ALGEBRA I : FIELD THEORY AND GALOIS THEORY

J. K. VERMA
Lecture 1 : Overview
Objectives
(1) A historical sketch of main discoveries about formulas for roots of
polynomials.
(2) Problems of classical Greek geometry.
(3) Discussion about the main themes of the course.
Key words and phrases: Quadratic, cubic and quartic equations, du-
plication of cube, trisection of an angle, construction of regular polygons,
Galois group.
1. Introduction and Overview
In this course we will study elds, Galois theory of eld extensions and
applications to geometry and theory of equations. We outline the main
topics that we will study.
The formula
x =
b

b
2
4ac
2a
for the roots of the quadratic equation ax
2
+ bx + c = 0 was known to
Babylonians. During the reign of King Hammurabi (1750 B. C.), Babylonian
mathematicians found methods of solving linear and quadratic equations in
one and two variables. They described algorithms to solve specic examples.
From these examples it is clear that they knew the formula for the roots of
quadratic equations.
In 1494, an Italian mathematician Franciscan Luca Pacioli published the
book Summa de Arithmetica, Geometria, Proportioni et Proportionalita con-
taining all that was known in that period in arithmetic, algebra, geometry
2
and trigonometry. Paciolo ended his book with a remark that solutions of
cubic equations seemed impossible.
Generations of mathematicians at the University of Bologna in Italy tried
to nd solutions of cubic equations. This was the largest and one of the most
famous universities at the turn of sixteenth century in Europe. Scipio Del
Ferro at this univerisity solved the cubic but never published his ndings.
In 1535, Niccolo Tartaglia, a mathematician from Venice proved in a
public demonstration that he could solve cubic equations. But He kept
his formula a secret. But a doctor from Milan Gerolamo Cardano obtained
these formulas from Tartaglia under an oath that he will keep them a secret.
Cardano wrote his textbook Ars Magna in 1545 which described Tartaglias
method and extended it to all cubic equations. It is easy to see that the
equation x
3
+ax
2
+bx+c = 0 is transformed into the equation x
3
+px+q = 0
by replacing x by xa/3. Let x
1
, x
2
, x
3
denote the roots of x
3
+px+q = 0,
= (4p
3
+ 27q
2
) and w = 1/2 1/2

3. Let
y
1
= x
1
+ w
2
x
2
+ wx
3
y
2
= x
1
+ wx
2
+ w
2
x
3
Then Cardanos formulas are
y
1
=
3

27
2
+
3
2

3
y
2
=
3

27
2

3
2

3
We also have x
1
+ x
2
+ x
3
= 0. These three linear equations determine the
roots x
1
, x
2
, x
3
.
In 1536 Lodovico Ferrari entered Cardanos house as a servant. Due to
his extraordinary mathematical abilities he became a mathematician under
Cardanos guidance. Ferrari showed that a quartic equation can be reduced
to a cubic equation and therefore it can be solved by means of four arithmetic
operations and extraction of square and cube roots. We will derive the
formulas of Cardano and Ferrari later.
Some of the greatest mathematicians, e.g., Euler and Lagrange attempted
to nd similar formula for the roots of quintic equations. Lagrange gave a
3
general method to solve equations of degree atmost four. But this method
did not work for quintic equations.
Mathematicians became skeptical about existence of such formulas for
equations of degree ve and higher. Paolo Runi, born 1765 was a student
of Lagrange. He published several papers(1802, 1813) about insolvability of
general quintic equation. His proof was not complete. The rst complete
proof was given by Neils Henrik Abel (1802-1829) in 1824. Abel also proved
that if the Galois group of the polynomial is commutative then the poly-
nomial is solvable by radicals. Commutative groups are called Abelian to
honour Abel for his deep work in many branches of mathematics.
Gauss made two fundamental contributions to the theory of equations. He
provided complete solution by means of radicals of the cyclotomic equation
x
n
1 = 0.
The roots of this equation are the complex numbers represented by the ver-
tices of a regular polygon of n sides centered at the origin. Gauss analysis
of the roots of cyclotomic equation led him to nd a criterion of the con-
structibility of regular polygons of n sides. We will discuss this later.
The second contribution was rst rigorous proof of of the Fundamental
Theorem of Algebra: Every polynomial with complex coecients is a product
of linear factors with complex coecients.
The most decisive results in the theory of equations were found by Evariste
Galois (1811-1832). Modern algebra began with the work of Galois. He in-
troduced the Galois group of a polynomial which connected eld theory with
group theory. In 1829, Galois presented two papers to the Paris Academy of
Sciences. These were sent to Cauchy who lost them. In 1830 he sent another
paper to the Academy whose secretary was Fourier who died before he could
examine this paper. The manuscript has never been found. In 1830, Galois
published a summary of his results. The rst theorem in this account is:
An equation of prime degree is solvable by radicals if and only if if two of
its roots are known then the others are rational functions of them. This
implies that a general equation of degree ve cannot be solved by means of
radicals. The most decisive result is the solvability criterion: A polynomial
is solvable by radicals if and only if its Galois group is a solvable.
4
We will also study the solutions of several problems in Greek Geometry
using rudiments of eld theory. In Euclidean Geometry, we carry out sev-
eral geometric constructions with a ruler (unmarked) and compass such as
bisection of line segments and angles, constructions of certain angles, tri-
angles, quadrilaterals and circles. Ancient Greeks posed the following four
problems:
(1) The Delian Problem : Construct the side of a cube of volume 2.
(2) The angle trisection problem : Divide a given angle in three
equal parts.
(3) Squaring a circle : Construct a square having same area as that
of a given circle.
(4) Constructible regular polygons : Find n for which regular poly-
gon of n sides can be constructed by ruler and compass and describe
their constructions.
The above problems remained open for almost 2200 years. The nal solu-
tion employed techniques from abstract algebra and analysis. We will show
that it is impossible to construct side of a cube whose volume is 2 by ruler
and compass. The word Delian is derived from Delos which was a city in in
ancient Greece. It is said that almost a quarter of population of Delos died
of plague in 428 B.C. A delegation was sent to the oracle of Apollo at Delos
to enquire how the plague could be arrested. The oracle replied that the
cubical altar to the Sun God Apollo should be doubled. Instead of doubling
the volume the faithfuls doubled the sides of the cube thereby increasing the
volume eightfold.The second and the third problems also circulated among
Greek geometers around the same time. It is not known who solved the
Delian problem rst. The angle trisection problem was solved by Gauss as a
special case of his remarkable solution of the fourth problem. Gauss, barely
19, provided a construction of the 17-sided regular polygon. He also charac-
terized n for which regular ngons are constructible by ruler and compass.
Recall that a prime of the form 2
2
m
+ 1 is called a Fermat prime. Gauss
proved that a regular ngon is constructible if and only if n = 2
r
p
1
p
2
. . . p
g
where n 0 and p
1
, p
2
, . . . , p
g
are distinct Fermat primes. Gausss Theorem
solves the angle trisection problem. If 20
o
was constructible, then we can
construct a regular 18-gon. Since 3
2
|18, we have a contradiction.
5
The values of n for which regular ngons were known to be constructible
upto the time of Gauss were n = 2
m
, 2
m
.3, 2
m
.5 and 2
m
.15. No one was able
to construct a heptagon or a regular 17gon.
In March, 1796 Gauss made his rst mathematical discovery : construc-
tion of a 17-sided regular polygon by ruler and compass. He began noting
down his mathematical discoveries in a diary which he maintained for the
next 19 years. Gauss published Disquisitiones Arithmaticae in 1801
which has become a classic in mathematical literature. The last result of
this is his solution to the fourth problem. Gauss was very proud of this
discovery. He desired that a regular polygon of 17 sides be engraved on his
tombstone. This wish was not fullled. It was fullled when a monument
to Gauss was built in his birth place Braunschwig. Explicit construction of
17sided regular polygon was given by Erchinger in 1800. In 1892 Richelot
and Schwendenwein constructed a regular 257gon. Around 1900 Hermes
constructed a regular 65537-gon. The manuscript lls a box and it is found
in Gottingen. The construction has now been computerized. See an article
by Bishop in American Math. Monthly (1978).
Lindemann proved in 1882 that is not a root of any polynomial with
rational coecients. This proved the impossibility of squaring a circle.
6
Lecture 2 : Algebraic Extensions I
Objectives
(1) Main examples of elds to be studied.
(2) The minimal polynomial of an algebraic element.
(3) Simple eld extensions and their degree.
Key words and phrases: Number eld, function eld, algebraic element,
transcendental element, irreducible polynomial of an algebraic element, al-
gebraic extension.
2. Algebraic Extensions
The main examples of elds that we consider are :
(1) Number elds: A number eld F is a subeld of C. Any such eld
contains the eld Q of rational numbers.
(2) Finite elds : If K is a nite eld, we consider : Z K, (1) = 1.
Since K is nite, ker = 0, hence it is a prime ideal of Z, say generated by
a prime number p. Hence Z/pZ := F
p
is isomorphic to a subeld of K. The
nite eld F
p
is called the prime eld of K.
(3) Function elds: Let x be an indeteminate and C(x) be the eld of
rational functions, i.e. it consists of p(x)/q(x) where p(x), q(x) are poly-
nomials and q(x) = 0. Let f(x, y) C[x, y] be an irreducible polynomial.
Suppose f(x, y) is not a polynomial in x alone and write
f(x, y) = y
n
+ a
1
(x)y
n1
+ + a
n
(x), a
i
(x) C[x].
By Gauss lemma f(x, y) C(x)[y] is an irreducible polynomial. Thus
(f(x, y)) is a maximal ideal of C(x)[y]/(f(x, y)) is a eld. K is called the
function eld of the curve dened by f(x, y) = 0 in C
2
.
Characteristic of a eld : Let R be a commutative ring with identity e.
Dene the ring homomorphism f : Z R by f(n) = ne. Then ker f = (n)
for some integer n. If n = 0, then Z is isomorphic to a subring of R. In this
case we say that R has characteristic zero. If R is a domain then Z/(n)
7
is a domain as it is isomorphic to a subring of R. Hence n is a prime number,
say p. Therefore the nite eld F
p
is isomorphic to a subeld of R. In this
case, we say that R has characteristic p. Thus any eld F contains either
an isomorphic copy of Q or F
p
.
Denition 2.1. (i) Let K be a subeld of a eld F. We say F is an
extension eld of K. We also say that K is a base eld. We also write
this as F/K.
(ii) An element a F is called algebraic over K if there exists a nonzero
polynomial f(x) K[x] such that f(a) = 0. If every element of F is algebraic
over K then we say that F is an algebraic extension of K.
(iii) An element a F which is not algebraic over K is called a transcen-
dental element over K.
Example 2.2. It is known that the base e of the natural logarithm and
are transcendental over Q. Since (i)
2
=
2
, i is a root of x
2

2
R[x].
Hence i is algebraic over R. However i is not algebraic over Q. Thus the
property of being algebraic depends upon the base eld.
Example 2.3. Let K be a nite eld whose characteristic is a prime number
p. Then K has a subeld F with p elements. Since K is nite, it is a nite
dimensional F-vector space. If dim
F
K = n then K has p
n
elements. If a
K then the set {1, a, a
2
, . . . , a
n
} is linearly dependent. Let b
0
, b
1
, . . . , b
n
F,
not all zero, so that b
0
+b
1
a+ +b
n
a
n
= 0. Hence a is a root of the nonzero
polynomial b
0
+b
1
x + +b
n
x
n
. Therefore b is algebraic over F and hence
K/F is an algebraic extension.
Proposition 2.4. Let F/K be a eld extension and F be algebraic over
K. Then there exists a unique monic irreducible polynomial f(x) K[x]
such that f() = 0.
Proof. Dene : K[x] F by (g(x)) = g(). Since is a ring homomor-
phism and is algebraic, ker = I is a nonzero ideal of K[x]. Since K[x]
is a PID and K[x]/I is isomorphic to a subeld of F, I is generated by an
irreducible polynomial h(x) K[x].. If g() = 0 then g(x) = h(x)h
1
(x) for
some polynomial h
1
(x) K[x]. If g is irreducible, then g = h(x) for some
K

= K \ {0}. If g and h are taken to be monic, then g = h.

8
Denition 2.5. The irreducible monic polynomial in F[x] whose root is
K is denoted by irr(, F) and it is called the irreducible monic poly-
nomial of over F. The degree of irr (, F) is called the degree of
and it is written as deg
F
.
Example 2.6. (i)

i C satises f(x) = x
4
+ 1 = 0. Show that f(x) =
irr(

i, Q). Consider the eld Q(i) = smallest eld containing Q and i. Then
irr (

i, Q(i)) = x
2
i.
(ii) Let p be a prime number and
p
= e
2i/p
. Then x
p
1 = 0 is satised
by
p
. Since x
p
1 = (x 1)(x
p1
+ x
p2
+ + x + 1) and
p
(x) :=
x
p1
+ x
p2
+ + x + 1 is irreducible over Q, irr(
p
, Q) =
p
(x).
Simple eld extensions: Let K F be a eld extension. Let , F
be transcendental. Dene : K[x] F such that (g(x)) = g(). Then
ker = {0}. Thus K[x] K[] and hence K() K() by an isomorphism
such that () = and |
K
= id
K
. The situation is quite dierent for
algebraic elements.
Proposition 2.7. Let F K be a eld extension and K be algebraic
over F and f(x) = irr (, F). Let n = deg f. Then
(i) F[] = F() F[x]/(f(x)). (ii) dim
F
F() = n and {1, , . . . ,
n1
} is
an F- basis of F().
Proof. Consider the substitution homomorphism
: F[x] F[] such that (x) = , |
F
= id
F
Then ker = (f(x)) where f(x) = irr(, F). Hence F[x]/(f(x)) F[].
since (f(x)) is a maximal ideal, F[] is a eld, so F[] = F().
(ii) Let g() F[] and g(x) = f(x)q(x) + r(x) where q, r F[x], and
deg r(x) < degf(x) or r(x) = 0. Then g() = r(). Thus F[] is an
Fvector space generated by 1, , . . . ,
n1
where n = deg f(x). Suppose
that

n1
i=0
a
i

i
= 0. If a
i
are not all zero then

n1
i=0
a
i
x
i
is a nonzero poly-
nomial of degree less than deg f(x) satised by . This contradicts mini-
mality of deg f(x). Thus {1, ,
2
, . . . ,
n1
} is an F- vector space basis of
F[]. Hence dim
F
F[] = deg irr (, F).

9
Proposition 2.8. Let K/F be a eld extension and K be algebraic
over F. Then F()/F is an algebraic extension.
Proof. If F() and = 0 then {1, ,
2
, . . . ,
n
} is a linearly dependent
subset of F() since dim
F
F() = n. Hence there exist a
0
, a
1
, . . . , a
n
F
not all zero so that a
0
+a
1
+ +a
n

n
= 0. Hence is algebraic. Therefore
F()/F is an algebraic extension.
Proposition 2.9. Let , K F be algebraic over F. Then there exists
an F-isomorphism : F() F() such that () = if and only if
irr (, F) = irr (, F).
Proof. Let f(x) = irr (, F) and g(x) = irr (, F). Then (f()) = f() =
0. Thus g(x)|f(x). Since g, f are monic and irreducible, g(x) = f(x).
Conversely, suppose irr (, F) = irr (, F). Then F() F[x]/(f(x))
F() and the isomorphisms are F-isomorphisms. Hence F() and F() are
F-isomorphic.
Proposition 2.10. Let F K, K

be two eld extensions of F. Let :

K K

be an F- isomorphism. Let K be a root of f(x) F[x]. Then

() is a root of f(x).
Proof. (f()) = f(()) = 0
Example 2.11. (i) Let f(x) = x
3
2 Q[x]. By Eisensteins criterion
f(x) is irreducible over Q. The roots of f(x) are , w, w
2
where is the
real cube root of 2 and w is the complex cube root of 1. Thus the elds
Q(), Q(w), Q(w
2
) are Qisomorphic.
(ii) Since irr (i, R) = x
2
+ 1, R[x]/(x
2
+ 1) = R(i) = C.
(iii) The polynomial f(x) = x
2
+ x + 1 is irreducible over F
2
. Hence K =
F
2
[x]/(f(x)) is a eld which is a two dimensional F
2
vector space. Hence
K is a eld with four elements.
(iv) The polynomial g(x, y) = y
3
x(x + 1)(x 1) is irreducible in C(x)[y]
by Eisensteins criterion. Hence C(x)[y]/(g(x, y)) is a simple eld extension
of the function eld C(x).
10
Lecture 3 : Algebraic Extensions II
Objectives
(1) Degree of a eld extension and its multiplicative nature.
(2) A eld extension of nite degree is algebraic.
(3) Transitivity of algebraic extensions.
(4) Compositum of two elds.
Key words and phrases: Simple eld extension, degree of a eld exten-
sion, compositum of elds.
3. Degree of a field Extension
Denition 3.1. Let F K be a eld extension. The dimension of the
F-vector space K, denoted by [K : F] is called the degree of the eld
extension K/F.
For an algebraic element K, dim
F
F() = deg irr(, F). If [K : F] < ,
then F K is called a nite extension.
Proposition 3.2. A nite extension K/F is an algebraic extension.
Proof. Let [K : F] = n and K. Then 1, , . . . ,
n
are linearly dependent
over F. Hence there exist a
0
, a
1
, . . . , a
n
, not all zero in F such that a
0
+a
1
+
+a
n

n
= 0. Let f(x) = a
0
+a
1
x + +a
n
x
n
. Then is a root of f(x).
Hence is algebraic over F.

Corollary 3.3. Every irreducible polynomial over R has degree 2.

Proof. Let f(x) R[x] be irreducible and C a root of f(x). Then
R[] C. If R, deg f(x) = 1. If / R, then [R[] : R] 2. Thus
C = R[]. Since [C : R] = 2, deg f(x) = 2.

11
Example 3.4. (1) Since irr (i, R) = x
2
+ 1, [C : R] = 2 as C R(i).
(2) Since irr (
p
, Q) = x
p1
+ x
p2
+ + x + 1, [Q(
p
) : Q] = p 1.
(3) Algebraic extension of a eld may not be nite. Consider the chain of
elds Q Q(2
1/2
) Q(2
1/2
n
) . Their union K contains the
algebraic numbers
n
= 2
1/2
n
for all n and
n
is a root of the irreducible
polynomial f
n
(x) = x
2
n
2. Hence [K : Q] 2
n
for all n. Thus [K : Q] = .
(4) Quadratic Extensions: If [K : F] = 2 then K is called a quadratic
extension of F. Let K\F then {1, } is a basis of K over F. Hence
2
=
a+b for some a, b F. Therefore f(x) = irr(, F) = x
2
axb. The roots
of f(x) are (a

a
2
+ 4ab)/2 if char F = 2. Therefore K = F(

a
2
+ 4b).
Denition 3.5. A chain of elds F
1
F
2
F
n
is called a tower of
elds if F
i
is a subeld of F
i+1
, for all i = 1, 2, . . . , n 1.
Proposition 3.6. If K F L is a tower of elds then
[L : F][F : K] = [L : K].
Proof. If either F/K or L/F are innite dimensional, then L/K is also
innite dimensional. Thus we may assume that F/K and L/F are nite.
Suppose that [F : K] = m and [L : F] = n. Let x
1
, x
2
, . . . , x
n
be a basis of
L over F and y
1
, y
2
, . . . , y
m
be a basis of F over K.
We claim that the set
B = {x
j
y
j
| i = 1, 2, . . . n, and j = 1, 2, . . . , m}
is a vector space basis of L over K. Let z L. Thus z = f
1
x
1
+ +f
n
x
n
,
for some f
1
, . . . , f
n
F. We write f
i
=

m
j=1
k
ij
y
j
. Therefore
z =
n

l=1
x
l
f
l
=
n

l=1
m

j=1
x
l
k
lj
y
j
.
Thus B generates L as a K- vector space. Suppose

m
j=1

n
i=1
a
ij
x
i
y
j
= 0.
Then
n

i=1

j=1
a
ij
y
j

x
i
= 0.
Since x
1
. . . , x
n
are F-linearly independent. Therefore

n
j=1
a
ij
y
j
= 0 for
each i. By linear independence of y
1
, . . . , y
n
to see that all the a
ij
= 0.
12
Corollary 3.7. Let F K be a nite eld extension. Then deg irr(, F)
divides [K : F], for all K.
Proof. Since F F() K, we have
[K : F] = [K : F()][F() : F].
Thus deg irr(, F) divides [K : F].
Proposition 3.8. Let K/F be a eld extension. If a
1
, a
2
, . . . , a
n
K are
algebraic over F then F(a
1
, a
2
, . . . , a
n
) is a nite algebraic extension of F.
Proof. Since a
i
is algebraic over F, it is algebraic over F(a
1
, a
2
, . . . , a
i1
).
Thus [F(a
1
, a
2
, . . . , a
i
) : F(a
1
, a
2
, . . . , a
i1
)] is nite for all i. Therefore the
eld F(a
1
, a
2
, . . . , a
n
) is a nite extension of F. Hence it is algebraic.
Corollary 3.9. Let E/F and K/E be algebraic extensions. Then K/F is
an algebraic extension.
Proof. Let a K and let a be a root of f(x) = a
0
+a
1
x+ +a
n1
x
n1
+x
n

E[x]. Consider the eld L = F(a

0
, a
1
, . . . , a
n1
). Then a is algebraic over
L. Hence L(a) is a nite extension of L. Since a
0
, a
1
, . . . , a
n1
are algebraic
over F, L is a nite extension of F. Hence L(a) is a nite extension of F.
Hence a is algebraic over F.
Corollary 3.10. Let K/F be a eld extension. Then the set of elements of
K which are algebraic over F is a subeld of K.
Proof. Let a, b K be algebraic over F. Then F(a, b) is a nite extension of
F. Hence all elements of F(a, b) are algebraic over F. In particular, a b, ab
and a/b if b = 0, are all algebraic over F.
Compositum of elds: Let L/k be a eld extensions and E/k and F/k be
intermediate eld extensions. Then the smallest eld containing E and F,
to be denoted by EF, is called the compositum of F and F. Suppose E =
k(a
1
, a
2
, . . . , a
n
) and F is an extension of k. Then EF = F(a
1
, a
2
, . . . , a
n
).
Example 3.11. Let m and n be co prime positive integers. Consider the
subelds E = Q(
m
) and F = Q(
n
) of C. Then the compositum of E and
13
F is Q(
mn
). Indeed, as m and n are coprime, there exist p, q N such that
mp + nq = 1. Therefore

mn
= exp(2i/mn) = exp(2pi/n) exp(2qi/m) = (
n
)
p
(
m
)
q
.
We can estimate the degree of the compositum of two nite eld exten-
sions in terms of their degrees.
Proposition 3.12. Let L/k be a eld extension and E/k, F/k be interme-
diate nite extensions elds. Then
[EF : k] [E : k][F : k].
If [E : k] and [F : k] are coprime then equality holds.
Proof. Let x
1
, x
2
, . . . , x
m
and y
1
, y
2
, . . . , y
n
be bases of the k-vector spaces
E and F respectively. Then it is easy to see that E = k(x
1
, x
2
, . . . , x
m
)
and F = k(y
1
, y
2
, . . . y
n
). Therefore EF = k(x
1
, x
2
, . . . , x
m
; y
1
, y
2
, . . . y
n
).
We have the following diagram of eld extensions:
L
EF
{
{
{
{
{
{
{
{
C
C
C
C
C
C
C
C
E
C
C
C
C
C
C
C
C
F
{
{
{
{
{
{
{
{
k
Since EF = E(y
1
, y
2
, . . . , y
n
) we have [EF : F] n. Since the degree is
multiplicative in a tower of nite extensions, we have
[EF : k] = [EF : E][E : k] mn.
Since m and n both divide [EF : k], and (m, n) = 1, we get mn | [EF : k].
Hence [EF : k] = mn.
16
Lecture 4 : Ruler and Compass Constructions I
Objectives
(1) Describe standard ruler and compass constructions.
(2) The eld of constructible numbers is closed under taking square roots
of positive reals.
(3) Characterization of constructible real numbers via square root towers
of elds.
(4) The degree of a constructible real number is a power of 2.
(5) Impossibility of squaring the circle, trisection of angles and duplica-
tion of cubes by ruler and compass.
Key words and phrases: Ruler and compass constructions, constructible
real numbers, square root tower, trisection of an angle, duplication of cube,
squaring a circle.
4. Four Problems of Classical Greek Geometry
The four problems of classical Greek Geometry: duplication of cube, tri-
section of angles, squaring of circles and construction of regular polygons
can now be solved using the rudiments of algebraic extensions of elds. A
complete solution of the last problem about characterization of constructible
regular polygons will use fundamental theorem of Galois theory. This will
be discussed later. As we have remarked before, these problems remained
open for more than 2000 years. We will see that the language of eld exten-
sions provides the right framework for discussion of these problems. Once
translated into this language, the solutions are obtained quickly.
First we will precisely formulate constructibility by ruler and compass and
the concept of constructible points, lines and constructible real numbers.
A real number is called constructible if it is the length of a line segment
connecting two constructible points. We will then see that a constructible
real number is algebraic over Q and its degree over Q is a power of two. This
criterion leads to solutions of the rst three problems and a partial solution
of the fourth problem.
17
Constructible points, lines, circles and real numbers:
Given a nite set {P
1
, . . . , P
n
} of points in the Cartesian plane R
2
, dene
the set S
m
inductively. Put S
0
= {P
1
, . . . , P
n
}. Suppose S
m
has been de-
ned. Put S
m+1
= S
m
T
m
where T
m
is the set of points of intersection of
lines passing through points in S
m
and circles with center at one point in S
m
with radii equal to distance between points of S
m
. Let S =

m=0
S
m
. We say
that S = C(P
1
, . . . , P
n
) is the set of points constructible from P
1
, P
2
, . . . , P
n
by ruler and compass. A real number a is called constructible if |a| is the
distance between two constructible points. A line passing through two con-
structible points is called a constructible line. A circle is called constructible
if its center is constructible and its radius is a constructible real number.
We can reformulate the problems according to the denition of con-
structible points. Let P
1
= (0, 0) and P
2
= (0, 1). Is (
3

2, 0) C(P
1
, P
2
) ?
This is the Delian problem. For the squaring of the circle problem, if there
exists a square with side a such that a
2
= then a =

. So the problem
is asking whether (

, 0) C(P
1
, P
2
). For the angle trisection problem,
set P
3
= (cos , sin ). The problem asks whether C(P
1
, P
2
, P
3
) contains
(cos /3, sin /3). The problem of construction of regular n-gons asks for
which values of n, (cos
2
n
, sin
2
n
) C(P
1
, P
2
).
Trisection of an angle with a marked ruler and compass
Figure 1. Trisection of an angle with a marked ruler and compass
Let AOB = . Draw a unit circle centered at O. Suppose one end of a
ruler is E and point P is marked on the ruler such that EP = 1. Slide the
18
ruler in such a way that E is on X-axis and P is on the circle and the edge
passes through B. Then DCO gives + + = . Hence = 2.
The BOC gives 4 + ( +) = . Hence = /3
Duplication of a cube with a marked ruler and compass
Figure 2. Duplication of a cube with a marked ruler
Use a ruler with one end point marked as E and a point marked as P with
EP = 1. Let AB be a segment of unit length. Draw the angles BAD = 90
o
and BAE = 120
o
. We show that PB =
3

2. Let x be the length of PB

and z be the length of AP. Then x
2
= z
2
+ 1. Since QEB || APB, we
get
x + 1
a + 1
=
x
1
and
a

3
z
=
a + 1
1
.
Hence a = 1/x and

3/xz = x + 1/x. We also have

3x = x
2
z + xz and
hence z =

3x/x(x+1) =

3/x + 1. Since x
2
= 3/(x + 1)
2
+1, x
4
+2x
3
+
x
2
= 3 + x
2
+ 2x + 1. Therefore x
4
+ 2x
3
2x 4 = (x + 2)(x
3
2) = 0.
Therefore x =
3

2
Standard Constructions
(i) Bisecting a line segment: Suppose A and B are constructible points,
we show that the mid point of the line segment AB is also constructible
19
Figure 3. Bisection of a line segment
Figure 4. Bisection of an angle
Draw circles with centers A, B with radius AB. Then the intersection points
of these circles C, D are constructible. The mid point of AB is the inter-
section of CD and AB. It is the mid point since it is the intersection of
diagonals of the rhombus ACBD.
(ii) Bisection of an angle: Let A, O, B be three constructible points. They
determine the angle AOB.
20
Draw a circle with center O and radius OB. It meets OA at D. Then D is
constructible. Now bisect the segment BD at E. So E is also constructible.
Then line OE bisects AOB.
(iii) Drawing a right angle:
Figure 5. Drawing a right angle
Suppose O, A are constructible points. We wish to draw a perpendicular
at O which is also a constructible line. Draw a circle C(O, OA). It meets
the extended line OA at Q. Draw circles C(Q, QA) and C(A, QA). These
intersect at B and C. The triangle QAB is isosceles. Hence BOA = 90
o
(iv) Dropping a perpendicular:
Suppose L is a constructible line and P a point outside this line which is
constructible. Then we can draw a perpendicular onto L from P which is
also constructible. Draw the circle C(P, r) where r is a large constructible
number so that C(P, r) meets 2 points Q and R. Draw circles at centers Q
and R of radius PQ. Join PS and take the intersection of QR and PS.
(v) Drawing a parallel line.
Suppose L is a constructible line and P is a constructible point outside L.
Drop a perpendicular PO on L. Now draw 90
o
on OP at P to get a parallel
line.
Algebraic properties of Constructible Real Numbers
21
Figure 6. Construction of a perpendicular onto a line
Proposition 4.1. A point P = (a, b) is constructible if and only if a and b
are constructible real numbers.
Proof. Drop a perpendicular from P to X and Y axes to get constructible
points A and B. So a and b are constructible real numbers. If a and b are
constructible then we can draw circles C(0, a) and C(0, b) to get A and B.
Now draw perpendicular at A and B to get P.
Proposition 4.2. Constructible real numbers form a subeld of R.
Proof. It is easy to show that a b are constructible if a and b are so. To
show ab and a/b for b = 0 are constructible, use the constructions in the
gures below.

Proposition 4.3. If a is a positive constructible real number then so is

a
Proof. Since A = (a, 0) is constructible so is B = (a + 1, 0). Hence the
mid point C = (a + 1/2, 0) is constructible. Draw the circle C with center
(0,
a+1
2
) and radius (a + 1)/2. Draw a perpendicular at A which meets the
circle at D. Since ODA and DBA are similar, x/a = 1/x we have
x =

22
Figure 7. Construction of

a
Figure 8. Construction of ab
Corollary 4.4. Let F C be a subeld of the eld C of constructible real
numbers. Let k > 0 F. Then F(

k) C.
Proof. We need to show each number of F(

k) is constructible. Since an
arbitrary element of F(

k) is of the form a + b

k where a, b C, it is
constructible since

k is constructible.
Theorem 4.5. Let Q F
1
F
2
F
n
be a sequence of elds such
that
23
Figure 9. Construction of a/b
F
j+1
= F
j
(
_
b
j
) and 0 < b
j
F
j
for j = 0, 1, . . . , n 1.
Then all elements of F
n
are constructible.
Proof. Apply induction on n.

Denition 4.6. A tower of elds as in Theorem 4.5 is called a square root

tower over Q.
Denition 4.7. Let F be a eld. Then F
2
is called the plane of F. Let
a, b, c F then the set {(x, y) | ax +by +c = 0} is called a line in F
2
and
the set {(x, y) | x
2
+y
2
+ax +by +c = 0} is called a circle in F
2
.
The proof of the next lemma is left as an exercise.
Lemma 4.8. Let F be a subeld of R. (i) The point of intersection, if any,
of two lines in F
2
belongs to F
2
.
(ii) The points of intersection of a line and a circle or two circles in F
2
lies
in F
2
or F(

k)
2
where 0 < k F.
Theorem 4.9. A real number a is constructible if and only if there exists
a square root tower Q F
1
F
N
such that a F
N
.
Proof. We have already proved that numbers in F
N
are constructible. If a
is constructible then P = (a, 0) is a constructible point. We wish to show
24
(a, 0) F
2
N
where F
N
is the last eld in a square root over Q. Beginning
with O = (0, 0) and I = (1, 0), the point P is constructed in nite number
of steps
S
0
= {O, I} S
1
S
m

Let P S
m
. Apply induction on m. If m = 0 then we are done. Let m > 0.
By induction S
m1
F
2
N
, where F
N
is the last eld in a square root tower
over Q. The points in S
m
are intersections of lines and circles in F
2
N
. Hence
they are in F
N
(

k)
2
for more 0 < k F
N
. Therefore P is in the plane of
F
N
.
Theorem 4.10. Suppose a is a constructible real number, then
[Q(a) : Q] = 2
m
for some m N.
Proof. Let Q F
1
F
N
be a square root tower over Q and a F
N
.
Then
[Q(a) : Q][F
N
: Q(a)] = [F
N
: Q] = 2
N
Hence [Q(a) : Q] = 2
m
for some m.
Corollary 4.11. It is impossible to duplicate a cube with ruler and compass.
Proof. The number =
3

2 is a root of the irreducible polynomial x

3
2
over Q. Hence [Q() : Q] = 3 which is not a power of two. Therefore is
not constructible.
Corollary 4.12. It is impossible to trisect an arbitrary angle with ruler
and compass.
Proof. Suppose the angle is given. We may assume P = (cos , sin )
is given along with O = (0, 0) and I = (1, 0). We wish to show that
(cos /3, sin /3) is not constructible. If so, then cos /3 and sin /3
are constructible real numbers. Using the identity cos 3 = 4 cos
3
3 cos
we get cos /3 = 4 cos
3
/9 3 cos /9. Therefore u = cos /9 satises
8u
3
6u 1 = 0. Hence w
3
3w1 = 0 where w = 2u. As [Q(w) : Q] = 3,
u is not constructible.
25
Corollary 4.13. It is impossible to square the unit circle by ruler and com-
pass.
Proof. Suppose it is possible to construct a segment a by ruler and compass
such that a
2
= . Then a =

is algebraic over Q, hence so is . But
is transcendental over Q. Therefore

is not constructible by ruler and
compass.
26
Lecture 5 : Ruler and Compass Constructions II
Objectives
(1) Wantzels characterization of constructible regular p-polygons.
(2) Richmonds construction of a regular pentagon.
(3) Gauss criterion of constructible regular polygons.
Key words and phrases: Fermats primes, constructible regular polygons,
Gauss criterion.
5. Constructible regular polygons
In this section we discuss constructibility by ruler and compass of regular
polygons. Gauss proved that a regular polygon of n sides is constructible
by ruler and compass if and only if n = 2
m
p
1
p
2
. . . p
r
where m N and
p
1
, p
2
, . . . , p
r
are distinct Fermat primes. The number F
m
= 2
2
m
+ 1 is
called a Fermat prime whenever it is a prime. The known Fermat primes
are:
F
0
= 3, F
1
= 5, F
2
= 17, F
3
= 257 and F
4
= 65537.
Fermat showed that F
m
is a prime for m 4. Eisenstein conjectured that
there are innitely many Fermat primes. This conjecture is still open. Euler
showed that F
5
is divisible by 641.
Proposition 5.1 (1837, Wantzel). Let a regular polygon of n sides be con-
structible and p be an odd prime dividing n. Then p is a Fermat prime.
Proof. If p|n and a regular n-gon is constructible then a regular p-gon is
also constructible. Thus the point (cos 2/p, sin 2/p) is a constructible
point. Hence there exists a eld F Q such that [F : Q] = 2
m
and
cos 2/p, sin 2/p F. Then
p
= cos 2/p + i sin 2/p F(i) and hence
[Q(
p
) : Q] = 2
s
= p1 and therefore p = 1+2
s
. It follows that s is a power
of 2. Hence p is a Fermat prime.
27
Figure 10. Richmonds construction of a regular pentagon
Construction of a pentagon by ruler and compass
We show that a pentagon is constructible by ruler and compass. The vertex
(cos 72
0
, sin 72
0
) of a regular pentagon corresponds to the complex number
z = e
2i/5
which is the root of the irreducible polynomial

5
(x) = x
4
+x
3
+x
2
+x + 1.
Therefore
z
2
+z + 1 +
1
z
+
1
z
2
= 0.
Completing the square we get
_
z +
1
z
_
2
+
_
z +
1
z
_
1 = 0.
Put z + 1/z = y to get y
2
+ y 1 = 0 and z
2
zy + 1 = 0. Hence
y = (1

5)/2. The second equation gives z = (y

_
y
2
4)/2. Clearly
y is constructible as it belongs to Q(

5) and we have a square root tower

Q Q(

5) Q(

5,
_
y
2
4).
Thus all roots of
5
(x) are constructible real numbers.
Richmonds construction of a regular pentagon (1893)
28
Draw a unit circle with center O. Draw a perpendicular OR at O. Let Q
be the mid point of OR. Join Q and P and then bisect PQO. Let the
bisector meet OP at S. Construct a perpendicular at S and let T be its
intersection point with the circle. We show TOP = 72
o
. It is enough to
show that OS = cos 72
o
. Note that
OQS =
90
2
tan
_
45
o

2
_
=
OS
1
2
OS =
1
2
tan(45
o

2
).
Using tan =
1
2
=
2 tan /2
1tan
2
/2
we get tan

2
=

5 2. Therefore
OS =
1
2
_
tan 45
o
tan /2
1tan 45
o
tan(/2)
_
=

51
4
= cos 72
o
.
Proposition 5.2. A heptagon is not constructible by ruler and compass.
Proof. Let = 2/7 and
7
= cos + i sin . Then
7
is a root of the irre-
ducible polynomial

7
(x) = x
6
+x
5
+x
4
+x
3
+x
2
+x + 1.
Therefore [Q(
7
) : Q] = 6. Using
7
+
7
= 2 cos we get [Q(
7
) : Q(cos )] =
2 and hence [Q(cos ) : Q] = 3. Thus cos is not constructible. Therefore a
heptagon is not constructible by ruler and compass.
Proposition 5.3. Let p be a prime number and
p
= cos 2/p
2
+i sin 2/p
2
.
Then
[Q(
p
) : Q] = p(p 1).
Proof. Since
p
satises the equation
x
p
2
1 = (x
p
1)(x
p(p1)
+x
p(p2)
+ + (x
p
)
2
+x
p
+ 1) = 0
and
p
p
= 1,
p
is root of f(x) = (x
p
)
p1
+ (x
p
)
p2
+ + (x
p
)
2
+ x
p
+ 1.
We show that f(x) Q[x] is irreducible. Put x = u+1 and use Eisensteins
criterion:
29
f(u + 1) =
p

k=1
(u + 1)
p(pk)
=
p

k=1
(u
p
+ 1 +pg(u))
pk
=
p

k=1
(u
p
+ 1)
pk
+ph
k
(u)
where h
k
(u) Z[u] has degree p
2
pk 1. Since

p
k=1
(u
2
+ 1)
pk
=
(u
p
+1)
p
1
u
p
= u
p(p1)
+pH(u),
f(u + 1) =
p

k=1
(u + 1)
p(pk)
= u
p(p1)
+pG(u).
Since f(1) = p, the constant term of f(u + 1) is divisible by p and not by
p
2
. By Eisensteins criterion f(u + 1) and hence f(x) is irreducible. Thus
[Q(
p
) : Q] = p
2
p.
Theorem 5.4 (Gauss). If a regular polygon of n sides is constructible then
n = 2
r
p
1
p
2
. . . p
s
where p
1
, . . . , p
s
are distinct Fermat primes.
Proof. If p
2
n then p-gon is constructible. Hence [Q(cos 2/p
2
) : Q] = 2
u
for
some positive integer u. Thus p(p 1) = 2
u
, which is a contradiction.
This nishes the proof of one half of Gausss constructibility criterion for reg-
ular polygons. We shall prove the other half after we prove the fundamental
theorem of Galois Theory.
34
Lecture 6 : Symmetric Polynomials I
Objectives
(1) Examples of symmetric polynomials.
(2) The fundamental theorem of symmetric polynomials.
(3) Newtons identities for power sum symmetric polynomials.
Key words and phrases: Symmetric polynomial, symmetrization of a
monomial, power sum symmetric polynomials, Newtons identities.
6. Symmetric Polynomials
Our next goal is to prove the Fundamental Theorem of Algebra : Ev-
ery polynomial of positive degree with complex coecients has a complex
root. You must have seen its topological and complex analytic proofs. We
will present a proof which uses symmetric polynomials and the construc-
tion of the splitting eld of a polynomial. We will learn about symmetric
polynomials in this section and splitting elds of polynomials in the next
section.
Let R be a commutative ring with identity and S = R[u
1
, u
2
, . . . , u
n
] be the
polynomial ring in n variables over R. Let S
n
, the symmetric group
of all permutations of {1, 2, . . . , n}. A permutation S
n
gives rise to an
automorphism g

: S S, dened as
g

(f(u
1
, . . . , u
n
)) = f(u
(1)
, . . . , u
(n)
).
Denition 6.1. A polynomial f S is called a symmetric polynomial
if for all S
n
f(u
1
, . . . , u
n
) = f(u
(1)
, . . . , u
(n)
).
Example 6.2. (1) Consider the general polynomial
f(x) = (x u
1
)(x u
2
) . . . (x u
n
)
= x
n

1
x
n1
+
2
x
n2
+ + (1)
n

n
35
where

1
= u
1
+ +u
n
,
2
=

i<j
u
i
u
j
, . . . ,
n
= u
1
u
2
u
n
.
It is easy to verify that
1
, . . . ,
n
are symmetric. These are called the
elementary symmetric polynomials in u
1
, u
2
, . . . , u
n
.
(2) The symmetrization of a monomial u

1
1
. . . u
n
n
is dened as
S(u

1
1
u
n
n
) =

Sn
u

1
(1)
u

2
(2)
. . . u
n
(n)
.
It is clear that S(u

1
1
u
n
n
) is a symmetric polynomial. The symmetriza-
tion of u
2
1
u
2
is
S(u
2
1
u
2
) = u
2
1
u
2
+u
2
1
u
3
+u
2
2
u
3
+u
2
3
u
1
+u
2
3
u
2
+u
2
2
u
1
.
(3) For each k the polynomials w
k
= u
k
1
+ u
k
2
+ + u
k
n
are symmetric
polynomials.
(4) Let h
m
denote the sum of all monomials of degree m in u
1
, u
2
, . . . , u
n
.
It is called the complete homogeneous symmetric polynomial of degree m.
Fundamental Theorem for symmetric polynomials
Example 6.3. Consider the symmetric polynomial
f(u
1
, u
2
, u
3
) = u
2
1
u
2
+u
2
1
u
3
+u
2
2
u
1
+u
2
2
u
3
+u
2
3
u
1
+u
2
3
u
2
.
Then f(u
1
, u
2
, 0) = u
2
1
u
2
+u
2
2
u
1
= u
1
u
2
(u
1
+u
2
) =
0
1

0
2
, where

0
1
=
1
(u
1
, u
2
, 0) = u
1
+u
2
and
0
2
=
2
(u
1
, u
2
, 0) = u
1
u
2
.
Consider f
1

2
= g. Then g|
u
3
=0
= 0. Thus u
3
| g. Since g is symmetric
u
1
u
2
u
3
=
3
| g. This gives f
1

2
= 3u
1
u
2
u
3
= 3
3
and therefore
f =
1

2
3
3
.
Theorem 6.4 (Newton). Let R be a commutative ring. Then every sym-
metric polynomial in R[u
1
, u
2
, . . . , u
n
] is a polynomial in the elementary
symmetric polynomials in a unique way. In other words if f(u
1
, u
2
, . . . , u
n
)
is symmetric then there exists a unique polynomial g R[x
1
, . . . , x
n
] such
that
g(
1
,
2
, . . .
n
) = f(u
1
, u
2
, . . . , u
n
).
36
Proof. Apply induction on n. The n = 1 case is clear. Let the theorem be
true for symmetric polynomials in n 1 variables. To prove the theorem
in R[u
1
, u
2
, . . . , u
n
], apply induction on deg f. If deg f = 0 then f is a
constant. It is clear in this case. Consider f(u
1
, u
2
, . . . , u
n1
, 0) = f
0

R[u
1
, u
2
, . . . , u
n1
]. Then f
0
is symmetric. By induction hypothesis we have
f
0
= g(
0
1
,
0
2
, . . . ,
0
n1
). Then f g(
1
,
2
, . . . ,
n1
) = f
1
is symmetric
and f
1
(u
1
, . . . , u
n1
, 0) = 0. Thus u
n
| f
1
and hence
n
| f
1
, by symmetry.
So f
1
=
n
h(u
1
, . . . , u
n
). Since
n
is not a zerodivisor in R[u
1
, . . . , u
n
], h is
symmetric. Since deg h < deg f, by induction hypothesis h is a polynomial
in
1
, . . . ,
n
, hence f is so. Therefore f is a polynomial in
1
, . . . ,
n
.
Uniqueness : Use induction on n. the n = 1 case is obvious. Let us rst
prove that the map
: S = R[z
1
, z
2
, . . . , z
n
] R[
1
,
2
, . . . ,
n
] such that
(z
i
) =
i
, i = 1, 2, . . . , n and |
R
= id
R
is an isomorphism. If it is not an isomorphism, we pick a nonzero polynomial
f(z
1
, z
2
, . . . , z
n
) S of least degree such that
f(
1
,
2
, . . . ,
n
) = 0.
Write f as a polynomial in z
n
with coecients in R[z
1
, z
2
, . . . , z
n1
] :
f(z
1
, z
2
, . . . , z
n
) = f
0
(z
1
, z
2
, . . . , z
n1
) + +f
d
(z
1
, z
2
, . . . , z
n1
)z
d
n
.
Then f
0
= 0. If so, then f = z
n
g where g S. Then
n
g(
1
, . . . ,
n
) = 0.
Hence g(
1
, . . . ,
n
) = 0. This contradicts the minimality of deg f. Therefore
we have
0 = f
0
(
1
, . . . ,
n1
) + +f
d
(
1
, . . . ,
n1
).
d
n
.
In this relation put u
n
= 0 to get
f
0
((
1
)
0
, (
2
)
0
, . . . , (
n1
)
0
) = 0.
This is a nontrivial relation among the elementary symmetric polynomials
in u
1
, u
2
, . . . , u
n1
. This is a contradiction.
Newtons identities for power sum symmetric polynomials
By the Fundamental Theorem for symmetric polynomials the symmetric
polynomials w
k
= u
k
1
+ + u
k
n
, k = 1, 2, 3, . . . are polynomials in the
37
elementary symmetric polynomials. Isaac Newton found identities which
express w
k
in terms of
1
,
2
, . . . ,
n
.
Theorem 6.5 (Newton).
w
k
=
1
w
k1

2
w
k2
+ + (1)
k

k1
w
1
+ (1)
k+1

k
k if k n,
=
1
w
k1

2
w
k2
+ + (1)
n+1

n
w
kn
if k n.
Proof. Let z, y be indeterminate. Then
(y u
1
)(y u
2
) (y u
n
) = y
n

1
y
n1
+
2
y
n2
+ + (1)
n

n
Put y = 1/z to get
(1 u
1
z)(1 u
2
z) (1 u
n
z) = 1
1
z +
2
z
2
+ +(1)
n

n
z
n
:= (z)
Consider the generating function of w
1
, w
2
, . . .
w(z) = w
1
z +w
2
z
2
+w
3
z
3
+ =

k=1
w
k
z
k
=

k=1
n

i=1
u
k
i
z
k
=
n

i=1

k=1
(u
i
z)
k
=
n

i=1
u
i
z
1 u
i
z
Since (z) = (1 u
1
z) (1 u
n
z),

(z) =
n

i=1
u
i
(z)
1 u
i
z
and hence w(z) =
n

i=1
u
i
z
1 u
i
z
=
z

(z)
(z)
This implies that
w(z)(z) = z(
1
+
2
(2z)
3
(3z
2
) + + (1)
n
n
n
z
n1
)
=
1
z 2
2
z
2
+ 3
3
z
3
+ + (1)
n+1
n
n
z
n
if k n, equating the coecient of z
k
we get
(1)
k+1
k
k
= w
k

1
w
k1
+w
k
2
2
+ + (1)
k
w
1

k1
.
38
Hence
w
k
=
1
w
k1

2
w
k2
+ + (1)
k+1

k
k.
If k > n, equate coecient of z
k
to get
w
k
w
k1

2
+ (1)
n

n
w
kn
= 0.
Therefore
w
k
=
1
w
k1

2
w
k2
+ + (1)
n+1

n
w
kn
.

39
Lecture 7 : Symmetric Polynomials II
Objectives
(1) Discriminant in terms of power-sum symmetric polynomials.
(2) Discriminant of a cubic.
(3) Existence of a splitting eld of a polynomial.
(4) Fundamental theorem of algebra via symmetric polynomials.
Key words and phrases: Discriminant of a polynomial, splitting eld,
fundamental theorem of algebra.
Discriminant of a polynomial: We discuss a method to calculate the
discriminant of a polynomial by employing Newtons identities.
Denition 6.6. Let u
1
, u
2
, . . . , u
n
, x be indeterminate and
f(x) = (x u
1
)(x u
2
) . . . (x u
n
).
The discriminant of f(x) is the symmetric function
disc (f(x)) =
i<j
(u
i
u
j
)
2
It is clear that f(x) has a repeated root if and only if disc (f) = 0.
Since disc (f) is a symmetric polynomial with integer coecients, by the
fundamental theorem for symmetric polynomials, there exists a polynomial
g(X
1
, . . . , X
n
) Z[X
1
, X
2
, . . . , X
n
] such that disc (f) = g(
1
,
2
. . . ,
n
).
The van der Monde matrix
M =

1 1 1
u
1
u
2
u
n
u
2
1
u
2
2
u
2
n
.
.
.
.
.
.
.
.
.
.
.
.
u
n1
1
u
n1
2
u
n1
n

has determinant det M =

i>j
(u
i
u
j
). Hence
40
disc (f) = det(MM
t
) =

n w
1
w
2
w
n1
w
1
w
2
w
3
w
n
w
2
w
3
w
4
w
n+1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
w
n1
w
n
w
n+1
w
2n2

.
Example 6.7. Using Newtons identities, we calculate the discriminant of
the polynomial p(x) = x
3
+px +q. We have
1
= 0,
2
= p,
3
= q and
MM
t
=

3 w
1
w
2
w
1
w
2
w
3
w
2
w
3
w
4

Newtons identities in this case are

w
1
=
1
= 0
w
2
=
2
1
2
2
= 2p
w
3
=
1
w
2

2
w
1
+ 3
3
= 3q
w
4
=
1
w
3

2
w
2
+
3
w
1
= 2p
2
Therefore
disc (f) = det MM
t
=

3 0 2p
0 2p 3q
2p 3q 2p
2

= 4p
3
27q
2
.
7. The Splitting Field of a Polynomial
In this section we construct a eld extension K/F which contains all the
roots of a given polynomial f(x) F[x]. For simplicity, we want K to be
the smallest eld containing F with respect to this property.
Denition 7.1. Let F be a eld and f(x) F[x] be a monic polynomial
of degree n. A eld K F is called a splitting eld of f(x) over F
if there exist r
1
, r
2
, . . . , r
n
K so that f(x) = (x r
1
) . . . (x r
n
) and
K = F(r
1
, r
2
, . . . , r
n
).
Example 7.2. (i) Let f(x) = x
2
+ax +b F[x]. If f(x) is reducible then
F is a splitting eld of f(x). If f(x) is irreducible then (f(x)) is a maximal
41
ideal of F[x]. Hence F(x)/(f(x)) F(r) is a eld, where r = x +(f(x)). If
s is another root of f(x), then s + r = a, so s = a r F(r). Hence
F(r) is a splitting eld of f(x) over F.
(ii) Consider the irreducible polynomial f(x) = x
3
+ x + 1 F
2
[x]. Let
r = x + (f(x)) F
2
[x]/(f(x)) = F
2
(r). Since [F
2
(r) : F
2
] = 3, F
2
(r) has 8
elements. A basis of the F
2
vector space F
2
(r) is {1, r, r
2
}. Hence
F
2
(r) = {0, 1, r, r
2
, 1 +r, 1 +r
2
, r +r
2
, 1 +r +r
2
}
and we have the relation r
3
= 1+r. Check that f(r
2
) = f(r
4
) = 0. Therefore
x
3
+x + 1 = (x +r)(x +r
2
)(x +r
4
).
Thus F
r
(r) is a splitting eld of x
3
+x + 1 over F
2
.
We will later see that if f(x) F
q
[x], where F
q
is a nite eld with q
elements, then F
q
[x]/(f(x)) is a splitting eld of f(x), if f(x) is an irreducible
polynomial over F
q
.
Existence of Splitting eld
Theorem 7.3. Let F be a eld. Then any polynomial f(x) F[x] of positive
degree has a splitting eld.
Proof. Apply induction on deg f. If deg f = 1 then F is the splitting eld
of f over F. Suppose deg f > 1. If f(x) splits as a product of linear factors
in F[x] then F is the splitting eld of f(x) over F. Suppose g(x) is an
irreducible factor of f(x) with deg g 2. Then r = x + (g(x)) K :=
F[x]/(g(x)) is a root of g(x) and hence of f(x). Since f(x) = (x r)h(x)
for some h(x) K[x] and deg h(x) < deg f(x). By induction h(x) has a
splitting eld L over K. Let r
2
, r
3
, . . . , r
n
L be the roots of h(x). Then
L = K(r
2
, r
3
, . . . , r
n
) = F(r
1
, r
2
, . . . , r
n
) is the required splitting eld.
We end this section by presenting a proof of the fundamental theorem of
algebra due to Gauss.
Theorem 7.4 (The Fundamental Theorem of Algebra). Every com-
plex polynomial of positive degree has a complex root.
42
Proof. We shall use the following facts:
(i) Every odd degree polynomial with real coecients has a real root.
(ii) Every quadratic polynomial in C[x] has a complex root.
(iii) The fundamental theorem for symmetric polynomials.
(iv) Every polynomial f(x) has a splitting eld.
(i) This is a consequence of the Intermediate Value Theorem.
(ii) It is enough to show that complex numbers have a complex square root.
Indeed, let z = a + bi C, where a, b R and (c + di)
2
= a + bi. Then
c
2
d
2
+ 2cdi = a +bi. Thus a = c
2
d
2
and b = 2cd. Therefore
a
2
+b
2
= (c
2
+d
2
)
2
c
2
+d
2
=

a
2
+b
2
R.
Therefore c
2
=
1
2
[a +

a
2
+b
2
] 0 and d
2
=
1
2
[

a
2
+b
2
a] 0. Thus
c, d R.
The polynomial g(x) = f(x)

f(x) R[x]. Here

f denotes the polynomial
whose coecients are conjugates of the coecients of f(x). If g(x) has a
complex root z then either f(z) = 0 or

f(z) = 0. If

f(z) = 0, then f( z) = 0.
Thus by replacing f by g, we may assume that f(x) is a monic polynomial
with real coecients.
Let d = deg f = 2
n
q, where q is odd. We apply induction on n. If n = 0,
then f is a real odd degree polynomial, hence it has a real root. Now let
n 1. Let K = C(
1
, . . . ,
2
), be a splitting eld of f(x), over C, where

1
, . . . ,
d
are the roots of f(x) in K. Consider the elements
y
ij
=
i
+
j
+r
i

j
,
where r R is xed and 1 i j d. There are

d+1
2

such pairs (i, j).

Hence
deg h(x) =

1ijd
(x y
ij
) =

d + 1
2

= 2
n1
q(d + 1).
The coecients of h(x) are elementary symmetric polynomials in y
ij
s. So
they are symmetric polynomials in
1
,
2
, . . . ,
d
. Hence they are polynomi-
als in the coecients of f(x). Hence h(x) R[x]. By induction on n, h(x)
43
has a complex root say z
r
. Since all the roots of h(x) K and z
r
C K,
z
r
=
i(r)
+
j(r)
+r
i(r)

j(r)
for some pair (i(r), j(r)) so that 1 i(r) j(r) d. Dene
: R {(i, j) | 1 i(r) j(r) d} = P, (r) = (i(r), j(r)).
Since R is innite and P is nite, there exists c = d R such that
(i(c), j(c)) = (i(d), j(d)) := (a, b). Therefore,
z
c
=
i(c)
+
j(c)
+r
i(c)

j(c)
=
a
+
b
+c
a

b
= z
d
=
a
+
b
+d
a

b
.
Therefore (d c)
a

b
= z
d
z
c
C. Hence
a

b
C so that
a
+
b
C.
But
a
and
b
are roots of
x
2
(
a
+
b
)x +
a

b
C[x].
Hence
a
,
b
C. Therefore f(x) has a complex root.
45
Lecture 8 : Algebraic Closure of a Field
Objectives
(1) Existence and isomorphisms of algebraic closures.
(2) Isomorphism of splitting elds of a polynomial.
Key words and phrases: algebraically closed eld, algebraic closure, split-
ting eld.
8. Algebraically Closed Fields
In the previous section we showed that all complex polynomials of positive
degree split in C[x] as products of linear polynomials in C[x]. While working
with polynomials with coecients in a eld F, it is desirable to have a eld
extension K/F so that all polynomials in K[x] split as product of linear
polynomials in K[x].
Denition 8.1. A eld F is called an algebraically closed eld if every
polynomial f(x) F[x] of positive degree has a root in F.
It is easy to see that a eld F is algebraically closed if and only if f(x) is
a product of linear factors in F[x]. The fundamental theorem of algebra
asserts that C is an algebraically closed eld. Let us show that any eld is
contained in an algebraically closed eld.
Existence of algebraic closure
Theorem 8.2. Let k be a eld. Then there exists an algebraically closed
eld containing k.
Proof. (Artin) We construct a eld K k in which every polynomial of
positive degree in k[x] has a root. Let S be a set of indeterminates which
is in 1 1 correspondence with set of all polynomials in k[x] of degree 1.
Let x
f
denote the indeterminate in S corresponding to f.
46
Let I = (f(x
f
) | deg f 1) be the ideal generated by all the polynomials
f(x
f
) k[S]. We claim that I is a proper ideal of k[S]. Suppose to the
contrary, I = k[S]. Then
1 = g
1
f
1
(x
f
1
) + + g
n
f
n
(x
fn
) (1)
for some g
1
, g
2
, . . . , g
n
k[S]. The polynomial g
1
, g
2
, . . . , g
n
involve only
nitely many variables. Put x
f
i
= x
i
for i = 1, 2, . . . , n and let x
n+1
, . . . , x
m
be the remaining variables in g
1
, g
2
, . . . , g
n
. Then
n

i=1
g
i
(x
1
, x
2
, . . . , x
n
, x
n+1
, . . . , x
m
)f
i
(x
i
) = 1.
Let E/k be an extension eld in which the polynomials f
1
(x
1
), . . . , f
n
(x
n
)
have roots
1
, . . . ,
n
respectively. Putting x
n+1
= . . . = x
m
= 0 and x
i
=
i
for all i = 1, 2, . . . , n in the equation 1 we get a contradiction. Hence I is a
proper ideal of k[S]. Let m be a maximal ideal of k[S] containing I. Then
K
1
= k[S]/m is a eld. We claim that x
f
+ m is a root of f(x). Indeed,
f(x
f
+m) = f(x
f
) +m = m. Thus each polynomial in k[x] has a root in K
1
.
Repeat the procedure on K
1
to get K
2
K
1
which has roots of all monic
polynomials in K
1
[x]. Let K =

i=1
K
i
. Then K is a eld. If f(x) K[x]
then f(x) K
n
[x] for some n. Hence f(x) has a root in K
n+1
K. Thus
K is algebraically closed.
Corollary 8.3. Let F be a Field. Then there exists a eld K F such that
K is algebraically closed and K is algebraic over F.
Proof. Let L F be an algebraically closed eld. Then the eld
K = {a L | a is algebraic over F}
is algebraically closed and it is algebraic over F.
Denition 8.4. Let F be a eld. An extension K/F is called an algebraic
closure of F if K is algebraically closed and K/F is an algebraic extension.
Isomorphism of algebraic closures
We now show that if E
1
and E
2
are algebraic closures of a eld F then they
are Fisomorphic. As a consequence we also prove that any two splitting
47
elds of a polynomial f(x) F[x] are Fisomorphic. Extensions of embed-
dings of elds is one of the main observations in various arguments in Galois
theory. The next result prepares us for the theorem about isomorphism of
algebraic closures of a eld.
Proposition 8.5. Let : k L be an embedding of elds where L is
algebraically closed. Let be algebraic over k and p(x) = irr(, k). Let
p(x) =

a
i
x
i
k[x] and p

(x) =

(a
i
)x
i
. Then () is a bijection
between the sets
{ : k() L | is an embedding and |
k
= } { L | p

() = 0}.
Proof. Let : k() L be an embedding extending . Then
(p()) = p

(()) = 0.
Hence () is a root of p

(x). Conversely let L and p

() = 0. Dene
: k() L by (f()) = f

(). We show that is well dened.

Suppose f() = g(). Then (f g)() = 0, so p(x) | (f(x) g(x)). Hence
p

(x) | (f g)

(x). Thus p

() = (f

() g

() = 0. Hence f

() =
(f()) = g

() = (g()). Thus is well-dened. Suppose that f

() =
(f()) = 0. Then p

(x) | f

(x). Since is an embedding, p(x) | f(x).

Thus f() = 0.

Proposition 8.6. Let : k L be an embedding of elds where L is

algebraically closed. Let E be an algebraic extension of k. Then there exists
an embedding : E L extending . If E is an algebraic closure of k and
L is an algebraic closure of (k) then is an isomorphism extending .
Proof. Consider the set
S = {(F, ) | k F E are elds and : F L such that |
k
= }.
Since (k, ) S, it is nonempty. Let (F, ) and (F

) S. Dene
(F, ) (F

) if and only if F F

and

|
F
= .
Then S is a partially ordered inductive set. Indeed, if {(F

)}
I
is a
chain in S then F =
I
F

is a subeld of E. Dene : F L as
(x) =

(x) if x F

. Then is well-dened.
48
By Zorns Lemma there exists a maximal element (F, ) S. We claim
that F = E. Suppose there exists E \ F. Since is algebraic over
F, : F L can be extended to F() L. This contradicts maximality of
(F, ). Thus E = F. Hence can be extended to an embedding of E into L.
Now suppose E is an algebraic closure of k and L is an algebraic closure of
(k). Since (E) is algebraically closed and L is algebraic over (E), L =
(E). Thus : E L is an isomorphism.
Theorem 8.7. If E
1
and E
2
are algebraic closures of a eld k then they
are kisomorphic.
Proof. The identity map k E
2
can be extended to : E
1
E
2
by the
above proposition, is a kisomorphism.
Theorem 8.8. Let E and F be splitting elds of polynomial f(x) k[x]
where k is a eld. Then they are k-isomorphic.
Proof. Let F
a
be an algebraic closure of F. Then it is also an algebraic
closure of k. Thus there exists an embedding : E k
a
extending id
k
:
k k
a
. Let f(x) = (x
1
) (x
n
) be a factorization of f(x) in E[x].
Then
f

(x) = (x (
1
)) (x (
n
)) F
a
[x].
Thus F = k((
1
), . . . , (
n
)) = (E) as k
a
contains a unique splitting eld
of any polynomial in k[x].
54
Lecture 9 : Separable Extensions I
Objectives
(1) Criterion for multiple roots in terms of derivatives
(2) Irreducible polynomials are separable over elds of characteristic
zero.
(3) Characterization of perfect elds of positive characteristic,
Key words and phrases: Separable polynomial, separable element, sepa-
rable extensions, derivative of a polynomial, perfect elds.
9. Separable Extensions
Let F be a eld. We have seen that the discriminant of a polynomial
f(x) F[x] vanishes if and only if f(x) has a repeated root. Calculation
of discriminant can be dicult. In this section we discuss an eective cri-
terion in terms of derivatives of polynomials whether certain root of f(x) is
repeated. We will also study elds F so that no irreducible polynomial in
F[x] has repeated roots.
Let E be a splitting eld of a monic polynomial f(x) F[x] of degree n.
Write in E[x] the unique factorization of f(x).
f(x) = (x r
1
)
e
1
(x r
2
)
e
2
(x r
g
)
en
.
where r
1
, . . . , r
g
E and e
1
, e
2
, . . . , e
g
are positive integers.
Denition 9.1. The numbers e
1
, e
2
, . . . , e
n
are called the multiplicities of
r
1
, r
2
, . . . , r
n
respectively. If e
i
= 1 for some i, then r
i
is called a simple
root. If e
i
> 1 then r
i
is called a multiple root. A polynomial f(x) with no
multiple roots is called a separable polynomial.
Proposition 9.2. The numbers of roots and their multiplicities are inde-
pendent of a splitting eld chosen for f(x) over F.
55
Proof. Let E and K be splitting elds of f(x) over F. Then there is an
Fisomorphism : E K. This isomorphism gives rise to an isomorphism

: E[x] K[x],

i
a
i
x
i

i
(a
i
) x
i
.
Let f(x) =

g
i=1
(x r
i
)
e
i
be the unique factorization of f(x) E[x]. Then

(f(x)) =

g
i=1
(x (r
i
))
e
i
. Since K[x] is UFD, (r
1
), . . . , (r
g
) are the
roots of

(f(x)) = f(x) with multiplicities e

1
, . . . , e
g
in K respectively.
The derivative criterion for multiple roots
Let f(x) = a
0
+ a
1
x + + a
n
x
n
F[x]. We can dene derivative of f(x)
without appealing to limits. This is preferable since F may not be equipped
with a distance function.
The derivative of f(x), is dened by f

(x) :=

m
i=0
ia
i
x
i1
. It is easy to
check that the usual formulas for (f(x)g(x))

, (f(x)g(x))

and (f(x)/g(x))

where g(x) = 0 hold for derivatives of polynomials.

Theorem 9.3. Let f(x) F[x] be a monic polynomial.
(1) If f

(x) = 0 then every root of f(x) is a multiple root.

(2) If f

(x) = 0 then f(x) has simple roots if and only if gcd(f, f

) = 1.
Proof. (1) Let f(x) = (x r)g(x). Then
0 = f

(x) = g(x) + (x r)g

(x).
Thus g(x) = (xr)g

(x), so r is a root of g(x). Hence r is a multiple root.

(2) () Let gcd(f, f

) = 1 and let r be a multiple root of f(x). Then

f(x) = (x r)
2
g(x) in some splitting eld E of f(x) over F. Thus
f

(x) = (x r)
2
g

(x) + 2(x r)g(x).

Hence f

(r) = 0. If d(x) = gcd(f(x), f

(x)) F[x] then

d(x) = p(x)f(x) +q(x)f

(x)
for some p(x), q(x) F[x]. Hence d(r) = 0. Therefore, deg d(x) 1, so
gcd(f, f

) = 1, which is a contradiction. Therefore f(x) has only simple

roots.
56
(). Let r
1
, r
2
, . . . , r
n
be the roots of f(x) and assume that they are simple.
Then
f(x) = (x r
1
)(x r
2
) (x r
n
) and f

(x) =

n
i=1
f(x)
(xr
i
)
.
Therefore (x r
i
) does not divide f

(x) any i. Hence f and f

have no
common root. Therefore gcd(f, f

) = 1.

Proposition 9.4. (1) Let f(x) F[x] be an irreducible polynomial. Then

f(x) is separable if and only if f

= 0.
(2) Irreducible monic polynomials over a eld of characteristic zero are sep-
arable.
Proof. (1) () If f

= 0, then every root of f(x) is a multiple root.

() Suppose r is a multiple root of f(x). Then f

(r) = 0. Since f(x)

is irreducible, f(x) | f

(x). But this is a contradiction since deg f

(x) <
deg f(x). Therefore f(x) is separable.
(2) If char F = 0, and f(x) is of positive degree, then f

(x) = 0.
Proposition 9.5. Let F be a eld of positive characteristic p. Then x
p
a
F[x] is either irreducible in F[x] or a F
p
.
Proof. Suppose f(x) = x
p
a = g(x)h(x) where 1 deg g = m < p
Let b be a root of f(x) in a splitting eld E of f(x). Then a = b
p
, so
f(x) = (x b)
p
. Hence b is also a root of g(x). Thus g(x) = (x b)
m
Then
b
m
F. Since (p, m) = 1, there exists x, y Z such that px+my = 1. Hence
b = b
px+my
= a
x
(b
m
)
y
F. Thus b
p
= a F
p
.
Example 9.6. We construct an irreducible polynomial with a multiple root.
Let F = F
p
(t) be the quotient eld of the polynomial ring F
p
[t]. Let f(x) =
x
p
t F[x]. Then t / F
p
. Suppose t is a p
th
power and
t =
g(t)
p
h(t)
p
=

i
a
i
t
i

p
(

i
b
i
t
i
)
p
.
Then t(

b
p
i
t
ip
) =

i
a
p
i
t
ip
. Hence a
i
= b
i
= 0 for all i. Thus x
p
t
is irreducible. Another way to see that x
p
t is irreducible is to apply
57
Eisensteins Criterion with t as a prime element. Let E be a splitting eld
of f(x) over F and u be a root of f(x). Then u
p
= t so x
p
t = (x u)
p
,
Hence f(x) has only one root in E.
Proposition 9.7. Let f(x) F[x] where char F = p, be an irreducible
polynomial. If f(x) is not separable then there exists g(x) F[x] such that
f(x) = g(x
p
).
Proof. Since f(x) =

i
a
i
x
i
is irreducible and inseparable, we have f

(x) =

(ia
i
)x
i1
= 0. Therefore i = pt
i
for some t
i
N. Hence
f(x) =

a
pt
i
x
pt
i
=

a
pi
(x
p
)
t
i
.

Perfect Fields
We have seen that irreducible polynomial over elds of characteristic 0 are
separable. But over a eld of positive characteristic, irreducible polynomial
may not be separable. We now discuss a condition on a eld F of positive
characteristic which will ensure that irreducible polynomials in F[x] are
separable.
Denition 9.8. Let F K be a eld extension. An algebraic element
K is a called separable element over F if irr(, F) is separable.
We say K/F is a separable algebraic extension if each element of K
is separable. We say F is a perfect eld if each algebraic extension is
separable.
Any eld of characteristic zero is perfect. By the previous example F
p
(t) is
not perfect. This is basically due to t not being a p
th
power in F
p
(t).
Theorem 9.9. Let F be a eld of positive characteristic p. Then F is perfect
if and only if
F = F
p
= {a
p
| a F}.
Proof. Suppose a F\F
p
. Then x
p
a F[x] is irreducible and inseparable.
Hence F is not perfect.
58
() Let F = F
p
and f(x) F[x] be an irreducible polynomial. If f(x) is
inseparable, then f(x) = g(x
p
) =

a
i
(x
p
)
i
=

(b
i
)
p
(x
p
)
i
= (

b
i
x
i
)
p
for
some b
i
F. This contradicts irreducibility of f(x). Hence f(x) is separable.

Corollary 9.10. Every nite eld is perfect.

Proof. Let |F| = p
n
. By Lagrange theorem applied to the multiplicative
group F

we get
p
n
1
= 1 for all F

. Hence
p
n
= for all F.
Therefore = (
p
n1
)
p
.
59
Lecture 10 : Separable Extensions II
Objectives
(1) Roots of an irreducible polynomial have equal multiplicity.
(2) Separable nite algebraic extensions and separable degree.
(3) Transitivity of separable extensions
Key words and phrases: Separable degree, extensions of embeddings.
Proposition 9.11. Let F be a eld and f(x) F[x] be a monic irreducible
polynomial. Then all roots of f(x) have equal multiplicity. If char F = 0
then all roots of f(x) are simple and if char F = p > 0 then all roots of
f(x) have multiplicity p
n
for some n.
Proof. Let , be roots of f(x) in F. Consider the F-isomorphism :
F() F() given by () = . Then can be extended to an automor-
phism of F. Let f(x) = (x)
m
h(x) where h(x) F[x] and is not a root
of h(x). Then f(x) = (f(x)) = (x )
m
(h(x)). Hence the multiplicity of
is at least m. By symmetry we conclude that both roots have the same
multiplicity.
We know that irreducible polynomials are separable if char F = 0. Let
char F = p > 0. Suppose f(x) has roots of positive multiplicity. Then there
exists a separable polynomial g(x) so that f(x) = g(x
p
n
). Let r
1
, r
2
, . . . , r
g
be distinct roots of f(x) in F. Then
f(x) = (x r
1
)
p
n
(x r
2
)
p
n
. . . (x r
g
)
p
n

Separability and extensions of embeddings

Let E = k(a) be an algebraic extension of a eld k. Let p(x) = irr(a, k).
We have seen that if : k L is an embedding of elds where L is alge-
braically closed then the number of embeddings : E L extending is
equal to the number of distinct roots of p

(x) in L. Hence if p(x) is separa-

ble, then the number of extensions of to embeddings of E into L is [E : k].
Conversely, if has [E : k] extensions, then for any such extension , () is
60
a root of p

(x). Hence p(x) is separable. We now discuss this phenomenon

for nite algebraic extensions.
Let : F L be an embedding of elds where L is an algebraic closure
of (F). Let : F L

be an embedding of elds where L

is an algebraic
closure of (F). Let E be an algebraic extension of F. Let S

(resp. S

)
denote the set of extensions of (resp. ) to embeddings of E into L (resp.
L

). Consider the following diagram of elds and embeddings:

L

oo

(E) E

oo

//

(E)
(F) F

oo

//
(F)
Let be an extension of the embedding
1
: (F) (F) to an
isomorphism : L L

.
Theorem 9.12. The map
: S

, (

) =

is a bijection.
Proof. If

then for any x F we have

(x) = (x) =
1
((x)) = (x).
Hence

is an extension of to an embedding of E into L

. Hence
induces a mapping : S

dened by (

) =

. Since is an
isomorphism, is a bijection.

Denition 9.13. If E/F is an algebraic extension then the cardinality of

S

is called the separable degree of E/F and it is denoted by [E : F]

s
.
Proposition 9.14. Let k F E be a tower of nite algebraic extensions.
Then [E : k]
s
[E : k] and
[E : k]
s
= [E : F]
s
[F : k]
s
.
61
Proof. First we show that the separable degree is multiplicative in a tower
of eld extensions. Let : k L be an embedding into an algebraically
closed eld L. Let (
i
)
iI
be distinct extensions of to embeddings of F
into L. Each
i
has [E : F]
s
extensions to embeddings of E into L. Let these
be (
ij
). Hence (
ij
) has cardinality [F : k]
s
[E : k]
s
. If : E L is an
embedding extending , then |
F
is an extension of to an embedding of F
into L. Hence |
F
=
ij
. This proves the multiplicativity of separable degree
in a tower of eld extensions.
Since E/k is nite, there exist elements a
1
, a
2
, . . . , a
n
such that
k k(a
1
) k(a
1
, a
2
) k(a
1
, a
2
, . . . , a
n
).
Each step in the above tower is a simple algebraic extension. Hence the
separable degree of each step is atmost its degree. Since the separable degree
and degree of a eld extension are multiplicative, we have [E : k]
s
[E : k].

Corollary 9.15. Let k F K be a tower of nite extensions. Then

[E : k]
s
= [E : k] if and only if the corresponding equality holds in each step
of the tower.
Theorem 9.16. Let E/k be a nite extension. Then E/k is separable if
and only if [E : k]
s
= [E : k].
Proof. Let E/k be nite separable extension. Then E = k(a
1
, a
2
, . . . , a
n
)
for some a
1
, a
2
, . . . , a
n
E. Then each a
i
is separable over k. Hence a
i
is
separable over k(a
1
, a
2
, . . . , a
i1
) for i = 1, 2, . . . , n.
[k(a
1
, a
2
, . . . , a
i
) : k(a
1
, a
2
, . . . , a
i1
)]
s
= k(a
1
, a
2
, . . . , a
i
) : k(a
1
, a
2
, . . . , a
i1
)].
for i = 1, 2, . . . , n, whence [E : k]
s
= [E : k].
Conversely let [E : k]
s
= [E : k]. Then using the fact that the separable
degree and degree are multiplicative and the separable degree is atmost the
degree, we conclude that for any a E, [E : k(a)]
s
[k(a) : k]
s
= [E : k]
s
=
[E : k]. Hence [k(a) : k]
s
= [k(a) : k]. Hence irr(a, k) is separable. Thus E/k
is a separable extension.

63
Lecture 11 : Finite Fields I
Objectives
(1) Existence and uniqueness of nite elds.
(2) Algebraic closure of a nite eld.
(3) Finite subgroup of the multiplicative group of a eld is cyclic.
(4) Gauss formula for the number of monic irreducible polynomials of
a given degree over a nite eld.
Key words and phrases: Finite eld, Gauss formula for irreducible poly-
nomials,
10. Finite Fields
A nite eld F of prime characteristic p contains a prime eld F
p
. Since F
is a nite dimensional vector space over F
p
, |F| = p
n
, where n = [F : F
p
].
We usually write p
n
= q.
Proposition 10.1. Let K and L be nite elds of cardinality q = p
n
, where
p is a prime number. Then K and L are isomorphic.
Proof. Since |K

| = q 1, by Lagranges theorem x
q1
= 1 for all x K

.
Thus every element of K is a root of the polynomial x
q
x = f(x). Hence
K is a splitting eld of f(x) over F
p
. Since any two splitting elds of f(x)
over F
p
are isomorphic, K L.
Notation: We shall denote a nite eld with p
n
elements by F
p
n.
Corollary 10.2 (Wilson). Let p be a prime number. Then
(p 1)! 1( mod p).
64
Proof. The assertion is clear for p = 2. Let p be odd. Since F
p
is the set of
roots of x
p
n
x, taking n = 1 we get
x
p1
1 = (x 1)(x 2) . . . (x (p 1)).
Putting x = 0 we obtain (p 1)! 1( mod p).
Proposition 10.3. For any prime p and any n N, there exists a nite
eld of cardinality p
n
. An algebraic closure F
a
p
of F
p
has a unique subeld
with p
n
elements for every n N and
F
a
p
=

nN
F
p
n.
Proof. Let q = p
n
. Then F
a
p
contains a unique splitting eld of x
q
x = f(x)
over F
p
. Let
K = { F

p
| f() = 0}.
Then K is a eld. The polynomial x
q
x is separable since its derivative is
1. Hence K has q elements. Therefore K is the required nite eld with q
elements. Let a be algebraic over F
p
and [F
p
(a) : F
p
] = n. Then F
p
(a) has
p
n
elements. Hence a F
p
n. Thus F
a
p
=

nN
F
p
n.
Theorem 10.4. Let U be a nite subgroup of the multiplicative group F

of a eld F. Then U is cyclic.

Proof. Let |U| = n. By Lagranges Theorem x
n
= 1 for all x U. Since
U is an abelian group, by the structure theorem for nite abelian groups,
there exist d
1
, d
2
, . . . , d
r
N such that n = d
1
d
2
. . . d
r
, d
1
| d
2
| | d
r
, and
U Z/d
1
Z Z/d
2
Z Z/d
r
Z.
Thus each x U satises x
dr
1 = 0. But x
dr
1 has atmost d
r
roots.
Thus n = d
r
and so U is cyclic.
Counting irreducible polynomials over nite elds
Let N
q
(n) denote the number of irreducible polynomials of degree n over a
nite eld F
q
. We derive a formula, due to Gauss, for N
q
(n). Let be a cyclic
generator of F

q
n. Then F
q
n = F
q
() and deg irr(, F
q
) = [F
q
n : F
q
] = n.
65
Hence N
q
(n) 1. Recall the Mobius inversion formula. Let f, g : N N be
functions so that
f(n) =

d|n
g(d).
Then
g(n) =

d|n
f(n/d)(d),
where is the Mobius function : N N dened as
(n) =

1 for n = 1
(1)
r
if n = p
1
. . . p
r
, where p
1
, . . . , p
r
are distinct primes,
0 if p
2
|n for some prime p.
Theorem 10.5 (Gauss). The number of irreducible monic polynomials of
degree n over F
q
is given by
N
q
(n) =
1
n

d|n
(d)q
n/d
.
Proof. Let f(x) be a monic irreducible polynomial in F
q
[x]. We show that
f(x) | x
q
n
x if and only if deg f | n.
Suppose f(x) | x
q
n
x. Then F
q
n contains all the roots of f(x). Let be a
root of f(x). Then F
q
() F
q
n. Thus [F
q
() : F
q
] = d = deg f | n.
Conversely, let d = deg f | n. Let f() = 0, where F
a
q
. Let be another
root of f(x) in F
a
q
. Then there exists an embedding : F
q
() F
q
()
such that () = . Since F
q
() is a splitting eld of x
q
d
x over F
q
, so
is (F
q
()) = F
q
(). But F
a
q
has only one splitting eld of x
q
d
x, hence
F
q
(). Thus f(x) | x
q
n
x.
Notice that x
q
n
x is a separable polynomial. Hence
x
q
n
x =

d|n
f
(d)
1
(x) f
(d)
Nq(d)
(x),
where f
(d)
1
(x), . . . , f
(d)
Nq(d)
(x) are all the degrees d irreducible polynomials over
F
q
. Equate degrees to get q
n
=

d|n
dN
q
(d). By Mobius inversion
nN
q
(n) =

d|n
(n/d)q
d
.

71
Lecture 12 : The Primitive Element Theorem
Objectives
(1) Factorization of polynomials over nite elds.
(2) The Primitive element theorem.
(3) Finite separable extensions have a primitive element.
Key words and phrases: Primitive element, nite separable extensions,
factorization.
Example 10.6. We know that the polynomial x
p
n
x is the product of
all the degree d monic irreducible polynomials in F
p
[x] where d | n. This
is useful for constructing irreducible polynomials over F
p
. Let us factorize
x
16
x over F
2
. The irreducible quadratic polynomials are factors of x
4
x =
x(x+1)(x
2
+x+1). Hence there is only one quadratic irreducible polynomial
over F
2
. The cubic irreducible are factors of
x
8
x = x(x
7
+ 1) = x(x + 1)(x
6
+ x
5
+ x
4
+ x
3
+ x
2
+ x + 1).
By Gauss formula N
2
(3) = 2. Therefore the irreducible cubics over F
2
are
x
3
+x
2
+1 and x
3
+x+1. By Gauss formula, we count irreducible quartics
over F
2
:
4N
2
(4) =

d|4
(4/d)2
d
= (4)2 + (2)2
2
+ (1)2
4
= 4 + 16 = 12.
Hence N
2
(4) = 3. These quartics are factors of x
16
x. The irreducible
factors of this polynomial have degrees 1, 2 and 4. Therefore the irreducible
quartics are factors of
x
16
x
x(x + 1)(x
2
+ x + 1)
= (x
4
+ x + 1)(x
4
+ x
3
+ 1)(x
4
+ x
3
+ x
2
+ x + 1).
We end this section by an interesting application of nite elds.
Proposition 10.7. The polynomial x
4
+ 1 is irreducible in Z[x] but it is
reducible over F
p
for every p.
72
Proof. Let f(x) = x
4
+1. Then f(x+1) is irreducible over Z by Eisensteins
criterion. For p = 2, we have x
4
+ 1 = (x + 1)
4
. Now let p be odd. Then
8 | p
2
1. Hence
x
4
+ 1 | x
8
1 | x
p
2
1
1 | x
p
2
x.
The splitting eld of x
p
2
x over F
p
is the nite eld F = F
p
2. Hence
[F : F
p
] = 2. Therefore the roots of x
4
+1 in F have degree 1 or 2. Therefore
x
4
+1 cannot have a cubic or quartic irreducible factor over F
p
. Hence it is
reducible over F
p
for each prime p.

11. The Primitive Element Theorem

Since F

q
n is a cyclic group, F
q
n = F
q
() where is a generator of F

q
n. We say
that is primitive element of the eld extension F
q
F
q
n. In this section
we discuss existence of primitive elements in nite algebraic eld extensions.
We will show that in a nite separable extension, primitive elements always
exist.
Denition 11.1. Let E/F be a eld extension. An element E is called
a primitive element of E over F if E = F().
Example 11.2. (1) Let f(x) = x
3
2, =
3

2 and = e
2i/3
. Then
Q(, ) is a splitting eld of f(x). Moreover [Q(, ) : Q] = 6. Since
Q() R, + / Q().
Q(, )
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L

Q
w
w
w
w
w
w
w
w
w
w
Q() Q( + )
r
r
r
r
r
r
r
r
r
r
r
r
H
H
H
H
H
H
H
H
H
H
Q
id
Q
We know that the number of ways id : Q

Q can be extended to
an embedding : Q( + )

Q is deg irr( + , Q) = [Q( + ) : Q].
Since degree irr(, Q()) = 2, id : Q()

Q can be extended in two
73
ways:
2
or . Restriction of this embedding to Q( + ) maps
+ to +
2
or + . In a similar way we can embed Q( + ) onto
Q( + ), Q( +
2
), Q(
2
+
2
) and Q(
2
+ ). Thus [Q( + ) :
Q] = 6. So Q(, ) = Q( + ). Therefore + is a primitive element.
(2) An algebraic extension need not have a primitive element. Let eld k be
a eld with char(k) = p and let u, v be indeterminates. Let E = k(u, v) and
F = k(u
p
, v
p
). Then f(u, v)
p
F for any f(u, v) E. But [E : F] = p
2
. If
y E is a primitive element of E/F then deg irr(y, F) = p
2
. But y
p
F.
This is a contradiction.
Theorem 11.3 (The Primitive Element Theorem). Let E/k be a nite
extension.
(1) There is a primitive element for E/k if and only if the number of
intermediate subelds F such that k F E is nite.
(2) If E/k is a nite and separable extension then it has a primitive element.
Proof. (1) If k is a nite eld then E is nite and hence E

is a cyclic group.
Thus E/k has a primitive element.
Let k be innite and let E/k have nitely many intermediate elds. Suppose
, E. As c varies in k, k( + c) varies over nitely many intermediate
subelds of E/k. Hence, there are c
1
= c
2
k such that k( + c
1
) =
k(+c
2
) := L. Thus (c
1
c
2
) L. Therefore L. Hence L. Thus
k(, ) = k(+c
1
). Proceed inductively to show that E = k(
1
, . . . ,
n
) =
k(
1
+ c
2

2
+ +
n
c
n
) for some c
2
, . . . , c
n
k.
Conversely, let E = k() for some E and f(x) = irr(, k). Let k F
E be a tower of elds. Set h
F
= irr(, F). Then h
F
| f(x) as F varies over
all the intermediate subelds of E/k.
Since h
F
is irreducible over F, it is also irreducible over F
0
, a subeld of F
generated by the coecients of h
F
(x) over k. Since deg h
F
(x) = [E : F] =
[E : F
0
], it follows that F = F
0
. Since there are nitely many divisors of
f(x), there can be only nitely many intermediate elds of E/k.
(2) Now let E/k be a nite separable extension. Then E = k(
1
,
2
, . . . ,
n
).
To show that E/k has a primitive element it is enough to nd a primitive
74
element when n = 2 and then apply induction on n. So let E = k(, ). We
look for a primitive element of the form + c where c k.
Let [E : k] = n. If +c generates E/k, then +c must have n conjugates
(images of + c under the action of n embeddings of E into k
a
). Hence
there exist n k-embeddings
1
,
2
, . . . ,
n
: E

k. which map + c to
n distinct roots of p(x) = irr( + c), k) in

k. Thus + c is a primitive
element if and only if there exist n embeddings
1
, . . .
n
: E

k such that

i
( + c) =
j
( + c), for all i = j, if and only if

i<j
(
i
()
j
()) + c(
i
()
j
()) = 0
if and only if c is not a root of the polynomial
f(x) =

i<j
(
i
()
j
()) + x(
i
()
j
()).
Since k is innite and f(x) has nitely many roots, such a c exists.
79
Lecture 13 : Normal Extensions
Objectives
(1) Normal extensions and their examples.
(2) Characterization of normal extensions in terms of embeddings and
splitting elds.
Key words and phrases: Normal extensions, Galois extensions, Galois
group of a Galois extension.
12. Normal extensions
Suppose F is a eld and E is a splitting eld of f(x) F[x]. Let r
1
, r
2
, . . . , r
n
be distinct roots of f(x) in F. Then E = F(r
1
, r
2
, . . . , r
n
). Suppose that
a E and g(x) = irr (a, F). Let b F be another root of g(x). Then the
map : F(a) F(b) given by (a) = b and (c) = c for all c F is an
F-embedding.
Let : E F be an extension of . Then (r
i
) = r
j
for each i and some
j. Hence (E) E. Since E is a nite dimensional F-vector space and is
injective, it is also surjective. Hence (E) = E. Therefore b E. This shows
that E contains splitting elds of irr (a, F) for all a E. This property is
the dening condition for normal algebraic extensions.
Denition 12.1. An algebraic extension E/F is called a normal exten-
sion if whenever f(x) F[x] is irreducible and has a root in E then f(x)
splits into linear factors in E[x].
Example 12.2. (1) The algebraic closure F
a
of a eld F is a normal ex-
tension of F.
(2) Let be a primitive n
th
root of unity in C. Then Q() is a normal ex-
tension of Q. An element Q() is of the form g() for some g(x) Q[x].
If is a root of irr (, Q) then there is a Qisomorphism : Q() Q()
such that () = . The isomorphism can be extended to an embedding
: Q()

Q. But () =
m
for some m. Hence = (f()) = f(
m
)
80
Q(). Hence Q()/Q is a normal extension. Alternatively, note that Q() is
a splitting eld of x
n
1 over Q.
(3) Every quadratic extension E/F is normal. Let a EF. Then irr (a, F) =
f(x) = x
2
+ bx + c for some b, c F. Let f(x) = (x a)(x s) for some
s E. Hence E is a splitting eld of f(x) over F. Hence E/F is a normal
extension.,
(4) The extensions Q(
4

2)/Q(

2) and Q(

2)/Q are normal but the ex-

tension Q(
4

2)/Q is not a normal extension since the complex roots of

irr (
4

2, Q) are not in Q(
4

2).
(5) If E/F is a normal extension and K is an intermediate subeld of E/F
then E/K is a normal extension.
Lemma 12.3. Let E/F be an algebraic extension. Let : E E be an
Fembedding, then is an automorphism of E.
Proof. We need to prove that (E) = E. Let a E and p(x) = irr (a, F).
Let K be the subeld of E generated by the roots of p(x) in E. Then K is
a nite dimensional F-vector space. Since is an Fembedding, it maps
roots of p(x) to its roots. Hence (K) K. Since is an injective F
linear map of the Fvector space K, dim
F
K = dim
F
(K). Hence is
surjective.
Theorem 12.4. Let E/F be an algebraic extension such that E F
a
. Then
the following conditions are equivalent:
(1) Every F-embedding : E F
a
is an automorphism of E.
(2) E is a splitting eld of a family of polynomials in F[x].
(3) E/F is a normal extension.
81
Proof.
F
a
E

//
(E) = E
F(a)

//
F(b)
F
id
//
F
(1) (2): Let a E and p
a
(x) = irr (a, F). If b F
a
is a root of p
a
(x) then
there is an F-isomorphism : F(a) F(b). The embedding : F(a) F
a
can be extended to an embedding : E F
a
. But (E) = E. Hence b E.
Thus all roots of p
a
(x) are in E. Hence E is a splitting eld of the family of
polynomials (p
a
(x))
aE
.
(2) (3): Let E be a splitting eld of (p
i
(x))
iI
of polynomials in F[x].
Let a E and f(x) = irr (a, F). Let b be any other root of f(x) in F
a
.
Then there is an F-isomorphism : F(a) F(b) so that (a) = b. The
map can be extended to an F-embedding : E F
a
. But maps roots
of (p
i
(x))
iI
to their roots. Hence (E) E. Hence b E. Thus f(x) splits
into linear factors in E[x].
(3) (1) : Let : E F
a
be an F-embedding. Let a E. Then
p(x) = irr (a, F) splits into linear factors in E[x]. Since (a) is a root of
p(x), (a) E. Hence (E) E. By Lemma 12.3, (E) = E.
Proposition 12.5. Let E
1
, E
2
be subelds of a eld E. Let E
1
, E
2
be normal
extensions of F. Then E
1
E
2
/F and E
1
E
2
/F are normal.
Proof. Let E
1
and E
2
be normal extensions of F. Let : E
1
E
2
F
a
be
an F-embedding. Then (E
1
E
2
) = (E
1
)(E
2
) = E
1
E
2
. Similarly observe
that (E
1
E
2
) = (E
1
) (E
2
) = E
1
E
2
. Hence E
1
E
2
is a normal
extension of F.

82
Lecture 14 : Galois group of a Galois Extension I
Objectives
(1) Galois extension and the Galois group of a Galois extension.
(2) Galois group of a nite extension of nite elds and quadratic ex-
tensions.
(3) Galois groups of biquadratic extension.
(4) Galois group of a separable cubic polynomial.
(5) Fundamental Theorem of Galois theory (FTGT).
Keywords and phrases: Biquadratic and cubic extensions, fundamental
theorem of Galois Theory.
13. The Galois Group of a Field Extension
Denition 13.1. A eld extension E/F is called a Galois extension if it
is normal and separable. The Galois group of a Galois extension E/F
denoted by G(E/F) or Gal(E/F) is the group of all F-automorphisms of E
under composition of maps.
Proposition 13.2. The Galois group of the Galois extension F
q
n/F
q
is a
cyclic group of order n generated by the Frobenious automorphism : F
q
n
F
q
n, dened as (a) = a
q
.
Proof. Note that is an F
q
-automorphism since any a F
q
is a root of
x
q
x. Let G = . Then
n
(x) = x
q
n
= x. Therefore [G[ n. Suppose
[G[ = d. Then
d
= id, so (x) = x
q
d
= x. But x
q
d
x has atmost q
d
roots.
Thus d = n.
We now show that G(F
q
n/F
q
) = . Since F
q
n/F
q
is a separable extension,
[F
q
n : F
q
]
s
= n. Hence the number of F
q
-automorphisms of F
q
n is n whence
= G(F
q
n/F
q
).
Example 13.3. Quadratic extensions: Let K/F be a separable qua-
dratic extension. Then for any K F we have irr (, F) = f(x) =
x
2
+ bx + c. Let be another root of f(x). Then + = b and = c
and f(x) = (x )(x ) K[x]. Hence K/F is a normal extension.
83
Let : K = F() K be a K- automorphism dierent from id
F
. Then
() = . Thus G(K/F) = id
F
, is a group of order 2.
Example 13.4. Biquadratic extensions: A eld extension K/F is called
biquadratic if [K : F] = 4 and K is generated by roots of two irreducible
quadratic separable polynomials. Let K = F(, ) and irr(, F) = x
2
a
and irr(, F) = x
2
b.
F(, )
I
I
I
I
I
I
I
I
I
u
u
u
u
u
u
u
u
u
F()
J
J
J
J
J
J
J
J
J
J
F()
t
t
t
t
t
t
t
t
t
t
F
Since [F(, ) : F] = 4, x
2
a is irreducible over F() and x
2
b is
irreducible over F(). Any F- automorphism of K maps to or and
to or . Let () = , () = and () = , () = . Then
= and
2
=
2
= id. Therefore
G(K/F) = id, , , =
is the Klein 4-group.
Example 13.5. The Galois group of a separable cubic : Let F be
a eld of char ,= 2, 3. Consider an irreducible cubic polynomial f(x) =
x
3
+ px + q F[x]. Thus f(x) has no root in F. Let us observe that f(x)
is separable over F. Since f

(x) = 3x
2
+ p we have
f =
x
3
(3x
2
+ p) +
2p
3
x + q
and hence
gcd(f, f

) =

2p
3
x + q, 3x
2
+ p

.
Since f has no root in F, 2px/3 +q does not divide f(x). Hence (f, f

) = 1
and so f(x) is separable. Thus a splitting eld E of f must have degree 3 or
6. Let E = F(
1
,
2
,
3
) where
1
,
2
,
3
are the roots of f(x) in E. Then
any F- automorphism permutes the roots
1
,
2
,
3
.
84
Dene : G(E/F) S
3
by () = p

where p

is the corresponding
permutation. It is easy to check that is an injective group homomorphism.
Hence G(E/F) S
3
or A
3
. Let us see how disc (f(x)) determines the Galois
group. We identify G(E/F) with a subgroup of S
3
. Let
= (
1

2
)(
2

3
)(
1

3
).
Then
2
= disc (f(x)) = (4p
3
+ 27q
2
) F. Hence [F() : F] 2. If
disc (f(x)) is not perfect square in F then 2 [ [E : F]. hence G(E/F) = S
3
.
If disc (f) is a square in F then F and hence G(K/F) cannot have any
odd permutations since these do not x . Thus G(E/F) = A
3
. For example,
if f(x) = x
3
+ x + 1, then disc (f) = 31. Therefore G(E/F) = S
3
. If
f(x) = x
3
3x + 1, then G(E/F) = A
3
as disc (f) = 3
4
.
14. The Fundamental Theorem of Galois Theory
Let F be a eld. We know that a Splitting eld E of a polynomial
f(x) F[x] is a normal extension of F. If f(x) is separable then E/F is
separable. Thus a splitting eld of a separable polynomial f(x) F[x] is a
Galois extension of F. Conversely if E/F is a nite Galois extension then
by the Primitive Element Theorem there is an a E such that E = F(a).
Since E/F is normal, E is a splitting eld of irr (a, F). Thus a nite
extension E/F is Galois if and only if E is a splitting eld of a separable
polynomial f(x) over F. We say in this case that G(E/F) is the Galois
group of f(x). Since any two splitting elds of f(x) are F-isomorphic, we
write G(E/F) = G
f
.
Denition 14.1. Let G be a group of automorphism of a eld E. Then
E
G
= a E [ (a) = a for all G
is called the xed eld of G acting on E.
Theorem 14.2 (Fundamental Theorem of Galois Theory (FTGT)).
Let E/F be a nite Galois extension. Consider the sets:
J = K [ K is an intermediate eld of E/F and ( = H [ H < G(E/F).
(i) The maps:
K G(E/K) and H E
H
85
give a one-to-one correspondence, called the Galois correspondence be-
tween J and (.
(ii) K/F is Galois if and only if G(E/K) G(E/F) and in this case
G(K/F)
G(E/F)
G(E/K)
.
(iii) [E : K] = [G(E/K)[.
The FTGT will be proved in several steps. We shall prove parts of it for
innite Galois extensions.
Theorem 14.3. Let E/F be a Galois extension with G = G(E/F). Then
(1) F = E
G
.
(2) Let K be an intermediate subeld of E/F. Then E/K is Galois and the
map K G(E/K) is an injective map from J to (.
Proof. (1) Let a E
G
. Let : F(a) F be an F-embedding. Let :
E F be an extension of . Since E/F is Galois, is an automorphism
of E. Hence (a) = a. Therefore [F(a) : F]
s
= 1. But E/F is separable, so
F(a)/F is also separable. Thus [F(a) : F]
s
= [F(a) : F] = 1. So a F.
(2) Let K be an intermediate subeld of E/F. Then E/K is separable as
E/F is so. Let : E K = F be a K-embedding. Then it is also an
F-embedding. As E/F is normal, is an automorphism of E. Thus E/K
is a Galois extension. Let H = G(E/K). Then by (1), we have K = E
H
.
Let K and K

be intermediate subelds of E/F. If H = G(E/K) and H

=
G(E/K

) then K = E
H
and K

= E
H

. Hence the map K G(E/K) is an

injective map.

86
Lecture 15 : Galois group of a Galois Extension II
Objectives
(1) Artins Theorem about xed eld of a nite group of automorphisms.
(2) Behavior of Galois group under isomorphisms.
(3) Normal subgroups of the Galois groups and their xed elds.
Keywords and phrases: Fixed eld, Galois correspondence, normal sub-
groups of Galois group.
The next theorem is perhaps the most important ingredient of the Funda-
mental Theorem of Galois Theory (FTGT). We will need the following
Lemma 14.4. Let E/F be a separable algebraic extension. suppose that for
all E, deg irr (, F) n. Then [E : F] n.
Proof. Let E be such that deg irr(, F) is maximal among deg irr(, F)
for E. We claim that E = F(). Suppose E ,= F() and choose
EF(). Then F(, ) is a nite separable extension. By the Primitive
Element Theorem, there exists F(, ) such that F(, ) = F(). But
then deg > deg .
The above lemma is not true without separability assumption. For example,
deg
F
p for all k(u, v), where F = k(u
p
, v
p
), where k is a eld of
char p > 0. But [k(u, v) : k(u
p
, v
p
)] = p
2
.
Theorem 14.5 (Emil Artin). Let E be a eld and G a nite group of
automorphisms of E. Then
(1) E/E
G
is a nite Galois extension.
(2) G(E/E
G
) = G. (3) [E : E
G
] = [G[.
Proof. (1) Let E and G =
1
,
2
, . . . ,
n
and S =
1
(), . . . ,
n
().
Suppose [S[ = r. Without loss of generality let S =
1
(), . . . ,
r
(). If
87
G then
1
(), . . . ,
r
() are distinct. Hence S = (S). So restricted
to S is a permutation of S. Consider the polynomial
f(x) = (x
1
())(x
1
())(x
2
()) (x
r
()).
The coecient of f(x) are elementary symmetric functions of
1
() . . . ,
r
().
Since (S) = S these elementary symmetric functions are in E
G
. Thus
f(x) E
G
[x] is a separable polynomial and f() = 0. Hence E/E
G
is a sep-
arable and normal extension. Moreover for all E, deg irr (, E
G
) [G[.
Hence [E : E
G
] [G[. Thus E/E
G
is a nite Galois extension.
(2) and (3) : Since E/E
G
is a nite separable extension, [E : E
G
] is the
number of E
G
-embeddings of E E
a
. These embeddings are automor-
phisms of E as E/E
G
is a normal extension. Using (1) and the fact that
G G(E/E
G
), we get
[G[ [G(E/E
G
)[ = [E : E
G
] [G[.
Thus [G[ = [G(E/E
G
)[ = [E : E
G
] and so G = G(E/E
G
).
Theorem 14.6. Let E/F be a Galois extension with Galois group G. Let
K
1
and K
2
be intermediate subelds of E/F and H
1
= G(E/K
1
), H
2
=
G(E/K
2
). Let (H
1
, H
2
) denote the smallest subgroup containing H
1
and H
2
.
Then
K
1
K
2
= E
H
1
H
2
, K
1
K
2
= E
(H
1
,H
2
)
, and K
1
K
2
H
1
H
2
.
Proof. Since E/K
i
is Galois for i = 1, 2, we have K
i
= E
H
i
E
H
1
H
2
for
i = 1, 2. Therefore K
1
K
2
E
H
1
H
2
. Conversely, if G xes K
1
K
2
then it
xes K
1
and K
2
, consequently H
1
H
2
. Hence G(E/K
1
K
2
) H
1
H
2
.
Hence K
1
K
2
E
H
1
H
2
. The remaining statements are obvious.
Behavior of Galois groups under isomorphisms
Proposition 14.7. Let E/F be a Galois extension. Let : E (E) be
an isomorphism of elds. Then
(1) (E)/(F) is a Galois extension.
(2) G((E)/(F)) = G(E/F)
1
G(E/F).
88
Proof. (1) Since E/F is Galois, E is a splitting eld of a family of separable
polynomials f
i
(x) F[x] [ i . Then (E) is a splitting eld of of the
family of polynomials: f

i
(x) (F)[x] [ i . Hence (E) is a Galois
extension of (F).
(2) Dene : G(E/F) G(E/F) by the rule
1
.
(E)

1
//
E

//
E

//
(E)
(F)

1
//
F
//
F

//
(F)
The inverse of is given by the rule
1
. Hence is an isomorphism.

Theorem 14.8. Let E/F be a Galois extension. Let K be an intermediate

subeld of E/F. Then
(1) K/F is Galois if and only if G(E/K) G(E/F).
(2) If K/F is Galois, then G(K/F) G(E/F)/G(E/K)
Proof. (1) and (2) : Let K/F be Galois. Dene
: G(E/F) G(K/F) by () = [
K
.
Since K is a normal extension of F, [
K
G(K/F). Since
Kernel = G(E/F) [ [
K
= id
K
= G(E/K),
G(E/K) is a normal subgroup of G(E/F).
Conversely, let G(E/K)G(E/F). Let : E E be any F-automorphism.
We show that K = K. Now
G(E/K)
1
= G(E/K) = G(E/K),
provided G(E/F). Thus K = K. Let : K F
a
be an F-embedding.
Then can be extended to an embedding : E F
a
. Since E/F is Galois,
(E) = E. Thus (K) = K. Hence K/F is Galois.

96
Lecture 16 : Applications and Illustrations of the FTGT
Objectives
(1) Fundamental theorem of algebra via FTGT.
(2) Gauss criterion for constructible regular polygons.
(3) Symmetric rational functions.
(4) Galois group of some binomials.
Keywords and phrases: Fundamental theorem of algebra, constructible
regular polygons, symmetric rational functions.
15. Applications and Illustrations of the FTGT
The Galois correspondence between the set of subelds of a nite Galois
extension E/F and the set of subgroups of the Galois group G(E/F) con-
verts problems about roots of a separable polynomial to problems about the
Galois group of its splitting eld. We shall see that dicult problems about
polynomials are converted into much simpler problems about nite groups.
The Galois correspondence is perhaps the rst example of a well-established
technique in mathematics: nd a suitable formulation for a problem in one
branch of mathematics in another branch where the problem becomes much
easier to solve.
We will see that the Galois correspondence is powerful enough to provide
new ways to prove old results and solve new problems as well. This will
be demonstrated here by giving a new proof of the fundamental theorem of
algebra. We will also nish the proof of Gauss criterion for constructibility
of regular polygons. We shall derive an expression for cos 2/17 in terms
of square roots which proves that a seventeen sided regular polygon is con-
structible by ruler and compass.
We will provide concrete examples of Galois correspondence for some poly-
nomials. In later sections we will derive formulas for the roots of cubic and
quartic polynomials as a consequence of the Galois correspondence. Let us
begin by proving:
97
The Fundamental Theorem of Algebra
Theorem 15.1. The eld of complex numbers is algebraically closed.
Proof. Let f(x) =

a
i
x
i
C[x]. Write

f(x) =

a
i
x
i
where denotes the
complex conjugation. Then g(x) = f(x)

f(x) R[x]. Hence it is enough to
prove g(x) has a complex root.
The splitting eld E of g(x) over C is a splitting eld of (x
2
+ 1)g(x) over
R. Hence E/R is a Galois extension. Since 2 [ [E : R], the Galois group
G = G(E/R) has a 2- Sylow subgroup say S. If S < G then E E
S
R.
We know [E : E
S
] = [S[. Thus [E
S
: R] is odd. But R admits no proper odd
degree algebraic extensions. Hence S = G. Thus G is a 2-group. If [G[ = 2,
then E = C and we are done. If [G[ = 4, then [E : C] = 2. But C admits
no quadratic extension. Thus [G[ 8. Let H < G(E/C) of index 2. Then
[E
H
: C] = 2, which is a contradiction. Hence E = C.
Gauss Criterion for Constructible Regular Polygons
Lemma 15.2. Let m, n be coprime natural numbers. If regular polygons of
m sides and n sides are constructible then so is a regular mn-gon.
Proof. There exist integers x, y so that xm+yn = 1. Hence
2
mn
=
2x
n
+
2y
m
.
Since 2x/n and 2y/m are constructible, so is 2/mn.
Proposition 15.3. Let be a complex primitive p
th
root of unity where p
is a prime number. Then G(Q()/Q) is a cyclic group of order p 1.
Proof. If G, then restricted to the cyclic group U = () is an auto-
morphism. Hence () =
i
for some i = 1, 2, . . . , p 1. Dene a group
homomorphism : G U(Z/pZ) = 1, 2, . . . , p 1 by () = i

. It is
easy to see that is an isomorphism.
Theorem 15.4 (Gauss). A regular polygon of n sides is constructible if
and only if n = 2
r
p
1
p
2
. . . p
s
where r N and p
1
, p
2
, . . . , p
s
are distinct
Fermat primes.
98
Proof. We have already proved the necessity. For suciency, note that by
the above lemma and the fact that angles can be bisected by ruler and
compass, it is enough to prove that if p is a Fermat prime then cos(2/p)
is a constructible real number. Let be a primitive p
th
root of unity. Then
[Q() : Q] = p 1 = 2
t
for some t, the Galois group G = G(Q()/Q) is
cyclic of order 2
t
. Hence every intermediate subeld of Q()/Q is a Galois
extension of Q. In particular K = Q(cos 2/p) is a Galois extension of Q
of degree 2
t1
. Since G(K/Q) is a 2-group of order 2
t1
, there a chain of
subgroups G
i
having order 2
i
for i = 0, 1, . . . , t 1. Hence
Q K
G
t2
K
G
t3
K
G
0
= K
is a tower of real quadratic extensions terminating with K. Hence cos 2/p
is a constructible real number.
Example 15.5. Let K be a splitting eld of x
4
2 over Q. We nd the
Galois group G = G(K/Q) and show how to nd subelds of K/Q.
The polynomial f(x) = x
4
2 is irreducible over Q by Eisensteins criterion.
Let a =
4

2 be the real 4th root of 2. Then the roots of f(x) in C are

a, a, ia, ia. The splitting eld of f(x) over Q is Q(a, i) and [K : Q] = 8.
Hence G = G(K/Q) is a group of order 8. An automorphism in G maps a
to one of the four roots of f(x) and it maps i to either i or i. Let be the
conjugation map and be dened by (a) = ia. Check that
o() = 4, o() = 2 and = id.
The lattice of the subgroups of G is:
G
m
m
m
m
m
m
m
m
m
m
m
m
m
m
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
1,
2
, ,
2

q
q
q
q
q
q
q
q
q
q
P
P
P
P
P
P
P
P
P
P
P
P
1, ,
2
,
3
1,
2
, ,
3

m
m
m
m
m
m
m
m
m
m
m
m
m
O
O
O
O
O
O
O
O
O
O
O
1,
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
W
1,
2

P
P
P
P
P
P
P
P
P
P
P
P
P
P
1,
2
1,
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
1,
3

g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
g
1
99
By Galois correspondence, there are 10 intermediate subelds of K/Q. These
are all xed elds of the subgroups displayed above. Set H = 1, ,
2
,
3
.
Since [K : K
H
] = o(H) = 4 we see that [K
H
: Q] = 2. Since i is xed
by each element of H, we conclude that K
H
= Q(i). Set L = 1, . Since
[K : K
L
] = o(L) = 2, we see that [K
L
: Q] = 4. Since (a) = a, K
L
= Q(a).
Set M = 1, . Since [K : K
M
] = o(M) = 2, [K
M
: Q] = 4. The orbit
of a under the action of M is a, ia. Adding the elements of this orbit we
get b = a + ia. Hence a + ia K
M
. To nd g(x) = irr (b, Q), we nd all
the conjugates of b by applying the automorphisms in G. This way we see
that the orbit of b under the action of G is b, b, a ia, a + ia. Hence
deg
Q
(b) = 4. Hence K
M
= Q(b). The other xed elds can be found by
similarly.
Example 15.6. We discuss the Galois group of x
p
2, where p is an odd
prime, We will show that it is isomorphic to the group
G =
__
a b
0 1
_
: a, b F
p
and a ,= 0
_
.
Put = e
2i/p
and =
p

2. The roots of x
p
2 are , ,
2
, . . . ,
p1
.
Thus K = Spl(x
p
2, Q) = Q(, ) and [K : Q] = p(p 1). If G(K/Q),
then () =
i()
and () =
j()
, where 1 j() (p 1) and
i() = 0, 1, . . . , (p 1). Dene
: G(K/Q) G by () =
_
j() i()
0 1
_
.
Dene , G(K/Q) by
() =
a
, () =
b
, () =
c
, and () =
d
.
Therefore
() =
_
d c
0 1
_
, () =
_
b a
0 1
_
, ()() =
_
bd c +ad
0 1
_
.
Since
() = (
a
) =
c+ad
() = (
d
) =
bd
100
we have
() =
_
bd c +ad
0 1
_
= ()().
Therefore is a group homomorphism. As
Ker = : d = 1 and c = 0 = id,
we conclude that is an isomorphism.
Example 15.7. Let x
1
, x
2
, . . . , x
n
be indeterminates over a eld F. The
symmetric group S
n
acts on E = F(x
1
, x
2
, . . . , x
n
), the fraction eld of the
ring of polynomials F[x
1
, . . . , x
n
]. If S
n
then

: E E dened
by

(x
i
) = x
(i)
is an automorphism of E. If
1
,
2
S
n
then

2
=

2
. Thus G =

: S
n
is a group of automorphism of E and it is
isomorphic to S
n
. Let x be a variable over E and consider the polynomial
ring E[x]. Then
g(x) = (x x
1
)(x x
2
) (x x
n
) E[x]
= x
n

1
x
n1
+
2
x
n2
+ (1)
n

n
Where
i
s are the elementary symmetric functions of x
1
. . . , x
n
. The auto-
morphism

: E E can be extended to E[x] by xing x which we again

denote by

. Therefore

(g(x)) = (x x
(1)
)(x x
(2)
) (x x
(n)
) = g(x)
Thus

(
i
) =
i
for all i = 1, 2, . . . , n. Hence F(
1
,
2
, ,
n
) E
G
.
Notice that E = F(
1
, . . . ,
n
, x
1
, . . . x
n
). So E is a splitting eld of g(x)
over F(
1
, . . . ,
n
) and g(x) is separable. If G(E/F(
1
, . . . ,
n
)) then
permutes the roots of g(x), hence =

for some . Thus G =

G(E/F(
1
. . . ,
n
)). Therefore symmetric rational functions are rational
functions of symmetric functions.
101
Lecture 17 : Cyclotomic Extensions I
Objectives
(1) Roots of unity in a eld.
(2) Galois group of x
n
a over a eld having n
th
roots of unity.
(3) Irreduciblilty of the cyclotomic polynomial
n
(x) over Q.
(4) A recursive formula for
n
(x).
Keywords and phrases : Roots of unity, Galois group of x
n
a, cyclo-
tomic polynomials.
16. Cyclotomic Extensions
Roots of unity in any eld
Let F be a eld. A root z F of x
n
1 is called an n
th
root of unity in
F. Roots of unity play important role in algebra and number theory. Their
analysis led Gauss to his rst mathematical discovery: construction of a
regular polygon of 17 sides.
Suppose that char F = p and n = p
e
m where (m, p) = 1. Then x
n
1 =
(x
m
1)
p
e
. By the derivative criterion, x
m
1 is separable. Hence the
splitting eld of x
n
1 is equal to that of x
m
1. Therefore we consider
elds of chararacteristic zero or of characteristic p where (p, n) = 1.
Let k be a eld and (n, char k) = 1. Then x
n
1 is separable by the
derivative criterion. Let Z = z
1
, z
2
, . . . , z
n
be the set of its roots of in k
a
.
Then Z is a multiplicative subgroup of (k
a
)

. Hence it is cyclic. Any of

the (n) generators of Z is called a primitive n
th
root of unity. Let z
be any such generator. Then k(z) is a splitting eld of x
n
1 over k. Let

n
(x) = irr (z, Q). A primitive n
th
root of unity over Q is denoted by
n
.
Denition 16.1. A splitting eld of x
n
1 over F is called a cyclotomic
eld of order n over F.
102
Proposition 16.2. Let (char F, n) = 1 and f(x) = x
n
1 F[x]. Then G
f
is isomorphic to a subgroup of U(n). In particular G
f
is an abelian group
and o(G
f
) [ (n).
Proof. As f(x) is separable, it has n distinct roots. Let z
1
, z
2
, . . . , z
n
= Z
be the set of roots of f(x) in F
a
and E = F(z
1
, z
2
, . . . , z
n
). Since Z E

is
a subgroup, it is cyclic. The map : G(E/F) Aut(Z) such that [
Z
is an injective group homomorphism. Since Aut(Z) m [ (m, n) = 1 :=
U(n) is an abelian group, G(E/F) is also an abelian group whose order
divides (n).
Example 16.3. Let F = F
2
. Then x
3
1 = (x1)(x
2
+x+1). Any root z
of x
2
+x +1 is a primitive cube root of unity over F. Hence [F(z) : F] = 2.
To nd the degree of a primitive seventh root of unity over F, consider the
factorization of x
7
1 into irreducible polynomials over F :
x
7
1 = (x 1)(x
3
+x
2
+ 1)(x
3
+x + 1).
Therefore there are 6 primitive 7th roots of unity over F with two minimal
polynomials. In contrast to this, we shall see that all the primitive n
th
roots of unity over Q have the same irreducible polynomial called the n
th
cyclotomic polynomial
n
(x).
Proposition 16.4. Let x
n
a = f(x) F[x] and suppose F has n distinct
roots of x
n
1. Then G
f
is a cyclic group and o(G
f
) divides n.
Proof. Let Z = z
1
, z
2
, . . . , z
n
F be the set of roots of x
n
1. Let r be a
root of f(x) in a splitting eld E of f(x). Then rz
1
, rz
2
, . . . , rz
n
are roots of
f(x). Thus E = F(r). Let , G(E/F). Then (r) = z

r and (r) = z

r
for some z

, z

Z. Hence (r) = (z

r) = z

r. Dene
: G(E/F) Z such that () = z

.
Then is a group homomorphism. The map is clearly injective. Since Z
is a subgroup of F

, it is a cyclic group of order n. Hence [G

f
[ divides n
and G
f
is cyclic.

103
Theorem 16.5. (1)
n
(x) Z[x], (2) [Q(
n
) : Q] = (n) = deg
n
(x) and
(3) G(Q(
n
)/Q) U
n
.
Proof. Let
n
(x) = irr (
n
, Q). Then x
n
1 =
n
(x)h(x), where h(x) is
monic in Q[x]. By Gauss Lemma
n
(x), h(x) Z[x]. We show that each
primitive n
th
root of unity is a root of
n
(x). Let p be a prime number and
(p, n) = 1. Suppose
n
(
p
n
) ,= 0. Hence h(
p
n
) = 0. Hence
n
is a root of
h(x
p
). Thus
h(x
p
) =
n
(x)g(x) for some monic g(x) Z[x].
Reduce mod p to get
(

h(x))
p
=

n
(x) g(x) ,
where denotes reduction of coecients mod p. Hence

n
(x) and

h(x)
have a common root mod p. But x
n
1 has distinct roots over F
p
. Hence

p
n
is a root of
n
(x). Hence all primitive n
th
roots of unity are roots
of
n
(x). Since G = G(Q(
n
)/Q) is isomorphic to a subgroup of U(n),
[Q(
n
) : Q] = [G[ (n). But deg
n
(x) (n). Hence [G[ = (n). Hence
G U(n).
Computation of Cyclotomic Polynomials
Let
n
be a primitive n
th
root of unity. Then the other roots of
n
(x) are

i
n
such that (i, n) = 1. Thus

n
(x) =

(i,n)=1
(x
i
n
).
Since the roots of x
n
1 form a cyclic group of order n, the order of any
root divides n. Since
d
(x) =

o(z)=d
(x z), it follows that
x
n
1 =

d|n

d
(x).
Therefore

n
(x) =
x
n
1

d|n,d<n

d
(x)
.
104
This is a recursive formula for computation of
n
(x). First few cyclotomic
polynomials are:

1
(x) = x 1

2
(x) =
x
2
1

1
(x)
= x + 1

3
(x) =
x
3
1

1
(x)
= x
2
+x + 1

4
(x) =
x
4
1

1
(x)
2
(x)
= x
2
+ 1

5
(x) = x
4
+x
3
+x
2
+x + 1

6
(x) =
x
6
1

1
(x)
2
(x)
3
(x)
= x
2
x + 1
105
Lecture 18 : Cyclotomic Extensions II
Objectives
(1) Discriminant of
p
(x).
(2) Subelds of Q(
p
).
(3) Kronecker-Weber Theorem for quadratic extensions of Q.
(4) Algorithm for construction of primitive elements of subelds of Q(
p
).
(5) Subelds of Q(
7
), Q(
13
) and Q(
17
).
Keywords and phrases : Discriminant of
p
(x), Kronecker-Weber The-
orem, subelds of Q(
p
).
17. Subfields of Q(
p
)
A celebrated theorem of Kronecker and Weber states that a Galois extension
E of Q with abelian Galois group is contained in a cyclotomic extension (an
extension of Q obtained by adjoining roots of unity.) We will prove this
theorem for quadratic extensions of Q. For this purpose, we show that the
square root of the discriminant of
p
(x) is a primitive element of the unique
intermediate subeld of K of Q(
p
) so that [K : Q] = 2.
Lemma 17.1. Let p be an odd prime. Then disc (
p
(x)) = (1)
(
p
2
)
p
p2
.
Proof. Let
p
be a primitive p
th
root of unity. Since x
p
1 =
p
(x)(x 1),
and px
p1
=
p
(x) + (x 1)
p
(x), we have for each i = 1, 2, . . . , p 1,
p(
i
p
)
p1
= (
i
p
1)
p
(
i
p
).
Therefore
p1

i=1

p
(
i
p
) =
p1

i=1
p(
i
p
)
p1
/(
i
p
1)
=
p
p1

p1
i=1
(
i
p
1)
=
p
p1
(1)
p1

p
(1)
= p
p2
.
106
Using the formula for discriminant in terms of derivatives, we get
disc (
p
(x)) = (1)
(
p
2
)
p
p2

Proposition 17.2. The eld Q(

p
) contains a unique quadratic extension
of Q, namely
Q
__
disc (
p
(x))
_
= Q
__
(1)
(
p
2
)
p
_
which is real if p 1(mod 4) and complex if p 3(mod 4).
Proof. The Galois group Gof Q(
p
) over Qis cyclic of order p1. Hence there
is a unique subgroup of G having index 2. Thus there is a unique subeld
of Q(
p
) which is a quadratic extension of Q. As
_
disc (
p
(x)) Q(
p
)Q
it generates the unique quadratic subeld of Q(
p
).
Corollary 17.3. Every quadratic extension of Q is contained in a cyclo-
tomic extension.
Proof. If p 3(mod 4), then Q(

p) Q(
p
) and if p 1(mod 4) then
Q(

p) Q(
p
). A quadratic extension of Q is of the form Q(

d) where
d is a squarefree integer. Suppose d = p
1
p
2
. . . p
r
where p
1
, p
2
, . . . , p
r
are
distinct primes. Then Q(

d) Q(
p
1
,
p
2
, . . . ,
pr
, i).

Proposition 17.4. Let L Q(

p
) be a subeld with [Q(
p
) : L] = 2. Then
L = Q(
p
+
1
p
).
Proof. As
p
is a root of x
2
(
p
+
1
p
)x+1 = 0,
_
Q(
p
) : Q(
p
+
1
p
)

2.
Since L = Q(
p
+
1
p
) R, we conclude that [Q(
p
) : L] = 2.
Proposition 17.5. Let p be a prime number. Let be a primitive p
th
root of unity. Let H be a subgroup of G = G(Q()/Q) = U(p). Put
H
=

H
(). Then
E
H
= Q(
H
).
Proof. Let H. Since H is nite, H = [ H. Hence (
H
) =
H
for all H. Hence Q(
H
) Q()
H
. Let / H. We show that (
H
) ,=
H
.
107
The set
B = 1, ,
2
, . . . ,
p2

is a basis of the Q-vector space Q(). If (

H
) =
H
then = () for
some H. Hence = 1 and so
1
= . Therefore H which is
a contradiction. If Q(
H
) ,= Q()
H
, then by FTGT, there is a subgroup
M > H such that Q(
H
) = Q()
M
Q()
H
. Hence
H
is xed by an
element M H. This is a contradiction.

Example 17.6. Let p = 7 and

7
= w. Then
_
Q(w +w
1
) : Q

= 3 Let us
nd the irreducible polynomial of w + w
1
= w + w
6
. To do this nd the
orbit of w +w
6
under the action of the Galois group G = G(Q(w)/Q). G is
is generated by the automorphism (w) = w
2
. Hence The orbit of w + w
6
under the action of G is
1
= w+w
6
,
2
= w
2
+w
5
,
3
= w
4
+w
3
. Hence
irr (w +w
6
, Q) =
3

i=1
(x
i
) = x
3
+x
2
2x 1.
Example 17.7. Put
13
= . We list all subelds of E = Q() by using
the procedure in the proposition above. Since Galois group G of the Galois
extension E/Q is cyclic of oder 12 it has proper subgroups of orders 2, 3, 4,
and 6. The automorphism () =
2
generates G. The action of powers of
on is described in the table:
i 1 2 3 4 5 6 7 8 9 10 11

i
() =
2

12

11

10

7
The unique quadratic extension of Q in E is Q(

13). The unique subeld of

degree 6 is Q( +
12
). The subgroup H of oder 4 is generated by
3
. Hence
H =
3
,
6
,
9
, id. Hence a primitive element of the degree 3 extension of
Q in E is

H
= +
3
() +
6
() +
9
() = +
8
+
12
+
5
.
The subgroup K of G of order 3 is generated by
4
. Hence a primitive
element of Q()
K
is

K
= +
4
() +
8
() = +
3
+
9
.
Hence the poset of intermediate subelds of Q() is
108
E = Q()
o
o
o
o
o
o
o
o
o
o
o
O
O
O
O
O
O
O
O
O
O
O
E
K
= Q(
K
) Q( +
1
)
Q(

13)
O
O
O
O
O
O
O
O
O
O
O
O
O
E
H
= Q(
H
)
o
o
o
o
o
o
o
o
o
o
o
o
o
Q
Example 17.8. Let E be the splitting eld of x
17
1 over Q generated by
a primitive seventeenth root z of 1. So irr(z, Q) = x
16
+ x
15
+ + x + 1
and E = Q(z). Therefore [E : Q] = 16. Thus [G(E/Q)[ = [U(17)[ = 16.
The multiplicative group of units mod 17 can be generated by 3+(17). Thus
: z z
3
is a generator of G(E/Q) = ,
2
, . . . ,
16
= 1. The subgroups
of G and their orders are:
G = G
1
= G
2
=
2
G
3
=
4
G
4
=
8
id
[G
1
[ = 16, [G
2
[ = 8, [G
3
[ = 4, and [G
4
[ = 2.
The chain of intermediate subelds is:
E
G
= Q E
G
2
E
G
3
E
G
4
E.
We determine the generators for these xed elds. Note that
(z) = z
3
,
2
(z) = z
3
2
, . . . ,
i
(z) = z
3
i
.
Let
x
1
=
8

i=1
(
2
)
i
(z), y
1
=
4

i=1
(
4
)
i
(z) and z
1
=
2

i=1
(
8
)
i
(z).
The xed elds are
E
G
2
= Q(x
1
) E
G
3
= Q(x
1
, y
1
) E
G
4
= Q(x
1
, y
1
, z
1
).
110
Lecture 19 : Abelian and Cyclic Extensions
Objectives
(1) Innitude of primes p 1 ( mod n ).
(2) Inverse Galois problem for nite abelian groups.
(3) Structure of some cyclic extensions.
Keywords and phrases : Primes of the form p 1 ( mod n ), abelian
extension, cyclic extension, inverse Galois problem.
18. The Inverse Galois Problem for Finite Abelian Groups
A Galois extension E/F is called abelian (resp. cyclic) if G(E/F) is
abelian (resp. cyclic). In this section we will show that any nite abelian
group is the Galois group of a Galois extension of Q. In other words, any
nite abelian group is the Galois group of a polynomial with rational coef-
cients. A proof of this theorem requires the theorem from number theory
that there are innitely many primes p 1( mod n). We shall prove this
using cyclotomic polynomials. This is a special case of Dirichlets theorem
about innitude of primes in the arithmetic progression a+nb where a, b are
coprime natural numbers and n = 1, 2, 3, . . . . We will also construct cyclic
extensions of elds having enough roots of unity.
Lemma 18.1. Let p be a prime number and n be relatively prime to p. Let

n
(x) have a root in F
p
. Then p 1 (mod n).
Proof. Let k Z,

k F
p
and

n
(

k) = 0. Then p |
n
(k). Hence p | k
n
1.
Thus k
n
1(mod p). We claim that o(

k) = n in the group (F
p
)

. Suppose
o(

k) = m < n. Then

k
m
= 1. Hence
x
n
1 =

d|n

d
(x) =
n
(x)

d<n

d
(x)
=
n
(x)

d|m

d
(x)h(x)
=
n
(x)(x
m
1)h(x)
111
Hence

k
n
1 =
n
(

k)(

k
m
1)h(

k). This means x

n
1 has a multiple root
in F
p
. This is a contradiction. Hence o(

k) = n. Hence n | p 1. Thus
p 1 mod n.
Theorem 18.2. There are innitely many primes p 1 (mod n).
Proof. Suppose to the contrary, p
1
, p
2
, . . . , p
g
are all such primes. Let m =
np
1
p
2
. . . p
g
. Since
m
(x) Z[x], is monic, lim
x

m
(mx) = . Hence
there exists k such that
m
(mk) 2. Let p be a prime factor of
m
(mk).
Then p | (mk)
m
1. Hence p does not divide mk. Hence (p, n) = 1 and
p = p
1
, . . . , p
n
. Moreover

m
(mk) = 0. Hence p 1(mod n). This is a
contradiction.
Theorem 18.3. Let G be a nite abelian group. Then there is a Galois
extension K/Q such that G(K/Q) = G.
Proof. We may assume that |G| 2. Then G C
n
1
C
n
k
. where
|G| = n = n
1
n
2
. . . n
k
and n
1
|n
2
| |n
k
. There exist innitely many primes
p
i
1 (mod n
i
) for i = 1, 2, . . . , k. We can nd subgroups H
1
< U(p
1
), H
2
<
U(p
2
), . . . , H
k
< U(p
k
) such that
U(p
1
)
H
1
C
n
1
,
U(p
2
)
H
2
C
n
2
, . . . ,
U(p
k
)
H
k
C
n
k
.
U(p
1
) U(p
2
) U(p
k
)
H
1
H
2
H
k
C
n
1
C
n
k
.
Let H < U(n) and H H
1
H
2
H
k
. Then
U(n)
H
G. By the FTGT
G(Q(
n
)
H
/Q) =
U(n)
H
G.

19. Cyclic Galois Extensions

In this section we discuss cyclic extensions of degree n if F has a primitive
n
th
root of unity or when F has characteristic p > 0 and E/F has degree p.
There is no simple description of cyclic extensions of Q or elds devoid of
roots of unity. We need a theorem of Dedekind about linear independence
over K of automorphims of a eld K.
112
Denition 19.1. Let G be a group and K a eld. By a character of
G in K we mean a homomorphism : G K

. We say that characters

1
,
2
, . . . ,
n
: G K

are linearly independent if for a

1
, . . . , a
n
K
a
1

1
+a
2

2
+ +a
n

n
= 0 if and only if a
i
= 0 for i = 1, 2, . . . , n.
Theorem 19.2 (Dedekind). Let
1
,
2
, . . . ,
n
be distinct characters of a
group G in a eld K. Then
1
,
2
, . . . ,
n
are linearly independent.
Proof. Apply induction on n. If n = 1, then
1
: G K

is clearly linearly
independent. Let n 2. Let n be the smallest positive integer such that
there exist a
1
, . . . , a
n
K, not all zero with
a
1

1
+ +a
n

n
= 0. (2)
Then a
i
= 0, for all i. Since
1
=
2
, there exists z G such that
1
(z) =

2
(z). Hence for all x G,
a
1

1
(xz) +a
2

2
(xz) + +a
n

n
(xz) = 0 (3)
a
1

1
(z)
1
+a
2

2
(z)
2
+ +a
n

n
(z)
n
= 0. (4)
Multiply (2) by
1
(z) and subtract (4) to get the relation :
(
1
(z)
2
(z))a
2

2
+(
1
(z)
3
(z))a
3

3
+ +(
1
(z)
n
(z))a
n

n
= 0.
The above relation has smaller length, which is a contradiction.

Lemma 19.3. Let F be a eld containing a primitive n

th
root of unity .
Suppose that E/F is a Galois extension of degree n and G = G(E/F) = ().
Then is an eigenvalue of .
Proof. The eld E is an n-dimensional F-vector space. Since has order n,
satises x
n
1 = 0. If is a root of a polynomial f(x) F[x] of degree
m < n then ,
2
, . . . ,
m
are linearly dependent over F. This contradicts
Dedekinds Theorem. Hence the minimal and the characteristic polynomials
of are equal to x
n
1. Hence is an eigenvalue of .

We now describe the structure of cyclic extensions of degree n over a eld

having a primitive n
th
root of unity.
113
Theorem 19.4. Let E/F be a cyclic extension of degree n with G =
G(E/F) = () and let F be a primitive n
th
root of unity. Then there
exists a b F so that E = F(a) where a
n
= b.
Proof. Since is an eigenvalue of , there exists an eigenvector a E

so
that (a) = a. Hence
i
(a) =
i
a for all i = 1, 2, . . . , n. Hence a has at
least n conjugates in E. As E/F is a Galois extension of degree n, and E
contains a splitting eld of f(x) = irr (a, F), it follows that E = F(a) and
a
n
F since (a
n
) =
n
a
n
= a
n
.
Intermediate subelds of a cyclic Galois extension
Let E/F be a cyclic Galois extension of degree n where F has a primitive n
th
root of unity. We have proved that E = F(a) where a
n
F. The number of
subgroups of the Galois group G = G(E/F) is d(n), the number of divisors
of n. Each of these subgroups is cyclic. Hence there are d(n) intermediate
subelds of E/F. We show that they are F(a
d
) where d is a divisor of n.
Proposition 19.5. Let E/F be a cyclic Galois extension of degree n where
F has a primitive n
th
root of unity. Let E = F(a) where a
n
F. Then The
intermediate subelds of E/F are F(a
d
) where d is a divisor of n.
Proof. The Galois group G has unique subgroup of order d for every divisor
d of n. Hence E/F has a unique subeld of degree d for each divisor d of n.
Consider the subeld K = F(a
d
). Then a is a root of x
d
a
d
K[x]. Thus
[E : F(a
d
)] d. Since a
n
F, we have (a
d
)
n/d
F. Hence [F(a
d
) : F] n/d.
It follows that [E : F(a
d
)] = d. Hence the intermediate subelds of E/F are
F(a
d
) where d varies over the divisors of n.

114
Lecture 20 : Cyclic Extensions and Solvable Groups
Objectives
(1) Cyclic extensions of degree p over elds with characteristics p.
(2) Solvable groups.
(3) Simplicity of S
n
and A
n
.
Keywords and phrases : Cyclic extensions, solvable groups, commuta-
tor, simple groups.
Structure of cyclic Galois extensions over elds of characteristic p
Let F be a eld of positive characteristic p. We discuss the structure of
Galois extensions of F of degree p. Consider the map : F F dened by
(a) = a
p
a. Then is a homomorphism of the additive group F. Moreover
F
p
= Ker . Note that
1
(a) = {a +i | i = 0, 1, . . . , p 1}.
Theorem 19.6 (Artin-Schreier). (1) Let E/F be a cyclic Galois exten-
sion of degree p where char F = p, a prime number. Then E = F(a) where
a is a root of x
p
x b for some b F.
(2) Suppose that a / F
p
F. Then f(x) = x
p
x a is irreducible over F
and a splitting eld of f(x) over F is cyclic of degree p.
Proof. (1) Let G = G(E/F) = () and let T : E E be the linear map of
the F-vector space E dened by T(a) = (a) a. Then
Ker T = {a E | (a) = a} = F.
Since T
p
= ( id)
p
=
p
id = 0, we have Im (T
p1
) Ker T = F. If
T
p1
= 0 then there is a nontrivial F-linear relation among
p1
,
p2
, . . . , , id.
This contradicts Dedekinds theorem. Hence Im T
p1
= Ker T = F. Let
b E so that T
p1
(b) = 1. Set = T
p2
(b). Then T() = () = 1.
Hence () = +1. Thus
i
() = +i for all i = 1, 2, . . . , p 1. Therefore
E = F().
Since (
p
) = ( + 1)
p
( + 1) =
p
, The element a =
p

E
G
= F. Hence is a root of x
p
x a.
115
(2) Conversely, suppose that a / F
p
F. Then we show that f(x) = x
p
xa
is irreducible over F. If is a root of f(x) then +i is a root of f(x) for all
i = 1, 2, . . . , p 1. Hence E = F() is a splitting eld of f(x). If we assume
that f(x) is irreducible over F then [E : F] = p and the Galois group is
generated by the automorphims () = + 1.
Suppose that f(x) = g
1
(x)g
2
(x) . . . g
n
(x) where each g
i
is irreducible over
F. If is a root of g
i
then E = F() as shown above. Hence each g
i
(x) has
same degree r and so deg f(x) = p = rn. Thus r = p and n = 1. Hence f(x)
is irreducible over F.

20. Solvable groups

Denition 20.1. Let G be a group. A sequence of subgroups
G = G
0
G
1
G
2
G
s+1
= 1 (5)
is called a normal series for G if G
i
is a normal subgroup of G
i1
for
i = 1, 2, . . . , s. The normal series (5) is called abelian (resp. cyclic ) if
the quotients G
i
/G
i+1
are abelian (resp. cyclic) for i = 0, 1, . . . s. A group
having an abelian series is called a solvable group.
Example 20.2. (1) Any abelian group is solvable.
(2) The group S
3
is solvable since S
3
A
3
1 is an abelian series.
(4) The group S
4
is solvable since
S
4
A
4
V
4
1
is an abelian series where V
4
= {(1), (12)(34), (13)(24), (14)(23)}.
Proposition 20.3. Any group G of order p
n
where p is a prime number is
solvable.
Proof. Apply induction on n. If n = 1 then G is cyclic and hence solvable.
Let n 2. Let C be the center of G. We know that o(C) > 1. Hence
o(G/C) < o(G). By induction, we have an abelian series
G/C G
1
/C G
2
/C G
s
/C = 1
Since (G
i
/C)/(G
i+1
/C) G
i
/G
i+1
for all i, we have an abelian series:
G G
1
G
2
G
s
C 1.
116
Thus G is solvable.

Denition 20.4. Let G be a group. The commutator [g, h] of g, h G

is dened as [g, h] = g
1
h
1
gh. The derived subgroup of G denoted by
G

is the subgroup generated by all the commutators in G. The k

th
derived
subgroup of G is dened inductively as G
(k)
= (G
(k1)
)

.
Proposition 20.5. Let f : G H be a homomorphism of groups.
(1) f(G

) H

. If f is onto then f(G

) = H

.
(2) If K G then K

G. In particular G

G.
(3) If K G then G/K is abelian if and only if G

K.
Proof. (1). Let g, h G. Then f([g, h]) = f(g)
1
f(h)
1
f(g)f(h) = [f(g), f(h)].
Hence f(G

) H

. It is clear that equality holds true if f is onto.

(2) Let a G. The inner automorphism i
a
: G G restricts to an automor-
phism of K as K G. Hence i
a
(K

) = K

. Therefore K

G. Since GG,
we have G

G.
(3) Let K G. Then G/K is abelian for all g, h G, ghK = hgK
h
1
g
1
hg K for all g, h G G

K.
Proposition 20.6. A group G is solvable if and only if G
(s)
= 1 for some
s N.
Proof. Let G be solvable. Then there is an abelian series for G
1 G
1
G
2
G
s
= G.
We show by induction on s that G
(s)
= 1. If s = 1, then G is abelian. Hence
[g, h] = 1 for all g, h G. Hence G

= 1. Now let s > 1. Then

1 G
1
G
2
G
s1
is an abelian series for G
s1
. Hence G
(s1)
s1
= 1. Since G/G
s1
is abelian,
G

G
s1
. Hence
G
(s)
= (G

)
(s1)
G
(s1)
s1
= 1.
Conversely suppose that G
(s)
= 1 for some s. Then
GG
(1)
G
(2)
G
(s)
= 1
is an abelian series for G. Thus G is solvable.
117
Proposition 20.7. Let G be a group and H be a subgroup.
(1) If G is solvable then so is H.
(2) If f : G H is a surjective group homomorphism and G is solvable
then H is so.
(3) If K G and G/K are solvable then G is solvable.
Proof. (1) If G is solvable then G
(s)
= 1 for some s. Since H
(s)
G
(s)
= 1,
we have H
(s)
= 1. Thus H is solvable.
(2) Let G
(s)
= 1. Since f is surjective, f(G
(s)
) = H
(s)
= 1. Hence H is
solvable.
(3) Let K G and K and G/K be solvable. Then there exist s and t such
that K
(s)
= 1 and (G/K)
(t)
= 1. Hence G
(t)
K. Thus G
(t+s)
K
(s)
= 1.
Hence G is solvable.

Lemma 20.8. The group A

n
is generated by 3-cycles. If n 5 then all
3-cycles are conjugates in A
n
.
Proof. Let be an even permutation. Let (ij)(rs) occur in a decomposi-
tion of as a product of transpositions. If (ij) and (rs) are disjoint then
(ij)(rs) = (ijr)(rsj). If j = r then (ir)(rs) = (irs). Hence every even per-
mutation is a product of 3-cycles. Now suppose that n 5. Let be any
permutation and (j
1
j
2
. . . j
p
) be a p-cycle. Then
(j
1
j
2
. . . j
p
)
1
= ((j
1
)(j
2
) . . . (j
p
)).
Let (ijk) and (rst) be any two 3-cycles. Dene by (i) = r, (j) =
s, (k) = t and let (u) = u for any u = i, j, k. Then
(ijk)
1
= ((i)(j)(k)) = (rst).
If is odd then put = (ij). Then is even and
(ijk)
1
= (ij)(ijk)
1
(ij) = (rst).

Theorem 20.9. The groups S

n
and A
n
are not solvable for n 5.
118
Proof. Since S
n
/A
n
is abelian, S

n
A
n
. Note that since n 5, any 3-cycle
is a commutator in view of :
[(jkv), (ikr)] = (vkj)(rki)(jkv)(ikr) = (vkj)(jiv) = (ikj).
Therefore S

n
= A

n
= A
n
. Thus S
(s)
n
= A
(s)
n
= A
n
for all s. Hence A
n
and
S
n
are not solvable for n 5.
Theorem 20.10 (Galois). The alternating group A
n
is simple for n 5.
Proof. (S. Lang) Suppose A
n
is not simple for n 5. Let N be a proper
normal subgroup of A
n
for some n 5. Let = 1 be a permutation in N
that has maximum number of xed points. We say that j is a xed point of
if (j) = j. Consider a decomposition of as a product of disjoint cycles
of length at least two: =
1

2
. . .
g
. Suppose the length of each
j
is two.
Since is an even permutation, g 2. Suppose that = (ij)(rs)
3
. . .
g
.
Let k be dierent from i, j, r, s and set = (rsk). Consider the commutator
= [, ] =
1

1
. Then 1 = N. Moreover (i) = i and (j) = j.
This is a contradiction since has maximum number of xed points among
the permutations in N \ {1}.
Now suppose that for some a,
a
= (ijk...) has length at least 3. If = (ijk)
then N has a 3-cycle and hence N = A
n
. If is not a 3-cycle then must
move at least two other elements r, s besides i, j, k. Put = (rsk) and
consider = [, ]. Then 1 = N. Moreover (j) = j and xes all the
elements that xes. This is a contradiction.
123
Lecture 21 : Galois Groups of Composite Extensions
Objectives
(1) Galois group of composite extensions
(2) Galois closure of a separable eld extension.
Keywords and Phrases : Composite extensions, normal closure.
21. Galois groups of composite extensions
Let K be a eld and K be an algebraic closure of K. Let E, F be intermediate
subelds of K/K. Recall that the compositum of E and F denoted by
EF is the smallest subeld of K containing E and F. In this section we
will discuss Galois groups of composite extensions and normal closure of an
algebraic eld extension.
Proposition 21.1. If E/K is a Galois extension and F/K is a eld exten-
sion, then EF/F is Galois. If F/K is Galois then EF/K and EF/K are
Galois.
Proof.
K
EF
F
F
F
F
F
F
F
F
F
x
x
x
x
x
x
x
x
x
E
F
F
F
F
F
F
F
F
F
F
x
x
x
x
x
x
x
x
x
E F
K
Consider the diagram above. As E/K is a separable and normal extension,
it is a splitting eld of a family f
i
(x) of separable polynomials over K.
Then EF/F is the splitting eld of the same family of polynomials. Hence
124
EF/F is a Galois extension. If F/K is Galois then it is a splitting eld of
a family of polynomials g
j
(x) over K. Hence EF/K is a splitting eld of
the polynomials f
i
(x) g
j
(x). Hence EF/K is Galois.
Now we show that if E/K and F/K are Galois then E F/K are Galois.
Let : E F K be a K-embedding. Let : EF K be an extension of
. Then (E) = E and (F) = F since E/K and F/K are Galois. Therefore
(E F) E F. Since E F/K is algebraic, (E F) = E F. Hence
E F/K is a Galois extension.

Proposition 21.2. Let E/K be a Galois extension and F/K be a eld

extension so that E, F K. Then the map : G(EF/F) G(E/K)
dened by () = [
E
is injective and it induces an isomorphism:
G(EF/F) G(E/E F).
Proof. Since is an F-automorphism of EF, it is also a K-automorphism.
Hence [
E
G(E/K). If [
E
= id
E
then = id
EF
. Hence is an injective
group homomorphism.
The image of is a subgroup H of G(E/K). By Artins TheoremG(E/E
H
) =
H. Hence E F E
H
. Let a E (E F). Then a EF F. Hence there
is a G(EF/F) so that (a) ,= a. Hence a / E
H
. Therefore E
H
= E F
and we conclude that G(E/E F) = H G(EF/F).
Corollary 21.3. Let E/K be a nite Galois extension and F as above.
Then
[EF : F] = [E : E F].
In particular, [EF : K] = [E : K][F : K] if and only if E F = K.
Proof. Since G(EF/F) G(E/E F), we obtain
[G(EF/F)[ = [EF : F] = [G(E/E F)[ = [E : E F].
Therefore we have:
[EF : K] = [E : E F][F : K] =
[E : K][F : K]
[E F : K]
.
The conclusion follows from the equation above.
125
Theorem 21.4. Let E/K and F/K be nite Galois extensions so that
E, F K. Then the homomorphism
: G(EF/K) G(E/K) G(F/K), () = ([
E
, [
F
)
is injective. If E F = K then is an isomorphism.
Proof. It is clear that is a group homomorphism. The kernel of consists
of G(EF/K) so that (a) = a for all a E and for all a F. Hence
such = id
EF
. Thus is injective.
Suppose that E F = K. Then by Corollary 21.3,
[G(EF/K)[ = [EF : E F] = [F : K][E : K] = [G(E/K)[[G(F/K)[.
This shows that is an isomorphism.
The Normal Closure of an Algebraic Extension
Let K/F be an algebraic extension and K F. The normal closure of
K/F in K is the splitting eld N over F of the polynomials irr (a, F) [ a
K. It is clear that N is a normal extension of F containing K. Moreover
any normal extension N

F of F containing K must contain the splitting

elds of irr (a, F) [ a K. Hence N = N

. If K = F(a
1
, a
2
, . . . , a
n
) then
N is the splitting eld of the polynomials irr (a
i
, F) for all i = 1, 2, . . . , n.
If K/F is separable then N/F is a separable extension as it is obtained by
adjoining roots of separable polynomials over F. Hence the normal closure
of K/F when K/F is separable, is a Galois extension.
Let K/F be a separable extension that is not normal. Let N be a normal
closure of K/F. Put H = G(N/K). Then K = N
H
. Let H

< H be a normal
subgroup of G = G(N/F). Then N
H

> N
H
= K and N
H

/F is a normal
extension of K. Thus N
H

= N by minimality of N. Hence H

= (id).
126
Lecture 22 : Solvability by Radicals
Objectives
(1) Radical extensions.
(2) Solvability by radicals and solvable Galois groups.
(3) A quintic polynomial which is not solvable by radicals.
Keywords and Phrases : Radical extensions, solvable Galois groups,
insolvable quintic.
22. Solvability by radicals
Let F be a eld and f(x) F[x]. If there is a formula for the roots of f(x)
which involves the eld operations and extraction of roots, then we say f(x)
is solvable by radicals over F. This can be made precise in eld theory by
introducing the notion of a radical extension.
Denition 22.1. A eld extension K/F is called a simple radical exten-
sion of F if K = F(a) where a
n
F for some positive integer n. We say
that K/F is a radical extension if there is a sequence of eld extensions
F = F
0
F
1
F
2
F
n
= K
such that each F
i
is a simple radical extension of F
i1
for 1 = 1, 2, . . . , n. A
polynomial f(x) F[x] is called solvable by radicals over F if a splitting
eld of f(x) over F is contained in a radical extension of F.
Proposition 22.2. Let E/F be a separable radical extension. Let L E
be the smallest Galois extension of F so that L F
a
. Then L is a radical
extension of F.
Proof. Since E/F is separable and [E : F] = n, there are n F-embedding of
E into F
a
:

1
,
2
, . . . ,
n
: E F
a
.
Then L =
1
(E)
2
(E)
n
(E) is the smallest Galois extension of F con-
taining E. Indeed, let E = F(a). Then the roots of f
a
(x) = irr (a, F)
in F
a
are
i
(a) for i = 1, 2, . . . , n. Hence L = F(
1
(a),
2
(a), . . . ,
n
(a)) is
127
the splitting eld of f
a
(x) over F. Since
i
(E) E,
i
(E)/F is a radical
extension for each i = 1, 2, . . . , n. Hence L/F is a radical extension.
Theorem 22.3. Suppose char F = 0. If f(x) F[x] is solvable by radicals
then G
f
is a solvable group.
Proof. Let F = F
0
F
1
. . . F
r
= E be a sequence of simple radical
extension with F
i
= F
i1
(a
i
) such that a
n
i
i
F
i1
, i = 1, 2, . . . , r and E
contains a splitting eld K of f(x) over F. We may assume E/F is Galois
by the above proposition. Let n = n
1
n
2
. . . n
r
and M be the splitting eld
of x
n
1 over E.
E = F
r
E(w) = M
r
= M (e) = H
r
F
r1

F
r1
(w) = M
r1

G(M
r
/M
r1
) = H
r1

F
F
F
F
F
F
F
F
F
E = F
1
F
1
(w) = M
1
G(M
1
/M
1
) = H
1
F = F
0
F
0
(w) = M
0
G(M
r
/M
0
) = H
0
Let w be a primitive n
th
root of unity. Then F(w) has primitive n
th
i
root of
unity for i = 1, 2, . . . , r. Since E/F is Galois, E is a splitting eld of some
polynomial g(x) over F. Then M is a splitting eld of (x
n
1)g(x) over F.
Thus M is Galois over F. By the FTGT, G(K/F) G(M/F)/G(M/K).
Hence it is enough to prove that G(M/F) is solvable.
Each M
i
/M
i1
is a Galois extension. Hence H
i
H
i1
for i = 1, 2, . . . r.
Moreover
H
i1
/H
i
G(M
i
/M
i1
).
Since M
i
= M
i1
(a
i
) where a
n
i
i
M
i1
and M
i1
has a primitive n
th
i
root
of unity, the group H
i1
/H
i
is cyclic. Thus G(M
r
/F) is a solvable group.
Hence G
f
is a solvable group.
We will now construct a quintic f(x) Z[x] which is not solvable by radicals.
128
Proposition 22.4. A subgroup of S
5
containing a 5-cycle and a transposi-
tion is S
5
.
Proof. By renumbering we may assume G contains = (12) and =
(12345). Then G has (12)
1
= (23), (23)
1
= (34), (34)
1
= (45).
It is easy to show that (12), (23), (34), (45) = S
5
.
Any irreducible quintic f(x) Q[x] which has exactly 3 real roots is the
polynomial we are looking for. G
f
has an element of order 5 and the con-
jugation automorphism gives an element of order 2 in G
f
. The polynomial
x(x
2
4)(x
2
+ 4) = x
5
16x = g(x) has exactly 3 real roots 0, 2, 2. Since
g(1) = 15, g(1) = 15, f(x) = g(x) +2 = x
5
16x +2 have exactly 3 real
roots and it is irreducible over Q, Thus f(x) = 0 is not solvable by radicals
over Q.
Theorem 22.5 (Galois). Suppose F is a eld of characteristic zero and
f(x) F[x]. If G
f
solvable then f(x) is solvable by radicals over F.
L = K(w) (e) = H
k
= G(L/E)
E
k1
= L
H
k1

H
k1

E
1
= L
H
1
H
1
E = L
H
0
= F(w) H
0
= G(L/E)
Proof. Let K be a splitting eld of f(x) over F and [K : F] = n. Let L be
a splitting eld of x
n
1 over K and w be a primitive n
th
root of unity over
K. Then L = K(w). Put E = F(w). Then L is a splitting eld of f(x) over
E. Since H = G(L/E) embeds into G(K/F) H is also a solvable group. It
is enough to show f(x) is solvable by radicals over E. Consider an abelian
series for H.
H = H
0
H
1
H
k
= (1)
By rening this we may assume H
i
/H
i+1
is cyclic of order p
i+1
for i =
129
0, 1, . . . , k 1 where p
1
, p
2
, . . . , p
k
are primes numbers. Let E
i
= L
H
i
for
1, 2 . . . k. Then [E
i
: E
i1
] = [H
i1
/H
i
[ = p
i
. Since E
i1
has a primitive p
th
i
root of unity for i = 1, 2, . . . , k, E
i
/E
i1
is a simple radical extension. Hence
L/F is a radical extension. Thus f(x) is solvable by radicals over F.
Example 22.6. In this example we show that a splitting eld of E over
a eld F of a polynomial f(x) F[x] solvable by radicals need not be a
radical extension of F. Consider the polynomial f(x) = x
3
3x + 1 Q[x].
Let E be a splitting eld of f(x) over Q. We argue that E is not a radical
extension of Q. Reducing f(x) modulo 2, we see that the reduced polynomial
has no root in F
2
. Hence f(x) is irreducible over Q. The discriminant of f(x)
is 81. Hence G
f
= A
3
and therefore f(x) is solvable by radicals by Galois
theorem. Suppose that E/Q is a radical extension. Since [E : Q] = 3, there
is no proper intermediate subeld of E/Q. So E = Q(a) where a
n
Q, for
some n. Let g(x) = irr (a, Q). Then E is a splitting eld of g(x). Moreover
g(x) [ x
n
a
n
. Hence any root r of g(x) satises r
n
= a
n
. Since f(x) is a real
root, we may assume that E = Q(r). Hence r/a is a real n
th
root of unity.
Hence r = a. Hence g(x) has only two roots. This is a contradiction as
g(x) is a separable cubic polynomial.
131
Lecture 23 : Solutions of Cubic and Quartic Equations
Objectives
(1) Cardanos method for roots of cubic equations.
(2) Lagranges method for roots of quartic equations.
(3) Ferraris method for roots of quartic equations.
Keywords and Phrases : Cubic equations, quartic equations.
23. Solutions of cubic and quartic equations
In this section we present algorithms for nding roots of cubic and quartic
polynomials over any eld F of characteristic dierent from 2 and 3. This
is to make sure that irreducible cubics and quartics are separable.
Cubic polynomials
Cardano published Tartaglias method to nd roots of cubic polynomials
in 1545. This is known as Cardanos method. We may assume that the
given cubic is of the form f(x) = x
3
+ px + q since a general cubic can be
transformed into this form without changing its splitting eld. One begins
by introducing two unknowns u and v. Put x = u +v into f(x) = 0 to get
u
3
+v
3
+ 3u
2
v + 3uv
2
+p(u +v) +q = u
3
+v
3
+q + (3uv +p)(u +v) = 0.
We set u
3
+ v
3
+ q = 0 and 3uv + p = 0. Hence v = p/3u. Put this into
the rst equation to get
u
6
+qu
3
p
3
/27 = 0.
This is a quadratic equation in u
3
. Put D = (4p
3
+27q
2
). By the quadratic
formula we get
u
3
=
q

q
2
+ (4p
3
/27)
2
=
q
2

D/108.
Set A = q/2 +

D/108 and B = q/2

D/108. By symmetry of u
and v, we set u
3
= A and v
3
= B. Let be a primitive cube root of unity.
Then
u =
3

A,
3

A,
2
3

A, and v =
3

B,
3

B,
2
3

B.
132
We must choose cube roots of A and B in such a way that
3

A
3

B = p/3.
Having chosen these we see that the three roots of f(x) are
3

A+
3

B,
3

A+
2
3

B,
2
3

A+
3

B.
Example 23.1. Consider the cubic f(x) = x
3
3x+1. Reducing modulo 2,
we see that f(x) is irreducible over Q. The discriminant of f(x) is D = 81.
Hence
A = q/2 +

D/108 = exp(2i/3), and B = exp(2i/3).

Substitute these values of A and B into the formula for the roots, we see
that the three roots of f(x) are 2 cos(2/9), 2 cos(8/9) and 2 cos(14/9).
Let f(x) = x
3
+px +q R[x]. If disc (f) < 0, then cube roots of A and
B can be chosen to be real. In this case
r
1
=
3

A+
3

B R,
r
2
=
3

A+
3

B
2
+i

A
3

B
2

,
r
3
= r
2
.
If D = disc (f(x)) > 0 then A = q/2 + i

D/108 and B = A. Suppose

that
3

A = a +ib then due to uv = p/3 we have

3

B = a ib. Hence the

roots of f(x) are r
1
= 2a, r
2
= a b

3 and r
3
= a +b

3.
Notice that in this case, all the roots are real. However, they are expressed
in terms of complex numbers. It can be proved that the roots cannot be ex-
presseed in terms of real radicals. Historically, this is called the irreducible
case. This fact forced mathematicians to accept complex numbers as a valid
mathematical constructs.
Quartic polynomials
We present Lagranges method for the roots of a quartic polynomials. We
continue with the assumption that F has characteristics dierent from 2, 3.
Consider a general quartic polynomial f(x) = x
4
+ax
3
+bx
2
+cx+d. We put
y = x a/4 to get the polynomial g(y) = y
4
+py
2
+qy +r. Let r
1
, r
2
, r
3
, r
4
be roots of g(y). Consider the quantities

1
= (r
1
+r
2
)(r
3
+r
4
),
2
= (r
1
+r
3
)(r
2
+r
4
),
3
= (r
1
+r
4
)(r
2
+r
3
).
133
The cubic polynomial whose roots are
1
,
2
and
3
is called the resolvent
cubic of the quartic polynomial. It turns out to be the polynomial
h(x) = x
3
2px
2
+ (p
2
4r)x +q
2
.
Using the relation r
1
+r
2
+r
3
+r
4
= 0 we get
r
1
+r
2
=

1
r
3
+r
4
=

1
r
1
+r
3
=

2
r
2
+r
4
=

2
r
1
+r
4
=

3
r
2
+r
3
=

3
One can show that

3
= q. Hence two of the square roots
determine the third. Adding the three equations on the left and using the
fact that r
1
+r
2
+r
3
+r
4
= 0, we get
2r
1
=

1
+

2
+

3
.
2r
2
=

3
.
2r
3
=

1
+

3
.
2r
4
=

2
+

3
.
This shows that the roots of the resolvent cubic determine the roots of the
quartic.
Proposition 23.2. The discriminant of the quartic g(y) = y
4
+py
2
+qy +r
and its resolvent cubic h(x) = x
3
2px
2
+ (p
2
4r)x +q
2
are equal.
Proof. The dierences of the roots of the resolvent cubic are:

2
= (r
2
r
3
)(r
4
r
1
),
1

3
= (r
2
r
4
)(r
3
r
1
),
2

3
= (r
3
r
4
)(r
2
r
1
).
Hence the quartic and the resolvent cubic have same discriminant.
Remark 23.3. In the literature, we nd that the term resolvent cubic is
also used for the cubic whose roots are
t
1
= r
1
r
2
+r
3
r
4
, t
2
= r
1
r
3
+r
2
r
4
, and t
3
= r
1
r
4
+r
2
r
3
.
It can be shown that this cubic is r(x) = x
3
px
2
4rx+4pr q
2
and h(x)
and r(x) have equal discriminant and the same splitting eld over F.
134
Ferraris method for solving quartic equations
Consider the general quartic equation
x
4
+bx
3
+cx
2
+dx +e = 0.
Rewrite this as x
4
+bx
3
= cx
2
dx e. Now complete the square to get

x
2
+
bx
2

2
=

b
2
4
c

x
2
dx e.
Let y be another variable and consider the equation:

x
2
+
bx
2
+
y
2

2
=

b
2
4
c

x
2
dx e +y

x
2
+
bx
2

+
y
2
4
= x
2

b
2
4
c +y

+x

by
2
d

+
y
2
4
e (6)
The right hand side of the last equation is a square of a linear polynomial
in x if and only if its discriminant is zero. i.e.
(
1
2
by d)
2
4(
1
4
y
2
e)(
1
4
b
2
c +y) = 0.
Therefore
y
3
cy
2
+ (bd 4e)y b
2
e + 4ce d
2
= 0.
Let y be any root of this cubic and substitute it in the equation (6) to get
x
2
+
1
2
bx +
1
2
y = mx +n (7)
Notice that the roots of the equation (7) are the roots of the given quartic.
Proposition 23.4. Let x
1
, x
2
, x
3
, x
4
be the roots of
f = x
4
+bx
3
+cx
2
+dx +e = 0.
Then y
1
= x
1
x
2
+ x
3
x
4
, y
2
= x
1
x
3
+ x
2
x
4
, y
3
= x
1
x
4
+ x
2
x
3
are roots of
resolvent cubic g(y) = y
3
cy
2
+ (bd 4e)y b
2
e + 4ce d
2
.
135
Lecture 24 : Galois Groups of Quartic Polynomials
Objectives
(1) Galois group as a group of permutations.
(2) Irreducibility and transitivity.
(3) Galois groups of quartics.
Keywords and phrases : Transitive subgroups of S
4
Galois groups of
quartics, irreducibilty and transitivity.
24. Galois Groups of Quartic Polynomials
Galois group as a group of Permutations
Let f(x) F[x] be a monic polynomial with distinct roots r
1
, r
2
, . . . , r
n
. Let
E = F(r
1
, r
2
, . . . , r
n
) and G = G(E/F). Any G permutes the roots of
f(x). Dene : G = G(E/F) S
n
by () = |
R
. Then is an injective
group homomorphism. The subgroup (G) is called the Galois group of
f(x), and it is denoted by G
f
. By the FTGT, there is an intermediate
subeld of E/F corresponding to G
f
A
n
.
Theorem 24.1. Let F be a eld of characteristic = 2 and f(x) F[x], a
monic polynomial of positive degree with distinct roots r
1
, r
2
, . . . , r
n
F
a
.
Put E = F(r
1
, r
2
, . . . , r
n
). Put =
1i<jn
(r
i
r
j
). Then
E
G
f
An
= F().
Proof. Any transposition acting on maps to . Hence all permutations
in G
f
A
n
x . Thus F() E
G
f
An
. Let |G
f
/G
f
A
n
| = d. Then d 2.
If d = 1 then G
f
A
n
= G
f
and so G
f
A
n
. Thus F. Let d = 2.
Then G
f
A
n
= G
f
. So G
f
has an odd permutation. Hence / F. Thus
E
G
f
An
= F().

Denition 24.2. A subgroup H S

n
is called a transitive subgroup if for
any i = j {1, 2, . . . , n}, there exists H such that (i) = j.
136
Theorem 24.3. Let f(x) F[x] be a polynomial of degree n with n distinct
roots r
1
, r
2
, . . . , r
n
in F
a
. Then f(x) is irreducible if and only if G
f
is a
transitive subgroup of S
n
.
Proof. () Suppose G
f
is a transitive subgroup of S
n
. If f(x) is reducible
in F[x] then f(x) = g(x)h(x) for some g, h F[x] of positive degree. Let
g(r) = h(s) = 0 where r, s F
a
. Let G
f
be a permutation which maps
r to s. We may assume that g(x) is irreducible. But then s has to be a root
of g(x). Since f(x) has no repeated roots, h(x) is a constant.
() Suppose f(x) is irreducible. Let r, s be roots of f(x). Then there exists
an F-isomorphism : F(r) F(s) such that (r) = s. It can be extended
to an automorphism of F(r
1
, . . . , r
n
). Hence G
f
is transitive.

Transitive Subgroups of S
4
Let H be a transitive subgroup of S
n
. The orbit of action of H on [n] is
[n]. Thus n = | orbit (1)| = |H|/| stab (1)|. Hence n | |H|. The orders of
possible Galois groups of irreducible separable quartics are 4, 8, 12 and 24.
These groups are listed below.
(1) C
4
= {(1234), (13)(24), (1432), (1)}.
A cyclic group of order 4 has two 4-cycles. There are six 4-cycles in
S
4
. Thus there are three transitive cyclic subgroups of order 4.
(2) Klein 4 -group V = {(1), (12)(34), (14)(32), (13)(24)} is a normal
subgroup of S
4
.
(3) There are 3- Sylow subgroups of order 8. They are all isomorphic to
D
4
. These are H
1
= V, (13), H
2
= V, (12), H
3
= V, (14).
(4) A
4
is the only subgroup of order 12 and it is normal in S
4
.
(5) S
4
is the only subgroup of order 24.
Calculation of Galois group of quartic polynomials
Let F be a eld of char = 2, 3. Let f(x) = x
4
+b
1
x
3
+b
2
x
2
+b
3
x+b
4
F[x]
be separable. By the change y = x +
b
1
4
we may assume that there is no
x
3
term. This change does not alter the Galois group and the discriminant.
So let f(x) = x
4
+ bx
2
+ cx + d F[x] be an irreducible polynomial with
137
roots r
1
, r
2
, r
3
, r
4
in a splitting eld E of f(x) over F. We write G
f
S
4
.
So G
f
G(E/F). Set
t = {t
1
= r
1
r
2
+r
3
r
4
, t
2
= r
1
r
3
+r
2
r
4
, t
3
= r
1
r
4
+r
2
r
4
}.
Proposition 24.4. E
G
f
V
= F(t) and G(F(t)/F) =
G
f
G
f
V
.
Proof. Clearly, F(t
1
, t
2
, t
3
) E
G
f
V
. The element t
1
is xed by H
1
=
(12), V ), a dihedral group of order 8 in S
4
. Moreover
S
4
= H
1
(13)H
1
(14)H
1
.
Thus H
1
is the stabilizer of t
1
. Similarly, H
2
= Stab (t
2
) = (13), V ), H
3
=
Stab (t
3
) = (14), V ). Since V = H
1
H
2
H
3
, if G
f
xes t
1
, t
2
, t
3
then V. Hence G(E/F(t)) G
f
V which gives F(t) E
G
f
V
. We
know that F(t
1
, t
2
, t
3
) is the splitting eld of the resolvent cubic over F,
hence it is Galois. Thus G(F(t)/F)
G
f
G
f
V
.
Proposition 24.5. The resolvent cubic of a separable irreducible quartic
has a root in F if and only if G
f
D
4
.
Proof. Let t
1
F. Then G(E/F(t
1
)) = G
f
= G
f
H
1
G
f
H
1
.
Conversely if G
f
H
i
for some i say i = 1, then each G
f
xes t
1
and
hence t
1
E
G
f
= F.
Theorem 24.6. Let f(x) be an irreducible separable quartic over a eld F
of char F = 2 and E = F(r
1
, r
2
, r
3
, r
4
) be a splitting eld where r
1
, . . . , r
4
are the roots of f(x). Let r(x) denote resolvent cubic of f(x).
(1) If r(x) is irreducible in F[x] and disc (r(x)) / F
2
then G
f
S
4
.
(2) If r(x) is irreducible in F[x] and disc (r(x)) F
2
then G
f
A
4
.
(3) If r(x) splits completely in F[x] then G
f
V.
(4) Let r(x) have one root in F. Then
(a) If f(x) is irreducible over F(t) then G
f
D
4
.
(b) If f(x) is reducible over F(t) then G
f
C
4
.
Proof. Since f(x) is irreducible over F, G
f
is a transitive subgroup of S
4
.
Hence |G
f
| = 4, 8, 12, or 24, |G
f
V | = 1, 2 or 4, and |G
f
/G
f
V | = |G
r(x)
| =
1, 2, 3, 6. Thus |G
f
V | > 1. We also have |V G
f
| |
G
f
V G
f
| = |G
f
|. Thus
{2, 4} {1, 2, 3, 6} = {4, 8, 12, 24}.
138
(1) If r(x) is irreducible over F and disc (r(x)) F
2
then G
r(x)
A
3
.
Hence |G
f
/G
f
V | = 3. Hence |G
f
| = 12 and therefore G
f
A
4
.
(2) If r(x) is irreducible over F and disc (r(x)) is not a square in F, then
G
r(x)
S
3
. Hence |G
f
/G
f
V | = 6. Thus |G
f
| = 12 or 24. If |G
f
| = 12 then
G
f
A
4
and |G
f
/G
f
V | = 3 which is a contradiction. Hence G
f
S
4
.
(3) If r(x) has all its roots in F, then E
G
f
V
= F = E
G
f
. Thus G
f
V .
Since 4 | |G
f
|, G
f
= V .
(4) Now let r(x) have exactly one root in F. Then [F(t) : F] = 2 =
|G
f
/G
f
V |. Thus |G
f
| = 4 or 8.
(a) Suppose f(x) is irreducible over F(t). Then
[E : F(t)] = |G
f
V | 4 |G
f
| = 8 G
f
D
4
.
(b) Suppose f(x) is reducible over F(t). If G
f
D
4
then
[E : F] = 8 [E : F(t)] = 4.
Hence G(E/F(t)) = V which is transitive. Hence f(x) is irreducible over
F(t). This is a contradiction. So |G
f
| = 4. If G
f
= V then G
r(x)
=
G
f
/G
f
V = {1}. But |G
r(x)
| = 2. Thus G
f
C
4
.
Example 24.7. (1) (G
f
= V ) Let f(x) = x
4
+1 Q[x]. Then the resolvent
cubic is r(x) = x(x 2)(x + 2). Since f(x) is irreducible over Q, G
f
= V .
(2) (G
f
= C
4
) Consider f(x) = x
4
+ 5x
2
+ 5 which is irreducible over Q by
Eisenstein criterion. Then
r(x) = x
3
5x
2
20x + 100 = (x 5)(x 2

5)(x +

5).
Thus t
1
= 5, t
2
= 2

5, t
3
= 2

5. Hence F(t) = Q(

5) and
x
4
+ 5x
2
+ 5 =

x
2
+
5+

5
2

x
2

5
2

.
Therefore f(x) is reducible over F(t). Thus G
f
C
4
.
(3) (G
f
= S
4
) Consider f(x) = x
4
x +1. Then f(x) is irreducible modulo
2, and hence it is irreducible over Q. The resolvent cubic r(x) = x
3
4x1
is irreducible over Q and disc (r(x)) = 229 / Q
2
. Hence G
f
= S
4
.
139
(4) (G
f
= D
4
) The polynomial f(x) = x
4
3 is irreducible over Q and
r(x) = x(x +i2

3)(x i2

3). Therefore F(t) = Q(i

3). Hence
f(x) = (x
2

3)(x
2
+

3) = (x i
4

3)(x +i
4

3)(x +
4

3)(x
4

3).
Thus f(x) has no root in Q(i

3). The splitting eld of f(x) over Q is

Q(i,
4

3) which is a degree 8 extension of Q. Hence G

f
= D
4
.
(5) (G
f
= A
4
) Let f(x) = x
4
8x + 12. Then r(x) = x
3
48x 64. Using
Eisensteins criterion, f(x) is irreducible over Q. Since disc (r(x)) = 2
12
3
4
is a perfect square in Q, G
f
= A
4
.
Example 24.8. Let p be a prime number and f(x) = x
4
+ px + p. Then
r(x) = x
3
4pxp
2
. Possible roots of r(x) in Q are 1, p, p
2
. Check that
1, p
2
are not roots for any p. But r(p) = p
2
(p5) and r(p) = p
2
(3p).
Hence r(x) has a rational root if and only if p = 3, 5. For p = 3, 5, the
resolvent cubic is irreducible over Q. Check that disc (f(x)) = p
3
(25627p)
is never a perfect square in Q. Let G be the Galois group of f(x). Then
G = S
4
if p = 3, 5. If p = 3 then r(x) = (x + 3)(x
2
3x 3). Hence the
splitting eld L of r(x) over Q is Q(

21). Check that x

4
+3x+3 is irreducible
over Q(

21). Hence G = D
4
. The p = 5 case has been considered in the
previous example.
145
Lecture 25 : Norm, Trace and Hilberts Theorem 90
Objectives
(1) The norm and the trace function.
(2) Multiplicative form of Hilberts Theorem 90.
(3) Cyclic extensions of degree n.
(4) Additive version of Hilberts 90.
(5) Cyclic extensions of prime degree: Artin-Schreier Theorem.
Keywords and phrases: Norm, trace, Hilberts theorem 90, cyclic exten-
sions, Artin-Schreier Theorem.
25. Norm, Trace and Hilberts Theorem 90
Denition 25.1. Let E/F be a nite separable extension of degree n. Let

1
, . . . ,
n
be the F-embeddings : E F
a
. For any a E, dene the norm
and trace of a by,
N
E/K
(a) =
1
(a)
2
(a)
n
(a)
Tr
E/K
(a) =
1
(a) + +
n
(a).
Example 25.2. Let m be a square free integer. Consider the quadratic
extension E = Q(

m) of Q. The Galois group G = G(E/Q) consists of

identity map and the automorphism (a +

m) = a b

m. Therefore
Tr(a + b

m) = 2a and N(a + b

m) = a
2
mb
2
.
Proposition 25.3. (1) N
E/K
: E

is a group homomorphism.
(2) Let E K F be a tower of nite separable extensions. Then
N
E/F
= N
K/F
N
E/K
, Tr
E/F
= Tr
K/F
Tr
E/K
(3) If E = F(a) and irr (a, F) = x
n
+ a
n1
x
n1
+ + a
0
then
N
E/F
(a) = (1)
n
a
0
, and Tr
E/F
(a) = a
n1
.
(4) Tr : E F is a surjective F-linear map.
Proof. (1) N
E/F
(ab) = N
E/F
(a)N
E/F
(b) for all a, b E is clear.
Let L =
1
(F) . . .
n
(F). Then L/F is a Galois extension. Let a E

.
Then N
E/F
(a) is xed under all G(L/F), thus it is in F

.
146
(2) Let {
j
} be the family of F-embeddings : K F
a
and {
i
} be the
family of all K-embeddings of : E F
a
. Each
j
can be extended to an
automorphism of F
a
. Let this extension be denoted by
j
. Then {
j

i
} is
the family of all F-embeddings of E F
a
. For any x E,
N
K/F
N
E/K
(x) = N
K/F

i=1

i
(x)

=
m

j=1
n

i=1

i
(x) = N
E/F
(x).
For any x E we have
Tr
K/F
Tr
E/K
(x) = Tr
K/F

i=1

i
(x)

=
m

j=1
n

i=1

i
(x) = Tr
E/F
(x).
(3) Suppose E = F(a) and f(x) = x
n
+ a
n1
x
n1
+ + a
n
= irr (a, F)
and f(x) = (x a
1
)(x a
2
) (x a
n
) where a
1
, . . . , a
n
are all the roots
in F
a
of f(x). Each a
i
= (a
1
) for some F-embedding : E F
a
. Thus
N
E/F
(a) = (1)
n
a
n
and Tr
E/F
(a) = a
n1
.
(4) Tr
E/k
(a) =
1
(a) + +
n
(a). By Dedekinds theorem on characters,

1
+ +
n
is not a zero map. Since Tr
E/F
is a linear map of F-vector
spaces, it is surjective.
Proposition 25.4. Let E/F be a nite separable extension of degree n and
a E. Let m
a
: E E be the F-linear map dened as m
a
(x) = ax for all
x E. Then
N
E/F
(a) = det(m
a
) and Tr
E/F
(a) = Tr(m
a
).
Proof. Let K = F(a) and f(x) = irr (a, F) = x
d
+a
d1
x
d1
+ +a
1
x+a
0
.
Then 1, a, a
2
, . . . , a
d1
is an F-basis for K. Let v
1
, v
2
, . . . , v
e
be a K-basis of
E. Then {v
i
a
j
| i = 1, 2, . . . , e; j = 0, 1, . . . , d 1} is an F-basis of E. We
order this basis as :
B = {v
1
, av
1
, a
2
v
1
, . . . , a
d1
v
1
; . . . ; , v
e
, av
e
, a
2
v
e
, . . . , a
d1
v
e
}.
Consider the matrix
A =

0 0 0 . . . 0 a
0
1 0 0 . . . 0 a
1
0 1 0 . . . 0 a
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 . . . 1 a
d1

.
147
Then the characteristic polynomial of A is f(x). The matrix of m
a
with
respect to B is the n n matrix:

A 0 0 . . . 0
0 A 0 . . . 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 . . . A

.
Therefore det m
a
= (det A)
e
and Tr m
a
= e Tr A. Therefore
N
E/F
(a) = N
K/F
N
E/K
(a) = N
K/F
(a
e
) = (det A)
e
= det m
a
,
Tr
E/F
(a) = Tr
K/F
Tr
E/K
(a) = Tr
K/F
(ea) = e Tr A = Tr m
a
.

Proposition 25.5. Let E/F be a nite separable extension. Then

(1) the map : E E F given by (x, y) = Tr(xy) is bilinear.
(2)The map T
x
: E F given by T
x
(y) = Tr(xy) is an F-linear map.
(3) The map : E Hom(E, F) given by (x) = Tr
x
is an isomorphism.
Proof. It is easy to see (1) and (2). For (3), if (x) = Tr
x
= 0 then Tr
x
(y) =
Tr(xy) = 0 for all y E. Hence for any e E, Tr
x
(x
1
e) = Tr(e) = 0.
Thus Tr is the zero functional. This is a contradiction. Hence is an
injective linear map. Since dimE = dimHom(E, F), we conclude that is
an isomorphism.

Theorem 25.6 (Hilberts Theorem 90 (multiplicative form)). Let

E/F be a cyclic extension. Let G(E/F) = (). Then for E,
N
E/F
() = 1 if and only if =

()
for some E

.
Proof. Let [E : F] = n. If =

()
, then
N
E/F
() = ()
n1
() =
()
()
2
()


n1
()

= 1.
Conversely, suppose N
E/k
() = 1. Consider
id + + ()
2
+ ()
2
()
3
+ + ()
n2
()
n1
is a nonzero map from E F due to Dedekinds independence theorem.
Let K be such that
148
= + () + ()
2
() + + ()
n2
()
n1
() = 0.
Then
() = () + ()
2
() + + ()
2
()
n1
() = .
Therefore =

()
.

Theorem 25.7. Let k be a eld, n a positive integer coprime with char k

and assume k has a primitive n
th
root w of 1. Let E/k be cyclic extension
of degree n. Then E is splitting eld of x
n
a k[x].
Proof. Let G(E/k) = (). Then N
E/k
(w
1
) = w
n
= 1. By Hilberts the-
orem 90, there exists E such that () = w. Thus
i
() = w
i
for
i = 1, ..., n. Hence has n distinct conjugates in E. Since [E : k] = n,
E = k(). Since (
n
) = (w)
n
=
n
:= a E
G
= k. Thus E is a splitting
eld of x
n
a.

We now discuss the additive form of Hilbert 90 and its application to cyclic
extension of degree p, where p is prime and is equal to the characteristic of
the base eld.
Theorem 25.8 (Additive form of Hilberts Theorem 90). Let E/k be
a cyclic extension of degree n with Galois group G. Let G = . Then for
E
Tr
E/k
() = 0 if and only if = () for some E.
Proof. Let = (). Then Tr() = Tr() Tr(()) = 0.
Let Tr() = 0. Since Tr : E k is a nonzero map, there exits E such
that Tr() = 0. For the element
=
1
Tr()
[ + ( + ())() + + ( + () + +
n2
())
n2
()],
() =
1
Tr()
[()() + (() +
2
())
2
() + + (() +
2
() + +

n1
())
n1
()]
As Tr() = 0, () =
1
Tr()
[ + () + +
n1
()] = .
149
Theorem 25.9 (Artin-Schreier). Let k be a eld of char p > 0. Let E/k
be a cyclic extension of degree p. Then E is a splitting eld of x
p
x a
for some a E and E = k() where
p
= a for some E.
Proof. Let E/k be cyclic of degree p. Then Tr(1) = 0. Hence there exists
E such that () = 1 where = G(E/k). Thus () = + 1.
Hence
i
() = a + i for i = 1, 2, . . . , p. Since char k = p, the elements
, + 1, ..., + p 1 are distinct. Hence [k() : k] = p and E = k(). As
(
p
) = (())
p
() = (+1)
p
(+1) =
p
,
p
E
()
= k.
Let a =
p
k. Then satises f(x) = x
p
x a = 0. The roots of
f(x) are , + 1, . . . , + p 1. Thus E is a splitting eld of f(x).

Example 25.10. (Pythagorean Triples) Let us nd all Pythagorean

triples (x, y, z) such that x
2
+ y
2
= z
2
where x, y, z N. Hence x
2
/z
2
+
y
2
/z
2
= N(x/z +iy/z) = 1. Let us apply Hilberts theorem 90 to the cyclic
extension Q(i)/Q. The Galois group of this extension is cyclic of order 2
generated by the conjugation automorphism. Hence N(a +ib) = a
2
+b
2
. So
there exists = c + id Q(i) such that
x/z + iy/z = (c + id)/(c id) = (c
2
d
2
+ 2icd)/(c
2
+ d
2
).
Thus x/z = (c
2
d
2
)/(c
2
+ d
2
) and y/z = 2cd/(c
2
+ d
2
). Putting c = s/u
and d = t/u where s, t, u N, we get
x = s
2
t
2
, y = 2st, z = s
2
+ t
2
.