M3P11 Galois Theory

Lectured by Prof Alessio Corti

Typed by David Kurniadi Angdinata
Spring 2019

K (α) K (α0 ) K (β, γ) K (δ) K (δ 0 )

K (β) K (βγ) K (γ)

hei

hτ i σ2 τ σ2 hστ i σ3 τ

σ2 , τ hσi σ 2 , στ

Syllabus
Field extensions. Irreducible polynomials. Normal extensions. Separable extensions. Fundamental the-
orem of Galois theory. Galois theory of quadratic polynomials. Galois theory of biquadratic polynomials.
Galois theory of finite fields. Galois theory of cubic polynomials. Reduction modulo prime.

1
M3P11 Galois Theory Contents

Contents
0 What is Galois theory? 3

1 Eisenstein numbers 6

2 Elementary facts 9

3 Axiomatics 11

4 Galois correspondence 14

5 Normal extensions 17

6 Separable polynomials 18

7 Separable degree 20

8 Separable extensions 21

9 Biquadratic polynomials 22

10 Finite fields 26

11 Symmetric polynomials 27

12 Irreducible polynomials 29

13 Reduction modulo prime 31

2
M3P11 Galois Theory 0 What is Galois theory?

0 What is Galois theory?

Lecture 1
References. Thursday
• E Artin, Galois theory, 1994 10/01/19

• A Grothendieck and M Raynaud, Revêtements étales et groupe fondamental, 2002

• I N Herstein, Topics in algebra, 1975
• M Reid, Galois theory, 2014
Notation. If K is a field, or a ring, I denote

K [X] = {a0 + · · · + an X n | ai ∈ K} ,

the ring of polynomials with coefficients in K.

Example.
• Q, R, C.
• Quadratic fields
√  n √ o Q [X]
Q 2 = a + b 2 | a, b ∈ Q = .
hX 2 − 2i
It is also a field, since √
1 a−b 2
√
= 2 .
a+b 2 a − 2b2

• If p is prime, Z/pZ = Fp is a finite field. If f (X) ∈ K [X] is irreducible, K [X] / hf (X)i is a field.
For example, X 2 − 2. Both Z and K [X] have a division algorithm. For example, let [a] ∈ Z/pZ and
[a] 6= 0, that is p | a. Since p is prime, gcd (p, a) = 1, so there exist x, y ∈ Z such that ax + py = 1.
Thus [a] · [x] = 1 in Z/pZ.
• For K a field, either for all m ∈ Z, m 6= 0 in K, so K has characteristic ch (K) = 0, or there exists p
prime such that m = 0 if and only if p | m, so K has characteristic ch (K) = p.
• For K a field,
( )
f (X)
K (X) = F rac (K [X]) = φ (X) = f, g ∈ K [X] , g 6= 0 .
g (X)

is also a field, the field of rational functions with coefficients in K. For example, Fp (X, Y ) = Fp (X) (Y ).
Example. Consider algebraic equations in a field K.
• Let aX 2 + bX 2 + c = 0 for a, b, c ∈ K be a quadratic. There is a formula
√
−b + b2 − 4ac
X= .
2a

• For a cubic Y 3 + 3pY + 2q = 0,

q p q p
3 3
Y = −q + q + p + −q − q 2 + p3 .
2 3

• There is a formula for quartic equations.

• It is a theorem that there can be no such formula for equations of degree at least five.
Galois theory deals with these easily.

3
M3P11 Galois Theory 0 What is Galois theory?

Lecture 2
Definition 0.1. A field homomorphism is a function φ : K1 → K2 that preserves the field operations, Friday
for all a, b ∈ K1 , 11/01/19

φ (a + b) = φ (a) + φ (b) ,
φ (ab) = φ (a) φ (b) ,

and φ (0K1 ) = 0K2 and φ (1K1 ) = 1K2 .

Remark. All field homomorphisms are injective. If a ∈ K1 \ {0}, then there exists b ∈ K1 such that ab = 1,
then φ (a) φ (b) = 1, so φ (a) 6= 0. This easily implies φ is injective. If a1 6= a2 , then a1 − an 6= 0, so
0 6= φ (a1 − a2 ) = φ (a1 ) − φ (a2 ). Then φ (a1 ) 6= φ (a2 ).
We concern ourselves with field extensions k ⊂ K, and every homomorphism is an extension. Consider a
field extension k ⊂ K and α ∈ K. Then k (α) ⊂ K denotes the smallest subfield of K that contains k and
α. Not to be confused with k (X).
Example. There are two very different cases exemplified in Q ⊂ C.
√ √

• α = 2, so Q 2 .
• α = π, so Q (π).

Definition 0.2.
• α is algebraic over k if f (α) = 0 for some 0 6= f ∈ k [X]. Otherwise we say that α is transcendental
over k.
• The extension k ⊂ K is algebraic if for all α ∈ K, α is algebraic over k.

Definition 0.3. Consider a field k and f ∈ k [X]. We say that k ⊂ K is a splitting field for f if
Qn
• f (X) = a i=1 (X − λi ) ∈ K [X] for a ∈ k \ {0}, and
• K = k (λ1 , . . . , λn ).

Example.
√

• If f (X) = X 2 − 2 ∈ Q [X], then K = Q 2 is a splitting field for f . Indeed
 √  √  √ 
X2 − 2 = X + 2 X − 2 ∈ Q 2 [X] .

√

• If f (X) = X 2 + 2, then K = Q −2 .
• If f (X) = X 3 − 2, then √  n
3
√
3
√
3
o
Q 2 = a + b 2 + c 4 | a, b, c ∈ Q
√ √
2, ω , where ω = −1+2 3 , is a splitting field.
3


is not a splitting field. K = Q
 √  √  √ 
X 3 − 2 = X − 2 X − ω 2 X − ω2 2 .
3 3 3

4
M3P11 Galois Theory 0 What is Galois theory?

Theorem 0.4 (Fundamental theorem of Galois theory, Galois correspondence). Assume characteristic zero.
Let k ⊂ K be the splitting field of f (X) ∈ k [X]. Let

G = {σ : K → K | σ field automorphism, σ|k = idk } .

We call this group the Galois group. There is a one-to-one correspondence

{k ⊂ K1 ⊂ K | K1 subfield} ↔ {H ≤ G | H subgroup}
K1 7→ {σ ∈ G | ∀λ ∈ K1 , σ (λ) = λ} .
{λ ∈ K | ∀σ ∈ H, σ (λ) = λ} ←[ H ≤ G

Why is this cool? Fields are hard, groups are easy. We will see that there is a good formula for the roots of
f (X) if and only if G is a soluble group. Lecture 3
Tuesday
Example. Let deg (f ) = 2 and f (X) = X 2 + 2AX + B ∈ K [X]. If K already contains the roots then 15/01/19
L = K and G = {id}. Suppose K does not contain the roots. We still have quadratic formula
p
λ1,2 = −A ± A2 − B.
√
If ∆ = A2 − B then ∆ does not exist in K. We must have
√  n √ o
L=K ∆ = a + b ∆ | a, b ∈ K .

Then K ⊂ L and
G = {σ : L → L | σ|K = idK } = C2
is generated by √ √
σ : a + b ∆ 7→ a − b ∆.
The following is further specialisation.
• Let K = R and ∆ = −1. Then
 √
L = C = a + b −1 | a, b ∈ R ,

and G = C2 is generated by √ √
σ : a + b −1 7→ a − b −1,
complex conjugation.
• Let K = Q and ∆ = 2. Then n √ o
L = a + b 2 | a, b ∈ Q ,

and G = C2 is generated by √ √
σ : a + b 2 7→ a − b 2.

The fundamental theorem implies there does not exist

√ 
K ( K1 ( K ∆ = L.
√
Is this obvious? Consider x ∈ L \ K, so x = a + b ∆, and b 6= 0, and then
√ x−a
∆= ,
b
so K (x) = L.

5
M3P11 Galois Theory 1 Eisenstein numbers

1 Eisenstein numbers
√ √
2, ω , where ω = −1+i 3


Let f (X) = X 3 − 2 ∈ Q [X] and L = Q 3
2 is a solution of X 2 + X + 1 = 0. Then
√ √  n √ √ o
3 3 3


Q (ω) = Q −3 , Q 2 = a + b 2 + c 4 | a, b, c ∈ Q .

Remark. For any splitting field of f , there is always a natural inclusion group homomorphism

ρ : G ,→ S (λ1 , . . . , λn ) ,

where S (λ1 , . . . , λn ) is the group of permutations of the roots of f = X n + a1 X n−1 + · · · + an .

• If σ ∈ G, f (λ) = 0, so λn + a1 λn−1 + · · · + an = 0.
n n−1
0 = σ (0) = σ λn + a1 λn−1 + · · · + an = σ (λ) + a1 σ (λ)


+ · · · + an .

• ρ is injective. If for all i, σ (λi ) = λi , then σ = id on K (λ1 , . . . , λn ) = L.

The fundamental theorem and remark implies that G = S3 . Lecture 4
Thursday
Definition 1.1. K ⊂ L is finite if L is finite-dimensional as a vector space over K. The degree of L over 17/01/19
K is
[L : K] = dimK (L) .
Theorem 1.2 (Tower law). Let K ⊂ L ⊂ F . Then

[F : K] = [F : L] [L : K] .

Theorem 1.3. Suppose f (X) ∈ K [X] is irreducible of degree d = deg (f ) and L = K (λ) where f (λ) = 0,
then [K (λ) : K] = d.
√
√

K = Q 3 2 is a field, and [K : Q] = 3. Let L = Q 3 2, ω be the splitting field of X 3 − 2 over Q. The
lattice of subfields is
L
2 2
2
√
√
√
3
Q 32 Q ω32 Q ω2 3 2
.

3
3 Q (ω)
3

2
Q
Then (Exercise)
√   √   √  √ √ 
Q ω 2 2 ∩ Q ω 2 = Q,
3 3 3 3 3
Q 2 + ω = L, Q 2, ω 2 = L.
 √
 √
√
√ √
√

What is L : Q 3 2 ? Note that L = Q 3 2 −3 . Could −3 ∈ Q 3 2 ? Consider X 2 +3 ∈ Q 3 2 [X].
By the tower law,
(
[L : Q] = [L : Q (ω)] [Q (ω) : Q] = 2 [L : Q (ω)] =⇒ 2 | [L : Q]
 √
  √
  √
 =⇒ 6 | [L : Q] .
[L : Q] = L : Q 3 2 Q 3 2 : Q = 3 L : Q 3 2 =⇒ 3 | [L : Q]
√
 √

• Either X 2 + 3 is irreducible over Q 3
2 , so by Theorem 1.3 L : Q 3 2 = 2 and [L : Q] = 6.
√

• Or X 2 + 3 is not irreducible, so Q 3 2 = L and [L : Q] = 3, a contradiction.

6
M3P11 Galois Theory 1 Eisenstein numbers

Are there any other fields? Claim that there are no other fields. Suppose Q ( K ( L is such a field. By the
tower law [K : Q] = 2 or [K : Q] = 3.
• Suppose [K : Q] = 2.
L

3
K (ω)
.
K Q (ω)

2 2
Q

– Either ω ∈ K, that is Q (ω) ⊂ K, so by the tower law Q (ω) = K.

– Or ω ∈/ K, so [K (ω) : K] = 2, so [K (ω) : Q] = 4 contradicts the tower law for Q ⊂ K (ω) ⊂ L.
• Suppose [K : Q] = 3.
L

√
3


2 K 2
.

K
3

Q
 √

Claim that X 3 − 2 ∈ K [X] splits. Suppose that it were irreducible, then K 3 2 : K = 3, which
√
√ √
contradicts the tower law for K ⊂ K 3 2 ⊂ L. So it has a root in K. Either 3 2 ∈ K, ω 3 2 ∈ K, or
√ √
√
√

ω 2 3 2 ∈ K. Thus Q 3 2 = K, Q ω 3 2 = K, or Q ω 2 3 2 = K.
I want to prove that n o
G = AutQ (L) = σ : L → L | σ|Q = idQ = S3 .
Lecture 5
Proof of Theorem 1.2. Suppose y1 , . . . , ym ∈ F is a basis of F as a vector space over L. Suppose x1 , . . . , xn ∈ Friday
L is a basis of L as a vector space over K. Claim that {xi yj } is a basis of F over K. 18/01/19
• {xi yj } generates F . Let z ∈ F . There exist µ1 , . . . , µn ∈ L such that
z = µ1 y1 + · · · + µn yn . (1)
µj ∈ L so for all j there exists λij ∈ K such that
µj = x1 λ1j + · · · + xm λmj . (2)
Plugging in (2) into (1), X
z= λij xi yj .
i,j

• {xi yj } are linearly independent over K. Suppose there exists λij ∈ K such that
!
X X X
0= λij xi yj = λij xi yj ,
i,j j i
P
so for all j, i λij xi = 0, so for all j and all i, λij = 0.

7
M3P11 Galois Theory 1 Eisenstein numbers

Example. To show G = S3 . Let σ = 1 2 . A basis of L/Q is
√
3
√
3
√
3
√
3
1, 2, 4, ω, ω 2, ω 4.

• σ (1) = 1.
√
√
• σ 3 2 = ω 3 2.
√
√
• σ ω 3 2 = 3 2.
√
√ √
√ √ √ √ √ √
• σ 3 4 = σ 3 2 · 3 2 = ω 3 2 · ω 3 2 = ω 2 3 4 = (−ω − 1) 3 4 = −ω 3 4 − 3 4.
√ √
√
√
√ √
• σ (ω) = σ ω 3 2/ 3 2 = σ ω 3 2 /σ 3 2 = 3 2/ω 3 2 = 1/ω = −1 − ω.
√
√ √
√
√
√ √ √
• σ ω 3 4 = σ ω 3 2 · 3 2 = σ ω 3 2 · σ 3 2 = 3 2 · ω 3 2 = ω 3 4.
Thus  
1 0 −1 0 0 0
0 0 0 1 0 0
 
0 0 0 0 0 −1
σ= .
0 0
 −1 0 0  0
0 1 0 0 0 0
0 0 0 0 1 −1


A question is if there were σ ∈ G such that ρ (σ) = 1 2 then we have written the matrix of σ as a Q-linear
map of L in a basis. But how to check that this Q-linear map is a field homomorphism? We know the Galois
correspondence for extensions of degree two.
  √    √    √ 
2 , Gal L/Q ω 2 2 , Gal L/Q ω 2 ⊂ G
3 3 3
Gal L/Q

contain an element of order two, and

√
3




ρ: Gal L/Q √ 2

7→ 2 3

Gal L/Q ω 2 √
3
2

7 → 1 2

Gal L/Q ω 3 2 7 → 1 3 .

The lattice of subgroups is

hei
2 2
2



3
2 3 1 2 1 3
.


3
3 1 2 3
3

2
G

Q (ω) /Q is the splitting field of X 2 + X + 1 and of X 2 + 3.

We can learn the following. Let k ⊂ L be a splitting field. Consider k ⊂ K ⊂ L. Then K ⊂ L is also a
splitting field. The corresponding H ≤ G is the Galois group Gal (L/K). On the other hand k ⊂ K is not
always a splitting field. It is a splitting field if and only if the corresponding H ≤ G is a normal subgroup
and in that case Gal (K/k) = G/H.

8
M3P11 Galois Theory 2 Elementary facts

2 Elementary facts
Lecture 6
Let K ⊂ L and a ∈ L. The evaluation homomorphism Tuesday
22/01/19
ea : K [X] −→ K [a] ⊂ L
f (X) 7−→ f (a)

is a surjective ring homomorphism, where K [a] is the smallest subring of L containing K and a.
Definition 2.1. f (X) = a0 X n + · · · + an ∈ K [X] is monic if a0 = 1.
Lemma 2.2.

• If a is transcendental, ea is injective and it extends to eea : K (X) → K (a), by

K (X)
e a .

⊂
f

K [X] ea L

• If a is algebraic, then Ker (ea ) = hfa i, where fa ∈ K [X] is irreducible, or prime, and unique if monic,
then called the minimal polynomial of a ∈ L/K. In this case
ea
K [X] K [a] ∼
= K (a) ⊂ L
∼
⊂

[ea ]
.
K [X]
hfa i

Proof. There is nothing to prove.

Remark. Let g (X) ∈ K [X] and g (a) 6= 0. Claim that 1/g (a) ∈ K [a]. Indeed gcd (f, g) = 1 in K [X] and
f - g. There exists φ, ψ ∈ K [X] such that f φ + gψ = 1 and g (a) ψ (a) = 1. All of this is saying
• K [a] ∼
= K (a), and
• K [X] / hfa i ∼
= K (a).
Let
EmK (K (a) , F ) = {σ : K (a) → F | σ field homomorphism, σK = idK } ,
where
K (a)
⊂

K σ .
⊂

F
Corollary 2.3. For K ⊂ L and a ∈ L algebraic over K,
• [K (a) : K] = deg (fa ), and

• If K ⊂ F is an extension,
EmK (K (a) , F ) = {b ∈ F | fa (b) = 0} .

9
M3P11 Galois Theory 2 Elementary facts

Proof. Since K (a) = K [a], [K (a) : K] = dimK (K (a)) = dimK [K (a)]. Suppose

f (X) = X n + µ1 X n−1 + · · · + µn ∈ K [X]

is the minimal polynomial of a over K. Claim that 1, . . . , an−1 is a basis of K [a] over K.
• The set generates K [a]. Let c ∈ K [a]. There exists g ∈ K [X] such that g (a) = c. Long division
implies that
g (X) = f (X) q (X) + r (X) , m = deg (r (X)) < n.
Then r (X) = λ0 + · · · + λm X m and g (a) = r (a) = λ0 + · · · + λm am .
• The set is linearly independent, otherwise there exists

g (X) = λ0 + · · · + λn−1 X n−1 ∈ K [X] , g (a) = 0,

and f was not the minimal polynomial.

σ (a) is a root of f , since applying σ to f (a) = 0 gives
n n−1
0 = σ an + µ1 an−1 + · · · + µn = σ (a) + µ1n−1 σ (a)


+ · · · + µn = f (σ (a)) .

Vice versa, if b ∈ F is a root of f ,

[eb ] K [X] [ea ]
K (b) ←−− −−→ K (a) ,
∼ hf i ∼
−1
then σ = [eb ] [ea ] . Thus there is a one-to-one correspondence

EmK (K (a) , F ) ↔ {b ∈ F | f (b) = 0}

σ 7→ σ (a) .
−1
[eb ] [ea ] ←[ b

Corollary 2.4. Let K be a field and f ∈ K [X]. Then there exists K ⊂ L such that f has a root in L.
Proof. Take g a prime factor of f . Take L = K [X] / hgi. In here a = [X] is a root of g hence a root of f .
Lecture 7
From now on in this course, we study field extensions K ⊂ L, always assumed to be finite, so [L : K] = Thursday
dimK (L) < ∞. 24/01/19
Remark. K ⊂ L is finite if and only if
• it is algebraic, that is for all a ∈ L, a is algebraic over K, and
• it is finitely generated, that is there exist a1 , . . . , am ∈ L such that L = K (a1 , . . . , am ).
An important point of view is that we study all possible field homomorphisms

Em (K, L) = {σ : K → L | σ field homomorphism} .

Often there is a field k ⊂ K, L in the background which we want to stay fixed, so let

Emk (K, L) = {σ : K → L | σ field homomorphism, σ|k = idk } .

√
√ √ √

Example. Let K = Q 3 2 . The minimal polynomial of 3 2 is X 3 − 2. Let L = Q 3 2, −3 be the
splitting field of X 3 − 2. Then
n√ √ √ o
EmQ (K, L) = Em (K, L) = roots of X 3 − 2 in L = 2, ω 2, ω 2 2 .
3 3 3


Remark. Suppose k ⊂ K. Then Emk (K, K) = G = Gal (K/k). Indeed every k-homomorphism σ : K → K
is automatically invertible. We know σ is injective. σ is also surjective because σ is a k-linear endomorphism
of a finite-dimensional k-vector space.

10
M3P11 Galois Theory 3 Axiomatics

3 Axiomatics
Proposition 3.1. Fix k ⊂ K and k ⊂ L. Then

#Emk (K, L) ≤ [K : k] .

Proof.
Special case. If K = k (a), let f (X) ∈ k [X] be the minimal polynomial of a. Then Emk (k (a) , L) is the
roots of f (X) in L, so

#Emk (K, L) = # {roots} ≤ deg (f ) = [k (a) : k] ,

as proved last time.

General case. If k = K, nothing to do. Otherwise choose a ∈ K \ k.

L
⊂ y .
k ⊂ k (a) ⊂ K

Consider the restriction map

ρ : Emk (K, L) → Emk (k (a) , L) .

Fix y ∈ Emk (k (a) , L). Then

n o
ρ−1 (y) = x : K → L | x|k(a) = idk(a) .

Since [k (a) : k] > 1, by the tower law [K : k (a)] < [K : k]. By induction we may assume
#ρ−1 (y) ≤ [K : k (a)]. So
X
#Emk (K, L) ≤ #ρ−1 (y) ≤ [k (a) : k] [K : k (a)] = [K : k] ,
y∈Emk (k(a),L)

by the tower law.

Proposition 3.2. Suppose given two field extensions k ⊂ K and k ⊂ L. Then there is a non-unique bigger
common field
K
σ1 φ1

k Ω
σ2 φ2
L
that contains both.

11
M3P11 Galois Theory 3 Axiomatics

Remark.
• More formally, suppose given σ1 ∈ Em (k, K) and σ2 ∈ Em (k, L), then there exist Ω, φ1 ∈ Em (K, Ω),
and φ2 ∈ Em (L, Ω) such that φ1 ◦ σ1 = φ2 ◦ σ2 .
√
√

• I never said that Ω is unique. For example, let K = Q 3 2 and L = Q 3 2 . One choice is Ω = k.
√ √

Another choice is Ω = Q 3 2, −3 , where
√
3


Q 2
√
3 √
3
σ 27→ 2

√ √

Q Q 3
2, −3 .
σ √
3 √
3
27→ω 2
√

Q 32

• Another more precise way to state this is there exists k ⊂ Ω such that Emk (K, Ω) and Emk (L, Ω) are
both non-empty.
Lecture 8
Friday
Proof.
25/01/19
Special case. If K = k (a), let f (X) ∈ k [X] be the minimal polynomial of a over k. Let L ⊂ E be
such that f (X) ∈ L [X] has a root α ∈ E. Then there exists σ ∈ Emk (k (a) , E) such that
σ (a) = α.
k (a)
σ
⊂
.
k E
⊂
⊂

General case. By induction on [K : k]. If [K : k] = 1, take Ω = L. If [K : k] > 1, take a ∈ K \ k.

K
⊂ ⊂

k (a) Ω
⊂ ⊂
.
⊂

k E
⊂
⊂

By special case there exists E as in the diagram. By tower law [K : k (a)] < [K : k] hence
by induction find Ω as in the diagram. Ω solves the original problem.

12
M3P11 Galois Theory 3 Axiomatics

Proposition 3.3. Let L be any field and G be a finite group acting on L as automorphisms. Let
K = G∗ = F ix (G) = LG = {λ ∈ L | ∀σ ∈ G, σ (λ) = λ} .
†
Consider AutK (L) = K † . Then the obvious inclusion G ⊂ K † = (G∗ ) is an equality, so G is all of K † .
Remark. Contextualising, this thing is half of the Galois correspondence.
{F | k ⊂ F ⊂ Ω} ↔ {G | G ≤ Autk (Ω)}
F 7→ AutF (Ω) = F † .
F ix (G) = G∗ ←[ G
†
∗
Then to prove the Galois correspondence, we need for all G, G = (G∗ ) . We also need for all F , F = F † .
Proposition 3.3 follows from the following lemma.
Lemma 3.4. K ⊂ L is a finite extension of degree [L : K] ≤ |G|.
Proof of Proposition 3.3. From Proposition 3.1, AutK (L) = EmK (L, L) because K ⊂ L is finite, and
#EmK (L, L) ≤ [L : K]. By Lemma 3.4,
[L : K] ≤ #EmK (L, L) ≤ [L : K] ,
so |G| = #EmK (L, L). By what we said, G ⊂ EmK (L, L), so G = EmK (L, L).
Lecture 9
Lecture 9 is a problem class. Tuesday
Proof of Lemma 3.4. Write G = {σ1 , . . . , σn } for n = |G|. Want that all (n + 1)-tuples a1 , . . . , an+1 ∈ L are 29/01/19
linearly dependent over K. Let a1 , . . . , an+1 ∈ L. Consider the n + 1 vectors in Ln . Let Lecture 10
    Thursday
σ1 (a1 ) σ1 (an+1 ) 31/01/19
a1 =  ...  , . . . , an+1 =  ..
∈L .
n
   
.
σn (a1 ) σn (an+1 )
These are linearly dependent over L. There exist x1 , . . . , xn+1 ∈ L not all zero such that
x1 a1 + · · · + xn+1 an+1 = 0.
By reordering the ai , may assume
x1 a1 + · · · + xk ak = 0, (3)
for some 1 ≤ k ≤ n + 1 with
• for all i ∈ {1, . . . , k}, xi 6= 0,
• such k is the smallest, and
• x1 = 1.
Claim that all these xi ∈ K. This does it, by reading j-th row where σj = idG . We need to show for all i,
xi ∈ LG . Take σ ∈ G.
   
σ (σ1 (a1 )) σ (σ1 (ak ))
.. .. n
σ (x1 )   + · · · + σ (xk )  =0∈L .
   
. .
σ (σn (a1 )) σ (σn (ak ))
Note that
G −→ G
τ 7−→ σ◦τ
is a bijective function and {σ ◦ σ1 , . . . , σ ◦ σn } = G. Multiplying by σ reshuffles the rows. So in fact,
σ (x1 ) a1 + · · · + σ (xk ) ak = 0. (4)
Claim that for all i, σ (xi ) = xi . Otherwise (3) − (4),
(x2 − σ (x2 )) a2 + · · · + (xk − σ (xk )) ak = 0
is a shorter solution, contradicting k minimal.

13
M3P11 Galois Theory 4 Galois correspondence

4 Galois correspondence
Definition 4.1. k ⊂ K is normal if

∀k ⊂ Ω, ∀σ1 , σ2 ∈ Emk (K, Ω) , ∃σ ∈ Emk (K, K) , σ2 = σ1 ◦ σ. (5)

Ω
⊂ σ
⊂

σ1 (K) σ1 K σ2 σ2 (K) .

⊂
⊂ ⊂

k
Equivalently, k ⊂ K is normal if

∀k ⊂ Ω, ∀σ1 , σ2 ∈ Emk (K, Ω) , σ2 (K) ⊂ σ1 (K) . (6)

√
√ √

Example. Q ⊂ Q 3 2 is not normal. Take Ω = Q 3 2, −3 .

(5) =⇒ (6) Indeed for all λ ∈ K, σ2 (λ) = σ1 (σ (λ)) ∈ σ1 (K), so σ2 (K) ⊂ σ1 (K).
(6) =⇒ (5) Work inside Ω, so k ⊂ σ2 (K) ⊂ σ1 (K) ⊂ Ω. Tower law implies that

[K : k] = [σ1 (K) : k] = [σ1 (K) : σ2 (K)] [σ2 (K) : k] = [σ1 (K) : σ2 (K)] [K : k] .

So [σ1 (K) : σ2 (K)] = 1, so σ1 (K) = σ2 (K). Take σ = σ1−1 ◦ σ2 . σ is clearly bijective and
it is more or less obvious that σ ∈ Emk (K, K).
Lecture 11
Equivalently, k ⊂ K is normal if for all K ⊂ Ω, for all σ ∈ Emk (K, Ω), σ (K) ⊂ K. Friday
01/02/19
Ω
⊂
⊂
σ
K σ (K)
⊂
⊂

Remark. We will see that k ⊂ K is normal if and only if there exists f (X) ∈ K [X] such that K is a splitting
field of f .
Lemma 4.2. Suppose k ⊂ K is normal. Consider k ⊂ L ⊂ K. Then also L ⊂ K is normal.
Proof. If σ ∈ EmL (K, Ω), then σ ∈ Emk (K, Ω).

Warning.
• It is not true in general that k ⊂ K is normal implies that k ⊂ L is normal. For example, let
√  √ √ 
3 3
k=Q⊂Q 2 ⊂Q 2, −3 = K.
√
3


k ⊂ K is normal because it is a splitting field but Q ⊂ Q 2 is not normal.
• Suppose k ⊂ L is normal and L ⊂ K is normal. This does not imply k ⊂ K is normal. This will be in
an example sheet.

14
M3P11 Galois Theory 4 Galois correspondence

Definition 4.3. k ⊂ K is separable if for all k ⊂ K1 ⊂ K2 ⊂ K, if K1 6= K2 , there exist k ⊂ Ω and

embeddings x ∈ Emk (K1 , Ω) and y1 , y2 ∈ Emk (K2 , Ω) such that

K Ω
⊂

⊂
⊂

y1
K2 y1 (K2 ) y2 (K2 )
y2 .

⊂
⊂ ⊂
x
K1 x (K1 )

⊂
⊂

That is, y1 |K1 = y2 |K1 = x but y1 6= y2 .

Slogan is that embeddings separate fields. We will see that
• in characteristic zero everything is separable, and
• in characteristic p we will have good ways to decide if something is separable.
Lemma 4.4. Suppose k ⊂ K ⊂ L. Then k ⊂ L is separable if and only if k ⊂ K and K ⊂ L are separable.
Proof.
=⇒ Obvious. K ⊂ K1 ⊂ K2 ⊂ L leads to k ⊂ K1 ⊂ K2 ⊂ L.
⇐= I will do later.

Theorem 4.5 (Fundamental theorem of Galois theory, Galois correspondence). Let k ⊂ K be normal and
separable. Let G = Emk (K, K). Then there is a one-to-one correspondence

{k ⊂ L ⊂ K} ↔ {H ≤ G}
L 7→ L† = {σ ∈ G | ∀λ ∈ L, σ (λ) = λ} .
∗
H = {λ ∈ K | ∀σ ∈ H, σ (λ) = λ} ←[ H
†
∗
Proof. We show that for all H ≤ G, (H ∗ ) = H and for all k ⊂ L ⊂ K, L† = L. We already did the

∗ We just prove the latter now. Note that L ⊂ K is normal and separable so all I need to show is
former.
k † = k, that is k = G∗ is the fixed field of G. That is, if λ ∈
/ k, there exists σ : K → K in G such that
σ (λ) 6= λ. By separability, there exists Ω and x1 6= x2 ∈ Emk (k (λ) , Ω) such that

K
x
f 1 ,f
x2

k (λ) x1 ,x2 Ω ,

so x1 (λ) 6= x2 (λ). Two steps.

• There exist x f2 : K → Ω extending x1 , x2 : k (λ) → Ω, by the following lemma.
f1 , x
• Because k ⊂ K is normal there exists σ ∈ Emk (K, K) such that x f1 ◦ σ then clearly σ (λ) 6= λ.
f2 = x

15
M3P11 Galois Theory 4 Galois correspondence

Lemma 4.6. Suppose k ⊂ K is normal. Then for all towers k ⊂ F ⊂ K ⊂ Ω, the natural restriction

ρ : Emk (K, Ω) → Emk (F, Ω)

is surjective.
Lecture 12
Lemma 4.6 says for all σ ∈ Emk (F, Ω), there exists σ
e ∈ Emk (K, Ω) such that σ
e|F = σ. Tuesday
05/02/19
K
σ
e

F σ Ω .

Proof. We know that there exists Ω

e as follows.

φ2
K Ω
e
ψ
σ
Ω .
e
F σ

There are two K ⊂ Ω,

e
ψ
φ1 : K ⊂ Ω ,−
→ Ω,
e φ2 : K ,→ Ω.
e

Because k ⊂ K is normal φ2 (K) ⊂ φ1 (K) ⊂ ψ (Ω). That proves that σ

e exists.
Corollary 4.7. Suppose k ⊂ K is normal. Then for all towers k ⊂ F ⊂ K ⊂ Ω,

Emk (F, K) → Emk (F, Ω)

is also surjective.

Corollary 4.7 states that for all σ ∈ Emk (F, Ω), σ (F ) ⊂ K.

σ
K e
σ
e (K)
.

F σ σ (F )

Proof. This clearly follows from Lemma 4.6. σ (F ) ⊂ σ

e (K) ⊂ K by definition of normal.

16
 M3P11 Galois Theory 5 Normal extensions

5 Normal extensions
Theorem 5.1. For finite k ⊂ K, the following are equivalent.
1. For all f ∈ k [X] irreducible either f has no root in K or f splits completely in K.
2. There exists f ∈ k [X] not necessarily irreducible such that K is a splitting field of f .
3. k ⊂ K is normal.

Proof.
1 =⇒ 2 There are λ1 , . . . , λm ∈ K such that K = k (λ1 , . . . , λm ). For all i let fi ∈ k [X] be the minimal
polynomial of λi . fi is irreducible and by 1 it splits completely. K is the splitting field of
m
Y
f (X) = fi (X) .
i=1

2 =⇒ 3 Suppose K ⊂ Ω. Let σ : K → Ω be another embedding. For all λi , σ (λi ) is a root of f , so

σ (λi ) ⊂ K hence σ (K) ⊂ K.

3 =⇒ 1 Let f (X) ∈ k [X] be irreducible. Suppose there exists λ ∈ K such that f (λ) = 0. Let Ω be a
splitting field of f (X) ∈ K [X]. Let µ ∈ Ω be a root of f . There exists a unique σ ∈ Emk (k (λ) , Ω)
such that σ (λ) = µ.
K

σ
F = k (λ) σ (F ) ⊂ Ω 3 µ .

k
By Corollary 4.7, σ (F ) ⊂ K, so µ ∈ K.

(Exercise: prove that any two splitting fields of f ∈ k [X] are k-isomorphic, not necessarily in a unique way) Lecture 13
Thursday
Proposition 5.2. Let k ⊂ L be a field extension. Then there exists a tower k ⊂ L ⊂ K such that k ⊂ K is 07/02/19
normal.
Proof. We use normal if and only if splitting field. Pick λ1 , . . . , λn ∈ L such that L = k (λ1 , . . . , λn ). Let
fi ∈ k [X] be the minimal polynomial of λi over k. Let K be the splitting field of
n
Y
f= fi ∈ L [X] .
i=1

Claim that K is the splitting field of f over k. Key point is argue that K is generated by the roots of f over
k.

17
M3P11 Galois Theory 6 Separable polynomials

6 Separable polynomials
Definition 6.1. A polynomial f ∈ k [X] is separable if it has n = deg (f ) distinct roots in any field k ⊂ K
such that f ∈ K [X] splits completely.
Remark. It is not completely obvious that this definition is independent of K. To see this, use the fact that
any two splitting fields are isomorphic.
Example.
p p
• Let k = Fp = Z/pZ. Then X p − a = (X − a) is not separable, since in characteristic p, (a + b) =
ap + bp .
• Let k = Fp (t). Then X p − t is an irreducible polynomial. Why? Let

Fp (t) [u]
K= = Fp (u) .
hup − ti
p
In K [X], X p − t = (X − u) .

For all k, define the derivation as

D : k [X] −→ k [X]
,
X n 7−→ nX n−1

and extend linearly to all of k [X]. The following are some properties.
• D is k-linear, that is for all λ, µ ∈ k, for all f, g ∈ k [X],

D (λf + µg) = λDf + µDg.

• Leibnitz rule, that is for all f, g ∈ k [X],

D (f g) = f Dg + gDf.

Most important thing to know in characteristic p, if p | n then DX n = nX n−1 = 0. If Df = 0 that does not
mean f is constant. This just means that there exists h ∈ k [X] such that f (X) = h (X p ).
Proposition 6.2. f (X) ∈ k [X] is separable if and only if gcd (f, Df ) = 1.
In R [x], f is inseparable if and only if there exists a multiple root, a critical point, which is a root of Df . Lecture 14
Friday
Lemma 6.3. Let f, g ∈ k [X] and c = gcd (f, g) in k [X]. Let k ⊂ L be an extension. Then c = gcd (f, g) in 08/02/19
L [X].
Proof. Indeed, if c | f and c | g in k [X] then also in L [X]. We also know that there exists φ, ψ ∈ k [X] such
that
f φ + gψ = c (7)
in k [X], and hence also in L [X]. Suppose that u | f and u | g in L [X], so u | c in L [X] by (7).

18
M3P11 Galois Theory 6 Separable polynomials

Proof of Proposition 6.2. Let k ⊂ L be any field where f splits completely. We can do the proof in L [X].
That is, we may assume that f splits completely, so
Y
f (X) = (X − λi ) .
i

2
⇐= Assume for a contradiction that f is not separable then f (X) = (X − λ) g (X). Then
2
Df (X) = 2 (X − λ) g (X) + (X − λ) Dg (X) = (X − λ) (2g (X) + (X − λ) Dg (X)) .

That is, (X − λ) | f and (X − λ) | Df , so gcd (f, Df ) 6= 1.

=⇒ For all i 6= j, λi 6= λj .  
j
X Y
Df =  (X − λj ) .
i=1 j6=i

Claim that for all i, (X − λi ) - Df . I hope you see this. This shows gcd (f, Df ) = 1.

Theorem 6.4. f ∈ k [X] irreducible is inseparable if and only if

• ch (k) = p > 0, and
• there exists h ∈ k [X] such that f (X) = h (X p ).
Proof. Indeed f is inseparable if and only if gcd (f, Df ) 6= 1, if and only if Df = 0, since f is irreducible so
gcd (f, Df ) 6= 1 if and only if f | Df , and deg (Df ) < deg (f ).
Definition 6.5. A field k in ch (k) = p > 0 is perfect if for all a ∈ k there exists b ∈ k such that bp = a.
Proposition 6.6. If k is perfect then f ∈ k [X] is irreducible implies that f (X) is separable.
Proof. If f were inseparable then f (X) = h (X p ). For all i, finding bpi = ai ,

h (X) = X n + a1 X n−1 + · · · + an = X n + bp1 X n−1 + · · · + bpn .

Thus
p
f (X) = h (X p ) = X n + b1 X n−1 + · · · + bn ,
so f is not irreducible.
Example. All finite fields are perfect. Suppose F is a finite field. Then ch (F ) = p > 0 so Fp ⊂ F therefore
[F : Fp ] = n < 0. dimFp (F ) = n < ∞, so F ∼
n
= (Fp ) as a vector space over Fp , so F has pn elements. The
n n
group F × = F \ {0} has pn − 1 elements. So for all a ∈ F × , ap −1 = 1. For all a ∈ F , ap = a, so
 n−1 p
ap = a,

and this shows F is perfect.

Definition 6.7. Consider k ⊂ K. An element a ∈ L is separable over k if the minimal polynomial
f (X) ∈ k [X] of a is a separable polynomial.
Lecture 15
Lecture 15 is a problem class. Tuesday
Lecture 16 is a problem class. 12/02/19
Lecture 17 is a class test. Lecture 16
Thursday
14/02/19
Lecture 17
Friday
19 15/02/19
M3P11 Galois Theory 7 Separable degree

7 Separable degree
Lecture 18
Definition 7.1. Let k ⊂ K. Choose K ⊂ Ω such that k ⊂ Ω is normal. Define the separable degree as Tuesday
19/02/19
[K : k]s = |Emk (K, Ω)| .

Remark. [K : k]s does not depend on K ⊂ Ω. Suppose k ⊂ Ω1 and k ⊂ Ω2 are normal. Then there exists a
e such that Ω1 ⊂ Ω
bigger field Ω e and Ω2 ⊂ Ω.
e Then
 
Emk (K, Ω1 ) = Emk K, Ω
e = Emk (K, Ω2 ) ,

by Corollary 4.7 a while ago,

Ω1
σ

⊂
e

K e .
Ω
σ

⊂
k
Remark. We can restate the definition of separable extension. Recall that k ⊂ K is separable if for all towers
k ⊂ K1 ⊂ K2 ⊂ K, there exist Ω, y : K1 → Ω, x1 , x2 : K2 → Ω such that x1 6= x2 and x1 |K1 = x2 |K2 = y, so

K2
x1 ,x2
⊂

K1 y Ω ,
⊂

that is [K2 : K1 ]s 6= 1. Thus k ⊂ K is separable if for all towers k ⊂ K1 ⊂ K2 ⊂ K,

[K2 : K1 ]s = 1 =⇒ K1 = K2 .

Theorem 7.2 (Tower law). For all k ⊂ K ⊂ L,

[L : k]s = [L : K]s [K : k]s .

Proof. Choose L ⊂ Ω and k ⊂ Ω normal, so

L
y
⊂

K Ω .
x= y|K
⊂

Study
ρ : Emk (L, Ω) → Emk (K, Ω) .
ρ is surjective. For all x ∈ Emk (K, Ω), there exists y ∈ Emk (L, Ω) such that y|K = x, so ρ−1 (x) =
EmK (L, Ω). Then
X X
[L : k]s = |Emk (L, Ω)| = ρ−1 (x) = [L : K]s = [L : K]s [K : k]s .
x∈Emk (K,Ω) x∈Emk (K,Ω)

20
M3P11 Galois Theory 8 Separable extensions

8 Separable extensions
Recall that for k ⊂ K, we said a ∈ K is separable over k if the minimal polynomial f (X) ∈ k [X] of a is a
separable polynomial.
Theorem 8.1. k ⊂ K is separable if and only if [K : k]s = [K : k].
Proof.

Step 1. [K : k]s = [K : k], so k ⊂ K is separable. Recall [K : k]s ≤ [K : k]. Statement follows from two
tower laws for k ⊂ K1 ⊂ K2 ⊂ K, so [K2 : K1 ]s = [K2 : K1 ]. So if [K2 : K1 ]s = 1 then [K2 : K1 ] = 1
then K1 = K2 .
Step 2. Suppose that k ⊂ k (a) is separable then a is separable. Let f (X) ∈ k [X] be the minimal polynomial.
Suppose for a contradiction that it is not a separable polynomial. f is irreducible and f | Df , so
Df ≡ 0, so ch (k) = p and there exists h (X) ∈ k [X] irreducible such that f = h (X p ). Let b = ap
and consider k ⊂ k (b) ⊂ k (a). a is a root of X p − b ∈ k (b) [X]. Then

p deg (h) = [k (a) : k] = [k (a) : k (b)] [k (b) : k] = [k (a) : k (b)] deg (h) ,
p
so [k (a) : k (b)] = p. Thus X p − b = (X − a) is the minimal polynomial of a over k (b), so
[k (a) : k (b)]s = 1 contradicts step 1 and two tower laws.
Step 3. For k ⊂ k (a), k ⊂ k (a) is separable, so [k (a) : k]s = [k (a) : k]. This is obvious from step 2. Then
[k (a) : k] is the degree of the minimal polynomial and [k (a) : k]s is the number of roots of minimal
polynomial.

Step 4. End of proof, by a familiar method. Let us do the general case by induction on [K : k]. If k = K then
there is nothing to prove. Otherwise pick a ∈ K \ k. We know that both k ⊂ k (a) and k (a) ⊂ K
are separable. Then [K : k (a)] < [K : k] by tower law, hence by induction [K : k (a)]s = [K : k (a)].
We also know [k (a) : k]s = [k (a) : k]. Two tower laws imply that [K : k]s = [K : k].

Lecture 19
Thursday
Corollary 8.2. For all towers k ⊂ K ⊂ L, if k ⊂ K and K ⊂ L are separable then k ⊂ L is separable. 21/02/19
Corollary 8.3. k ⊂ K is separable if and only if for all a ∈ K, a is separable over k.
Proof. Suppose k ⊂ K is separable. Pick a ∈ K then k ⊂ k (a) is also separable. By step 2 last time, a is
separable. Conversely, suppose for all a ∈ K, a is separable over k. Pick a ∈ K \ k. I claim k ⊂ k (a) is
separable. Then

[k (a) : k]s = |{roots of minimal polynomial f }| = deg (f ) = [k (a) : k] ,

so k ⊂ k (a) is separable. We want to show that k (a) ⊂ K is separable, by the following lemma.
Lemma 8.4. Let k ⊂ L ⊂ K. For λ ∈ K, λ is separable over k implies that λ is separable over L.

Proof. The minimal polynomial over L divides the minimal polynomial over k.

21
M3P11 Galois Theory 9 Biquadratic polynomials

9 Biquadratic polynomials
Let
√ √
q 
p p
K⊂K a± b = L, c = a2 − b, β= b∈
/ K, α= a + β ∈ L, α0 = a − β ∈ L.

We know that ±α, ±α0 are the roots of

f (X) = X 4 − 2aX 2 + c. (8)

This time we are not assuming (8) is irreducible. Let

√
δ = α + α0 , δ 0 = α − α0 , γ = αα0 = c.

Then
δ + δ0 δ − δ0
γ 2 = c, δ 2 = 2 (a + γ) , δ 02 = 2 (a − γ) , δδ 0 = 2β, α= , α0 = ,
2 2
and ±δ, ±δ 0 are the roots of
g (Y ) = Y 4 − 4aY 2 + 4b.
L is the splitting field of g. Assume
1. ch (K) 6= 2, and
2. b is not a square in K, that is [K (β) : K] = 2.
Claim that the extension K ⊂ L is separable. It is the splitting field of f (X). I need to check gcd (f, Df ) = 1.

Df = 4X 3 − 4aX = 4X X 2 − a .

√
f, Df have no common roots, since X = 0 is not a root of f and X = ± a is not a root of f , since b 6= 0.
Theorem 9.1. Assume 1 and 2.
1. Suppose bc, c are not squares. Then

[L : K] = 8, G = D8 ,

and f (X) is irreducible.

2. Suppose bc is a square, so c is not a square. Then

[L : K] = 4, G = C4 ,

and f (X) is irreducible.

3. Suppose c is a square, so bc is not a square. Then
• either 2 (a + γ) , 2 (a − γ) both not squares in K, then

[L : K] = 4, G = C2 × C2 ,

and f (X) is irreducible.

• or one of 2 (a + γ) , 2 (a − γ) is a square in K, but not the other, then

[L : K] = [K (β) : K] = 2, G = C2 ,

and f (X) is reducible.

22
M3P11 Galois Theory 9 Biquadratic polynomials

√  Lecture 20
Lemma 9.2. Let B ∈ F and A ∈ F be not square in F . If B is square in F A then either B is square Friday
in F or AB is square in F . 22/02/19
 √ 2
√
Proof. Let B = x + y A = x2 + Ay 2 + 2xy A. Then
2
• either x = 0, so B = Ay 2 , so AB = (Ay) is square in F ,
• or y = 0, so B = x2 , so B = x2 is square in F .

Proof of Theorem 9.1.

1. Strategy is [K (α) : K (β)] = [K (α0 ) : K (β)] = 2 and K (α) 6= K (α0 ).

L
2 2

K (α) K (α0 )
.
2 2
K (β)
2

K
• Key idea is that suppose α ∈ K (β) = {x + yβ | x, y ∈ K}. There exist x, y ∈ K such that
2 2
α = x + yβ. Then (x + yβ) = a + β and (x − yβ) = a − β, so

2 2
K 3 x2 − y 2 b = ((x + yβ) (x − yβ)) = (a + β) (a − β) = a2 − b = c,
so c is a square in K. Similarly, α0 ∈ K (β), so α ∈ K (β), so c is a square in K. c is not a square
therefore α ∈/ K (β) and α0 ∈
/ K (β), that is [K (α) : K (β)] = [K (α0 ) : K (β)] = 2.
√
• Suppose for a contradiction α0 ∈ K (α), that is a − β is square in K (α) = K (β) a + β . Apply

Lemma 9.2 with

F = K (β) , A = a + β, B = a − β.
Then either B is square in F , a contradiction, or AB is square in F , that is (a + β) (a − β) =
a2 − b = c is a square in K (β). Apply Lemma 9.2 again with
F = K, A = b, B = c.
Then either c is square in K or bc is square in K, which are contradictions. Thus K (α) 6= K (α0 ).
|G| = 8. Let σ ∈ G. Then
• either σ (β) = β, so there are four possibilities σ (α) = ±α and σ (α0 ) = ±α0 ,
• or σ (β) = −β, so there are four possibilities σ (α) = ±α0 and σ (α0 ) = ±α, since σ y 2 − a − β =

y 2 − a + β.
Because |G| = 8 all these permutations are elements of G. Thus G = D8 is the group of symmetries of
the square
α0

−α xσ α lτ .

−α0

23
M3P11 Galois Theory 9 Biquadratic polynomials

The lattice of subgroups is

hei

hτ i σ2 τ σ2 hστ i σ3 τ
.

σ2 , τ hσi σ 2 , στ

G
The lattice of subfields is
L
2 2
2 2
2

K (α) K (α0 ) K (β, γ) K (δ) K (δ 0 )

2 2 2 .
2 2 2 2
K (β) K (βγ) K (γ)
2
2 2
K
Lecture 21
2. K (βγ) = K, so K (β) = K (γ). β ∈ / K. Suppose a + β is square in K (β). There exist x, y ∈ K such Tuesday
2 2
that a + β = (x + yβ) = x2 + y 2 b + 2xyβ, so (x − yβ) = a − β, then 26/02/19

2 2
K 3 x2 − by 2 = ((x + yβ) (x − yβ)) = (a + β) (a − β) = a2 − b = c,
so c is square in K, a contradiction.

L = K (α) = K (α0 ) = K (δ) = K (δ 0 )

2

K (β) = K (γ) = K (β, γ) .

2

K = K (βγ)

Claim that G = C4 . What’s different? αα0 = γ and βγ ∈ K. Let σ ∈ G. If σ (β) = β then σ (α) = ±α.
• σ (α) = α gives σ (α0 ) = α0 , and
• σ (α) = −α gives σ (α0 ) = −α0 .
If σ (β) = −β then σ (α) = ±α0 .
• σ (α) = α0 gives σ (α0 ) = −α, and
• σ (α) = −α0 gives σ (α0 ) = α.

α0

−α xσ α .

−α0
Thus G = C4 .

24
M3P11 Galois Theory 9 Biquadratic polynomials

p 


3. L = K 2 (a ± γ) is the splitting field of g (Y ) = Y 2 − 2a − 2γ Y 2 − 2a + 2γ .

L = K (δ, δ 0 )

K (β) K (δ) K (δ 0 ) .
≤2
2 ≤2
K

If [L : K] = 4 then G = C2 × C2 . Note that [L : K] cannot be 1. Can it be [L : K] = 2? At least one

of 2 (a + γ) and 2 (a − γ)
p  in K. Suppose 2 (a − γ) is not square in K. Can it be that
is not square
2 (a + γ) is square in K 2 (a − γ) = K (δ 0 )? By Lemma 9.2

• either 2 (a + γ) is square in K, which is possible,

• or 2 (a + γ) 2 (a − γ) = 4 a2 − c = 4b is square in K, which is impossible.
Conclusion is
• either [L : K] = 4, f (X) is irreducible, and G = C2 × C2 ,
• or one of 2 (a + γ) or 2 (a − γ) is square in K but not the other, [L : K] = 2, f (X) is not
irreducible, and G = C2 .

Example. All over Q.

• Let
√
q 
4
f (X) = X − 2, L=Q ± 2 .

Then [L : Q] = 8 and G = D8 .
• Let
√
q 
4 2
f (X) = X − 4X + 2, L=Q 2+ 2 .

Then [L : Q] = 4 and G = C4 .
• Let    √ 
2πi
f (X) = X 4 − X 2 + 1, L = Q e 12 = Q i, 3 .

Then [L : Q] = 4 and G = C2 × C2 .
• Let
√ √ √  √ √ 
q 
f (X) = X 4 − 10X 2 + 1, L=Q 5+2 6 =Q 2+ 3 =Q 2, 3 .

Then [L : Q] = 4 and G = C2 × C2 .

• Let
√ √ 
q 
f (X) = X − 6X + 1 = X − 2X − 1 X 2 + 2X − 1 ,
4 2 2



L=Q 3+2 2 =Q 2 .

Then [L : Q] = 2 and G = C2 .

25
M3P11 Galois Theory 10 Finite fields

10 Finite fields
Lecture 22
If F is finite, then it has ch (F ) = p for some prime p. Then Fp ⊂ F . Because F is finite, it is a finite Thursday
dimensional vector space over Fp . As a vector space F ∼
m
= (Fp ) where m = dimFp (F ) = [F : Fp ], so |F | is a 28/02/19
power of p.
Theorem 10.1. Fix a prime p > 0. Then for all m ∈ Z≥1 , there exists a unique, up to non-unique
isomorphism, finite field with q = pm elements. Notation is Fq . Moreover,

Z
G = Gal (Fq /Fp ) = .
mZ
Proof. Suppose |F | = q. Then F × = F \ {0} is a group with q − 1 elements. That is, if λ ∈ F \ {0} then
λq−1 = 1. Y
Fp [X] 3 X q−1 − 1 = (X − λ) ∈ Fq [X] .
λ∈F \{0}

q−1
Every such field is a splitting field of X − 1. Any two splitting fields are isomorphic. This does the
uniqueness part. As for the existence part, let F be a splitting field over Fp of f (X) = X q−1 − 1 ∈ Fp [X].
Let us prove that F has q elements. Fp is a perfect field, so for all λ ∈ Fp there exists µ ∈ Fp such that
µp = λ. In particular f (X) has q − 1 distinct roots in F . Let us call them λ1 , . . . , λq−1 . Claim that

F 0 = {0, λ1 , . . . , λq−1 }

is a field, then clearly F 0 = F . We need to show that

• F is closed under addition,

• F is closed under multiplication, and

• things in F \ {0} have inverses.
F is closed under multiplication and inverses since for all n, {λ | λn = 1} is a group. F is closed under
q
addition since for all a, b ∈ F , (a + b) = aq + bq , for example
     
p p p p−1 p p
(a + b) = a + a b + ··· + abp−1 + bp , ∀1 ≤ k ≤ p − 1, p .
1 p−1 k

Claim that the function

F : Fq −→ Fq
a 7−→ ap
is a field automorphism, that is F ∈ G, of order exactly m. It is a field automorphism, since

p
a  a p ap F (a)
F (ab) = (ab) = ap bp = F (a) F (b) , F = = p
= ,
b b b F (b)
p
F (a + b) = (a + b) = ap + bp = F (a) + F (b) , F (1) = 1, F (0) = 0.
Certainly
F m = F ◦ · · · ◦ F = id,
k k
since for all λ ∈ Fq , λq = λ. Otherwise if order is k < m then for all λ ∈ Fq , λp = λ, so X p − X has q > pk
roots, a contradiction.

26
M3P11 Galois Theory 11 Symmetric polynomials

11 Symmetric polynomials
Lecture 23
Consider Friday
f (X) = (X − x1 ) . . . (X − xn ) = X n − σ1 X n−1 + · · · ± σn ∈ K (x1 , . . . , xn ) [X] , 01/03/19
where X X
σ1 = σ1 (x1 , . . . , xn ) = xi , σ2 = σ2 (x1 , . . . , xn ) = xi xj , ....
i≤i≤n i≤i≤j≤n

Here σ1 ∈ K [x1 , . . . , xn ] are the elementary symmetric polynomials. Let

Y Y 2
δ= (xi − xj ) , ∆ = δ2 = (xi − xj ) .
roots of f roots of f

Definition 11.1. σ ∈ K [x1 , . . . , xn ] is symmetric if and only if for all g ∈ Sn

σ xg(1) , . . . , xg(n) = σ (x1 , . . . , xn ) .

Example. Consider a degree two polynomial

(X − x1 ) (X − x2 ) = X 2 − σ1 X + σ2 .


• δ = x1 − x2 is not symmetric, for g = 1 2 ,


δ xg(1) , xg(2) = δ (x2 , x1 ) = x2 − x1 = −δ (x1 , x2 ) .

• But
2 2
∆ = δ 2 = (x1 − x2 ) = (x1 + x2 ) − 4x1 x2 = σ12 − 4σ2
is symmetric.
Example. Let
f (X) = X 3 − σ1 X 2 + σ2 X − σ3 .
The plan is to write an invariant telling us when a cubic has repeated roots.

δ = (x1 − x2 ) (x1 − x3 ) (x2 − x3 )

∼ C2 . Can I write δ 2 = ∆ as a


is not symmetric under S3 , but it is invariant under A3 = 1 2 3 =
polynomial in
σ1 = x1 + x2 + x3 , σ2 = x1 x2 + x1 x3 + x2 x3 , σ3 = x1 x2 x3 ?
Yes,
∆ = σ12 σ22 − 4σ13 σ3 − 4σ23 + 18σ1 σ2 σ3 − 27σ32 .
For X 3 + 3pX + 2q,
∆ = −22 33 p3 + q 2 .

The exact expression is totally relevant.

Can we find a formula for discriminant of a degree n polynomial? It is a general fact that all symmetric
polynomials are polynomials in the elementary symmetric polynomials, so
Sn
K [x1 , . . . , xn ] = K [σ1 , . . . , σn ] ⊂ K (σ1 , . . . , σn ) .

27
M3P11 Galois Theory 11 Symmetric polynomials

Theorem 11.2. Consider a degree n separable polynomial

f (X) = X n + a1 X n−1 + · · · + an ∈ k [X] .

Let k ⊂ L be the splitting field of f . Then G ⊂ An if and only if ∆ is a square in k.

By Galois theory ∆ ∈ k, because ∆ is symmetric, that is Sn -invariant, and hence G-invariant.

Proof. G ⊂ An if and only if δ is G-invariant, if and only if δ ∈ k.
Remark.
• We know K ⊂ L is a normal and separable splitting field of f ∈ K [X] implies that G ⊂ Sn .

• If in addition f ∈ k [X] is irreducible then G is transitive, that is for all λ and µ roots of f there exists
σ ∈ G such that σ (λ) = µ.
Theorem 11.3. Consider an irreducible cubic polynomial

X 3 − σ1 X 2 + σ2 X − σ3 ,

and k ⊂ L be the splitting field then G = S3 iff ∆ is not square in k, and G = A3 = C3 iff ∆ is square in k.
Example. For K = Q,
f (X) ∆ G
X3 − X − 1 −23 S3
X 3 − 2X − 1 5 S3
X 3 − 3X − 1 81 A3 .
X 3 − 4X − 1 229 S3
X 3 − 5X − 1 473 S3
X 3 − 6X − 1 837 S3

28
M3P11 Galois Theory 12 Irreducible polynomials

12 Irreducible polynomials
Lecture 24
Proposition 12.1. Suppose Tuesday
f (X) = a0 + · · · + ad X d ∈ Z [X] 05/03/19
has a root p/q ∈ Q with gcd (p, q) = 1, then p | a0 and q | ad .
Example.
• X 5 − 5 has no rational roots. Note that this does not show X 5 − 5 is irreducible over Q.
• X 3 − 2 is irreducible over Q.

Proof. Let
   d
p p
f = a0 + · · · + ad = 0.
q q
Multiplying by q d ,
a0 q d + · · · + ad pd = 0.
Then
q | a0 q d + · · · + ad−1 qpd−1 ,


so q | ad pd . Note that if c | ab and gcd (c, a) = 1 then c | b. Then gcd q, pd = 1, so q | ad . Similarly
p | a0 .
Remark. If K is a field then K [X] is a Euclidean domain, in particular unique factorisation holds, in
particular K [X] is an integral domain, that is for all a, b ∈ K [X] if ab = 0, then either a = 0 or b = 0.
√
Suppose that you want to show n ∈ Z is prime. Try to factor by all primes p < n.

Example. 97 is prime because 2 - 97, 3 - 97, 5 - 97, 7 - 97.

A question is can such a method work with polynomials f (X) ∈ Q [X]? Yes but we will not go there because
it is not very practical. This method is ok in Fp [X], where p is prime, and this leads to a useful strategy.
Example. Is X 5 − 5 ∈ Q [X] irreducible?

• Is it irreducible in F2 [X]? X 5 − 5 ≡ X 5 + 1 mod 2 and X = 1 is a root, so not irreducible.

• Is it irreducible in F3 [X]? X 5 − 5 ≡ X 5 + 1 mod 3 and X = −1 is a root, so not irreducible.
• It is not irreducible in F5 [X] since X 5 − 5 ≡ X 5 mod 5.

• Is it irreducible in F7 [X]? (Exercise)

29
M3P11 Galois Theory 12 Irreducible polynomials

Lemma 12.2 (Gauss’ lemma). Suppose

f (X) = a0 + · · · + ad X d ∈ Z [X] , gcd (a0 , . . . , ad ) = 1

factorises non-trivially in Q [X]. Then it factors non-trivially in Z [X].

Corollary 12.3. If f (X) is prime in Fp [X], for some p, then it is prime in Q [X].
Proof. Otherwise f (X) factors in Fp [X].
Proof of Lemma 12.2. Suppose f (X) = g (X) h (X) in Q [X] for g (X) , h (X) ∈ Q [X]. There is a c ∈ Z
such that

cf (X) = g 0 (X) h0 (X) , g 0 (X) , h0 (X) ∈ Z [X] , g 0 = λg, h0 = µh, λ, µ ∈ Q. (9)

There is a smallest such c. I claim c = 1. Otherwise there exists p prime such that p | c, so (9) is

0 = g 0 (X) h0 (X) ∈ Fp [X] .

Either p | g 0 (X), that is p divides all coefficients of g 0 , or p | h0 (X), a contradiction.

Corollary 12.4 (Eisenstein).
f (X) = a0 + · · · + ad X d ∈ Z [X]
is irreducible in Q [X] if there exists p prime such that p - ad but p | ai for i < d and p2 - a0 .
Proof. Work in Fp [X]. Then f (X) ≡ ad X d mod p. If f (X) = h (X) g (X) in Z [X], then

h (X) ≡ bk X k mod p, g (X) ≡ cd−k X d−k mod p,

and that means that

h (X) = b0 + · · · + bk X k , p | bi , i < k,
and
g (X) = c0 + · · · + cd−k X d−k , p | ci , i < d − k.
2
Thus p | a0 = b0 c0 , a contradiction.
Example. X 5 − 5 is irreducible by Eisenstein and p = 5.
Lecture 25
Lecture 25 is a problem class. Thursday
Lecture 26 is a problem class. 07/03/19
Lecture 26
Friday
08/03/19

30
M3P11 Galois Theory 13 Reduction modulo prime

13 Reduction modulo prime

Lecture 27
Theorem 13.1. Let f (X) ∈ Z [X] be monic of degree n, Q ⊂ K be the splitting field of f , and G = Tuesday
Gal (K/Q) ⊂ Sn . For p prime, denote by f as f viewed in Fp [X]. If there exists p such that f ∈ Fp [X] has 12/03/19
n distinct roots in a splitting field and
k
Y
f= fi (X) ∈ Fp [X] ,
i=1

with fi ∈ Fp [X] irreducible of degree ni , then there exists σ ∈ G ⊂ Sn of cycle decomposition type

(n1 ) . . . (nk ) .

The plan is to understand the statement and prove it, following Jacobson Basic Algebra I page 301. I want
to use Theorem 13.1 to write down an explicit degree five polynomial f ∈ Z [X] such that G = S5 .
Proposition 13.2. Suppose that r is prime and let G ⊂ Sr be a subgroup. If G contains an r-cycle and
one transposition, then G = Sr .


Proof. Say σ = (1 2 3 4 5) ∈ G and 1 2 ∈ G, by relabelling. Then

σ −i 1 2 σi = 1 − i



2−i

implies that




1 2 , 2 3 , 3 4 , 4 5 ∈ G.
A general fact is that for all n,


1 2 , ..., n−1 n
generate Sn , since a < b implies that





a b = a a+1 a+1 b a a+1 .

The following is the plan.

• Find irreducible monic degree five φ (X) ∈ F2 [X].

• Find irreducible monic degree two ψ (X) ∈ F3 [X].
• Find f ∈ Z [X] such that f ≡ φ in F2 [X] and f ≡ X (X − 1) (X + 1) ψ in F3 [X].
Theorem 13.1 implies that G = S5 .

• The irreducible degree two polynomial in F2 [X] is X 2 + X + 1.

• The irreducible degree two polynomial in F3 [X] are X 2 + 1, and two more.
Claim that
f ≡ φ (X) = X 5 + X 3 + 1 ∈ F2 [X]
is irreducible. Need to check that φ has no root in F2 and φ is not divisible by X 2 + X + 1. Then

f ≡ X (X − 1) (X + 1) X 2 + 1 = X 5 − X ∈ F3 [X] .

Thus
f (X) = X 5 + 3X 3 + 2X + 3.

31
M3P11 Galois Theory 13 Reduction modulo prime

Lecture 28
Definition 13.3. The character of a monoid to K is χ : P → K such that Thursday
14/03/19
• χ (0) = 1, and
• for all p1 , p2 ∈ P , χ (p1 + p2 ) = χ (p1 ) χ (p2 ).
Remark. If K is a field and A is a set then {f : A → K} is a K-vector space.
Theorem 13.4 (Linear independence of characters, Dedekind independence theorem). Let K be a field and
P be a monoid, such as P = N. Any set of distinct non-zero characters

χ1 : P → K, ..., χn : P → K, ...

is linearly independent in the vector space {f : P → K}.

Proof. Assume for a contradiction that

λ1 χ1 + · · · + λn χn = 0. (10)

We may assume (10) to be the shortest. We may assume for all i, λi 6= 0. n ≥ 2, since all are not zero.
There exists p such that χ1 (p) 6= χ2 (p), so

λ1 χ1 (p) χ1 + · · · + λn χn (p) χn = 0. (11)

Because (10) is the shortest then (10) and (11) are multiples of each other, so χ1 (p) = χ2 (p), a contradiction.

Theorem 13.5. Let f (X) ∈ Z [X] be degree n monic, Q ⊂ K be the splitting field of f , G = Gal (K/Q) ⊂
Sn , and λ1 , . . . , λn ∈ K be the roots of f (X). Let p be a prime. Denote by f image of f mod p. Assume f
is separable. Let Fp ⊂ F be a splitting field for f , so f has n distinct roots in F . Let R ⊂ K be the subring
generated by the roots of f , so R = Z [λ1 , . . . , λn ]. Then
1. there exists a ring homomorphism ψ : R → F ,
2. if ψ 0 : R → F is a ring homomorphism then ψ 0 induces a bijection

φ0 : {roots of f (X) in R} → roots of f (X) in F ,



3. ψ 0 : R → F is a ring homomorphism if and only if there exists σ ∈ G such that ψ 0 = ψ ◦ σ.

Lecture 29
Lecture 29 is a class test. Friday
15/03/19
Proof.
Lecture 30
Step 1. R is a finitely generated free Z-module, since R is generated as a Z-module by λe11 , . . . , λenn where Tuesday
0 ≤ ei < n. 19/03/19
Step 2. Let u1 , . . . , ud be a basis of R as Z-module. This is a basis of K as a Q-vector space, so d = [K : Q].
u1 , . . . , ud are clearly linearly dependent over Q. Next let Q ⊂ QR ⊂ K. Then QR is a subring
containing Q, so QR is a field. (Exercise: if K ⊂ L is finite and K ⊂ R ⊂ L then R is a ring implies
that R is a field) QR contains all roots of f , so QR = K, that is u1 , . . . , ud generate K over Q.

32
M3P11 Galois Theory 13 Reduction modulo prime

Step 3. Proof of 1. Let m ⊃ hpi be a maximal ideal in R then Fp ⊂ R/m is a finite field. π : R → R/m gives
n
Y n
Y
f (X) = (X − λi ) 7→ f (X) = (X − π (λi )) ,
i=1 i=1

that is f (X) splits in R/m [X]. R is generated by λi implies that R/m is generated by π (λi ), so
R/m is a splitting field for f ∈ Fp [X], so R/m ∼
= F . Thus

R
π ψ
.
R
∼ F
m

Step 4. Proof of 2. Easy. ψ 0 : R → F gives

n
Y n
Y
f (X) = (X − λi ) 7→ f (X) = (X − π (λi )) ,
i=1 i=1

so {λi } 7→ {π (λi )}.

Step 5. Proof of 3. Converse is obvious. Let σ ∈ G and G = {σ1 , . . . , σN } for rkZ (R) = N = [K : Q].
Consider ψi = ψ ◦ σi : P → F where P = (R \ {0} , ×) is a semigroup. Suppose ψN +1 : P → F is
another character. If for all i, ψN +1 6= ψi , we will derive a contradiction. Let r1 , . . . , rN be a basis
of R as a Z-module. Solve for xi ∈ F , for i = 1, . . . , N , in
(N +1 )
X
xi ψi (rj ) = 0 1 ≤ j ≤ N ,
i=1

which has N equations in (N + 1) variables in F . Since rj is a Z-basis this implies

N
X +1
xi ψj = 0,
i=1

and this contradicts Dedekind.

Proof of Theorem 13.1. Let R and ψ : R → F be as above. Consider F r ∈ Gal (F/Fp ) then ψ 0 = F r ◦ ψ :
R → F is a ring homomorphism. By Theorem 13.5.3 there exists σ ∈ G such that F r ◦ ψ = ψ ◦ σ. Thus

Fr
σ

ψ 
{roots of f in R} roots of f in F .

33