 q 2 . (9)
| {z }
−Discriminant
7
A substitution of variables given by x ← x + h does not affect the
calculation of the Galois group, as it is plainly evident.
Cubic Roots

I The equation x3 − 2 = 0:

R(t) = t6 + 108.

The group is S3 .
I The discriminant of x3 − 2 is −108. If we adjoin this value
we have
 √  √ 
R(t) = − −t3 + 6 i 3 t3 + 6 i 3 .

√
3 1 i√
Let w = 2, α = − + 3.
2 √2
I The polynomial t3 + 6i 3 has the root w − α · w and
describes the group.
Cyclic group and the group reduced to the identity
I The equation x3 − 3x + 1 = 0):

R(t) = t3 − 9 t − 9 t3 − 9 t + 9 .

The group is C3 .
I In this case the discriminant is 81.
: : : : :
[R] We can use the same Resolvent also in the reducible case,
provided that it has distinct values as it happens in this
case.

x3 − 7x + 6 = 0,

R(t) = (t − 1) (t − 4) (t − 5) (t + 5) (t + 4) (t + 1) .
[R] The group is reduced to the identity.
Galois Group of x4 − 2 =√0 √ √ √
I We list the roots: a = 4 2, b = i 4 2, c = − 4 2, d = −i 4 2.
I A resolvent can be chosen as
V : (x, y, z, t) 7→ 1 · x − 2 · y + 4 · z.
I R(t) = t24 − 1740 t20 − 996420 t16 + · · ·
I This polynomial factors into three polynomials of degree 8.
We have to chose the factor that has a root given by
√
4
√
4
√
4
1 · 2 − 2 · (i 2) + 4 · (− 2).
I The roots of the polynomial t8 + 476 t4 + 114244 describe
the Galois Group.
I Its roots correspond to the alignments:
[a, b, c, d], [a, d, c, b], [b, a, d, c], [b, c, d, a],
[c, b, a, d], [c, d, a, b], [d, a, b, c], [d, c, b, a].
I The group is generated by
   
a b c d a b c d
σ= , τ= .
b c d a b a d c
Geometrical interpretation of the Galois Group of the
equation x4 − 2 = 0

I An anticipation:
√ if we adjoin one of the two values
(±2i 2) that the function g : (x, y, z, t) 7→ xy + zt assumes
on the alignments of the roots the group reduces to
[a, b, c, d], [b, a, d, c], [c, d, a, b], [d, c, b, a].
E. Galois.
Écrits et mémoires mathématiques d’Évariste Galois.
Gauthier-Villars, Paris, 1976.
Publiés par R. Bourgne et J. P. Azra. Préface de J.
Dieudonné. Deuxième édition revue et augmenté.
Réimpression autorisée, Editions J. Gabay, 1997.
C. Jordan.
Traité des substitutions et des équations algébriques.
Gauthier-Villars, Paris, 1870.
Éditions Jacques Gabay, 1989.