Apuntes Introducion Teoria de Galois

Lectures notes in Introduction to Galois
Theory by Ekaterina Amerik

Ivan Murashko
November 9, 2018
2
Contents
1 Generalities on algebraic extensions 9

1.1 Field extensions: examples . . . . . . . . . . . . . . . . . . . . 9
1.1.1 K-algebra . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1.2 Field extension . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Algebraic elements. Minimal polynomial . . . . . . . . . . . . 13
1.2.1 K [X] / (P ) eld . . . . . . . . . . . . . . . . . . . . . . 13
1.2.2 Algebraic elements . . . . . . . . . . . . . . . . . . . . 15
1.2.3 Minimal polynomial . . . . . . . . . . . . . . . . . . . 15
1.3 Algebraic elements. Algebraic extensions . . . . . . . . . . . . 16
1.4 Finite extensions. Algebraicity and niteness . . . . . . . . . . 18
1.5 Algebraicity in towers. An example . . . . . . . . . . . . . . . 20
1.6 A digression: Gauss lemma, Eisenstein criterion . . . . . . . . 21
2 Stem eld, splitting eld, algebraic closure 25

2.1 Stem eld. Some irreducibility criteria . . . . . . . . . . . . . 25
2.1.1 Stem eld . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.2 Some irreducibility criteria . . . . . . . . . . . . . . . . 27
2.2 Splitting eld . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 An example. Algebraic closure . . . . . . . . . . . . . . . . . . 29
2.3.1 An example of automorphism . . . . . . . . . . . . . . 29
2.3.2 Algebraic closure . . . . . . . . . . . . . . . . . . . . . 30
2.3.3 Ideals in a ring . . . . . . . . . . . . . . . . . . . . . . 31
2.4 Extension of homomorphisms. Uniqueness of algebraic closure 32
3 Finite elds. Separability, perfect elds 35

3.1 An example (of extension)s. Finite elds . . . . . . . . . . . . 35
3.1.1 Finite elds . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Properties of nite elds . . . . . . . . . . . . . . . . . . . . . 37
3.3 Multiplicative group and automorphism group of a nite eld 40
3.4 Separable elements . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Separable degree, separable extensions . . . . . . . . . . . . . 44
3
4 CONTENTS
3.6 Perfect elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Tensor product. Structure of nite K-algebras 49

4.1 Denition of tensor product . . . . . . . . . . . . . . . . . . . 49
4.1.1 Summary for previous lectures . . . . . . . . . . . . . . 49
4.1.2 Tensor product . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Tensor product of modules . . . . . . . . . . . . . . . . . . . . 53
4.2.1 Advantages of the universal property . . . . . . . . . . 53
4.2.2 Several examples of universal property usage . . . . . . 53
4.3 Base change . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Examples. Tensor product of algebras . . . . . . . . . . . . . . 58
4.4.1 Tensor product of A-algebras . . . . . . . . . . . . . . 59
4.5 Relatively prime ideals. Chinese remainder theorem . . . . . . 61
4.6 Structure of nite algebras over a eld. Examples . . . . . . . 63
5 Structure of nite K-algebras continued 67

5.1 Structure of nite K-algebras, examples (cont'd) . . . . . . . . 67
5.2 Separability and base change . . . . . . . . . . . . . . . . . . . 69
5.3 Primitive element theorem . . . . . . . . . . . . . . . . . . . . 74
5.4 Examples. Normal extensions . . . . . . . . . . . . . . . . . . 75
5.4.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.4.2 Normal extensions . . . . . . . . . . . . . . . . . . . . 77
5.5 Galois extensions . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.6 Artin's theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6 Galois correspondence and rst examples 83

6.1 Some further remarks on normal extension. Fixed eld . . . . 83
6.2 The Galois correspondence . . . . . . . . . . . . . . . . . . . . 85
6.3 First examples (polynomials of degree 2 and 3) . . . . . . . . . 87
6.4 Discriminant. Degree 3 (cont'd). Finite elds . . . . . . . . . 88
6.4.1 Discriminant . . . . . . . . . . . . . . . . . . . . . . . 88
6.4.2 Finite elds. An innite degree example . . . . . . . . 90
6.5 Roots of unity: cyclotomic polynomials . . . . . . . . . . . . 92
6.6 Irreducibility of cyclotomic polynomial.The Galois group . . . 94
7 Galois correspondence and rst examples. Examples contin-

ued 97
7.1 Cyclotomic extensions (cont'd). Examples over Q . . . . . . . 97
7.2 Kummer extensions . . . . . . . . . . . . . . . . . . . . . . . . 99
7.3 Artin-Schreier extensions . . . . . . . . . . . . . . . . . . . . . 101
7.4 Composite extensions. Properties . . . . . . . . . . . . . . . . 104
CONTENTS 5
7.5 Linearly disjoint extensions. Examples . . . . . . . . . . . . . 106

7.6 Linearly disjoint extensions in the Galois case . . . . . . . . . 108
7.7 On the Galois group of the composite . . . . . . . . . . . . . . 109
8 Solvability by radicals, Abel's theorem. A few words on re-

lation to representations and topology 113
8.1 Extensions solvable by radicals. Solvable groups. Example . . 113
8.1.1 Extensions solvable by radicals . . . . . . . . . . . . . 113
8.1.2 Solvable groups . . . . . . . . . . . . . . . . . . . . . . 115
8.2 Properties of solvable groups. Symmetric group . . . . . . . . 115
8.3 Galois theorem on solvability by radicals . . . . . . . . . . . . 120
8.4 Examples of equations not solvable by radicals."General equa-
tion" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.5 Galois action as a representation. Normal base theorem . . . . 124
8.6 Relation with coverings . . . . . . . . . . . . . . . . . . . . . . 128
9 Ring extensions, norms and traces, reduction modulo p 131

9.1 Integral elements over a ring . . . . . . . . . . . . . . . . . . . 131
9.1.1 Ring extensions . . . . . . . . . . . . . . . . . . . . . . 132
9.2 Integral extensions, integral closure, ring of integers of a num-
ber eld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
9.2.1 Integral extensions and integral closure . . . . . . . . . 134
9.2.2 Ring of integers in a number eld . . . . . . . . . . . . 136
9.3 Norm and trace . . . . . . . . . . . . . . . . . . . . . . . . . . 137
9.3.1 Norms and traces . . . . . . . . . . . . . . . . . . . . . 137
9.3.2 Theorem about rings of integers . . . . . . . . . . . . . 141
9.4 Reduction modulo a prime . . . . . . . . . . . . . . . . . . . . 143
9.5 Finding elements in Galois groups . . . . . . . . . . . . . . . . 145
Appendices 147
A Course prerequisites 149
A.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
A.2 Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
A.2.1 Cyclic group . . . . . . . . . . . . . . . . . . . . . . . . 153
A.2.2 Group action . . . . . . . . . . . . . . . . . . . . . . . 153
A.2.3 Direct product . . . . . . . . . . . . . . . . . . . . . . 154
A.2.4 Sylow theorems . . . . . . . . . . . . . . . . . . . . . . 156
A.2.5 Abelian group . . . . . . . . . . . . . . . . . . . . . . . 156
A.3 Permutations . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
A.4 Rings and Fields . . . . . . . . . . . . . . . . . . . . . . . . . 162
6 CONTENTS
A.4.1 Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

A.4.2 Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
A.4.3 Polynomial ring K [X] . . . . . . . . . . . . . . . . . . 165
A.4.4 Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.4.5 Characters . . . . . . . . . . . . . . . . . . . . . . . . . 168
A.5 Modules and Vector spaces . . . . . . . . . . . . . . . . . . . . 168
A.5.1 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . 168
A.5.2 Linear algebra . . . . . . . . . . . . . . . . . . . . . . . 170
A.6 Functions aka maps . . . . . . . . . . . . . . . . . . . . . . . . 172
A.6.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . 172
A.6.2 Category theory . . . . . . . . . . . . . . . . . . . . . . 174
A.7 Number theory . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Index 175
Introduction
The document keeps lecture notes on Introduction to Galois theory that was
provided by Ekaterina Amerik (Higher School of Economics) via Coursera.
Each chapter corresponds to one lecture (or one week on Coursera). The
appendix keeps useful info for the course that is absent in it i.e. requirements
that are necessary for the course understanding.
I also tried to make all my comments as footnotes or inside brackets '()'
whenever it was possible.
All not clear (for me) or in-completed proofs are marked as ???
Several examples use GAP [8].
7
8 CONTENTS
Chapter 1
Generalities on algebraic
extensions
We introduce the basic notions such as a eld extension, algebraic element,
minimal polynomial, nite extension, and study their very basic properties
such as the multiplicativity of degree in towers.
1.1 Field extensions: examples

1.1.1 K-algebra
Denition 1.1 (K-algebra). Let K is a Ring and A is a Module over K 1
equipped with an additional binary operation A × A → A that we denote as ·
here. The A is an algebra over K if the following identities hold ∀x, y, z ∈ A
and for every elements (often called as scalar) a, b ∈ K
• Right distributivity: (x + y) · z = x · z + y · z
• Left distributivity: z · (x + y) = z · x + z · y
• Compatibility with scalars: (ax) · (by) = (ab)(x · y)
Example 1.2 (Field of complex numbers C). The eld of complex numbers
C can be considered as a K-algebra over the eld of real numbers R.
1
The denition in the lectures is more common and assumes that K is a ring but later
K will be mostly a Field. Therefore A has to be a Vector space over K in the case.
9
10 CHAPTER 1. GENERALITIES ON ALGEBRAIC EXTENSIONS
1.1.2 Field extension

Denition 1.3 (Field extension). Let K and L are elds. L is an extension
of K if L ⊃ K
and another denition
Denition 1.4 (Field extension). Let K is a eld then L is an extension of
2
K if L is a K-algebra
Why the 2 denitions are equivalent?
Lemma 1.5 (K-algebra and homomorphism). Given a K-algebra is the same
as having Homomorphism f :K→A of rings.
Proof. Really if I have a K-algebra I can dene the Homomorphism f (k) =

k · 1A , where 1A is an identity element of A. Thus k · 1A ∈ A.
And conversely if I have the Homomorphism f : K → A I can dene the
K-algebra structure by setting ka = f (k)a because f (k), a ∈ A and there is a
multiplication dened on A. As result I have a rule for multiplication a vector
(f (k) ∈ A) on a vector (a ∈ A) that is required for K-algebra setup.
Lemma 1.6 (About homomorphism of elds). Any Field homomorphism is

3
Injection.
Proof. Lets proof by contradiction. Really if f (x) = f (y) and x 6= y then
f (x) − f (y) = 0A ,
f (x − y) = 0A ,

−1 x−y
f (x − y)f ((x − y) ) = f = f (1K ) = 1A = 0A
x−y
that is impossible.
There are some comments on the results. We have got that a Homomor-
phism can be set between eld K and its K-algebra. The Homomorphism is
Injection therefore we can allocate a sub-eld A0 ⊂ A for that we will have
the Homomorphism is a Surjection and therefore we have an Isomorphism
between original eld K and a sub-eld A0 . This means that we can say that
the original eld K is a sub-eld for the K-algebra.
2L in the denition is not the same object with L from denition 1.3. Because L in
the denition is a K -algebra i.e. a ring but L in the denition 1.3 is a eld.
3 But the statement is not valid for groups. In that case the homomorphism is injection
if and only if the kernel is trivial and consists of only one element - identity. See theorem
A.136.
1.1. FIELD EXTENSIONS: EXAMPLES 11
Example 1.7 (Field extensions). C is a eld extension for R. R is a eld

extension for Q
Example 1.8 (K -algebra is not a eld). In the example
4
I will show that K
algebra is not a eld. Consider K = R. Vector space A = R2 i.e. A consists
of vectors of the following form

x1
x= ,
x2
where x1 , x2 ∈ R. I will dene the multiplication for L (our K algebra) as

follows
x1 y1 x1 · y 1
· =
x2 y2 x2 · y 2
It can be seen that all requirements of K-algebra are satised

x1 y1 z1
(x + y) · z = + =
x2 y2 z2

(x1 + y1 )z1 x1 z1 + y1 z1
= = =x·z+y·z
(x2 + y2 )z2 x2 z2 + y2 z2
z · (x + y) = z · x + z · y

ax1 by1 abx1 y1
(ax) · (by) = = (ab)(x · y)
ax2 by2 abx2 y2
The multiplication identity element of L is the following

1
1L =
1
The zero is the standard one from vector space

0
0L =
0
We can see that
1 0
= 0L
0 1
i.e. we have 2 divisor of zero which are not zero itself. The elements do not
have invert ones and as result the L is not a eld.
r
From other side if we dene L0 ⊂ L as follows L0 = , where
r
r ∈ R, then we will have that L0 is a eld and L0 ∼
= R.
4 the example was not present in the lectures
Denition 1.9 (Characteristic). If L is a eld there are 2 possibilities
1. 1 + 1 + · · · 6= 0. In this case Z ⊂ L but Z is not a eld therefore L is

an extension of Q. In the case charL = 0
2. 1 + 1 + · · · + 1 = m i=1 1 = 0 for some m ∈ Z. The rst time when

P
it happens is for a prime number i.e. minimal m with the property
is prime. In this case charL = p, where p = min m - the minimal m
(prime) with the property. In this case Z/pZ ⊂ L. The Z/pZ is a eld
denoted by Fp . The L is an extension of Fp .
No other possibilities exist. The Q and Fp are the prime elds. Any eld is
an extension of one of those.
Denition 1.10 (Prime eld). If L is a eld Q ⊂ L, i.e. charL = 0 then Q

is the prime eld for L. Otherwise, if charL = p > 0 then Fp is the prime
eld.
Claim 1.11. Let K [X] Ring of polynomials. The P ∈ K [X] is an Irre-

ducible polynomial. (P ) is an Ideal formed by the polynomial. 5 The set of
residues by the polynomial forms a eld that denoted by K [X] / (P ).
Proof.How we can see it? If Q ∈ K [X] is a polynomial that Q ∈

/ (P )
when Q is prime to P . 6 Then with Bezout (Lemma A.85) we can get
∃A, B ∈ K [X] such that
AP + BQ = 1,
or
BQ ≡ 1 mod P,
thus B is Q−1 in K [X] / (P ).
Claim 1.12. 7
Let K is a eld and a∈K then K [X] / (X − a) is also a
eld and there exists an Isomorphism between the eld and K i.e.
K [X] / (X − a) ∼
=K (1.1)
5
I.e. (P ) = {Q = GP } where G ∈ K [X]
6
As soon as P is irreducible in K [X] then there is only one possibility for Q and P
to have common divisors: if Q = GP where G ∈ K [X] but this is in contradiction with
Q∈ / (P )
7 The claim is not a part of the lectures but it's very usefully in future lectures.
1.2. ALGEBRAIC ELEMENTS. MINIMAL POLYNOMIAL 13
Proof. The K [X] / (X − a) is a eld just because X − a is Irreducible poly-

nomial (see example A.88 and claim 1.11).
For the proof the main statement (1.1) lets consider a polynomial P ∈
K [X] and dene the following Field homomorphism:
φ : K [X] / (X − a) −−−−−−−→ K (1.2)

P (X)→P (a)
The φ dened by (1.2) is Field homomorphism. For the proof of the claim
lets take P1 , P2 ∈ K [X] / (X − a). Clear that
φ (P1 + P2 ) = P1 (a) + P2 (a) = φ (P1 ) + φ (P1 ) .
The same holds with the multiplication:
φ (P1 · P2 ) = P1 (a) · P2 (a) = φ (P1 ) · φ (P1 ) .
That completes the proof of the Field homomorphism.

Lemma About homomorphism of elds (Lemma 1.6) says that any
Field homomorphism is Injection, i.e. φ is an injection.
Next we should show that φ is Surjection it's easy because ∀k ∈ K we
can consider constant polynomial P = k from K [X]. For the polynomial we
will have φ (k) = k .
As result φ is Injection as well as Surjection i.e. φ is Bijection. With the
fact that φ is Homomorphism it will gives us that φ is Isomorphism.
1.2 Algebraic elements. Minimal polynomial

1.2.1 K [X] / (P ) eld
Alternative proof of claim 1.11 i.e. the fact that K [X] / (P ) is the Field as
follows. The (P ) is a Maximal ideal 8 but a quotient by a Maximal ideal is
a Field (see About Quotient Ring and Maximal Ideal (Theorem A.94) ).
We also can say that K [X] / (P ) is an extension of K because it's K-
algebra.
Example 1.13 (K = F2 / (X 2 + X + 1)). Lets consider the following eld

F2 = Z/2Z = {0, 1} in the eld polynomial X 2 + X + 1 is irreducible. It's
very easy to verify it because F2 has only 2 elements that can be (possible) a
root:
02 + 0 + 1 = 1 6= 0
8 To prove that (P ) is a Maximal ideal we have to use Bezout (Lemma A.85) .
and
12 + 1 + 1 = 1 6= 0
The polynomial has the following residues: X̄ = X + (X 2 + X + 1) and
X + 1 = X + 1 + (X 2 + X + 1). Thus the eld F2 / (X 2 + X + 1) consists of
4 elements: {0, 1, X̄, X + 1}.
It's easy to see that the third element (X̄ ) is a root of P (X) = X 2 +X +1:
X̄ 2 + X̄ + 1 = P (X) + (P (X)) = (P (X)) ≡ 0 mod P.
X̄ 2 + X̄ + 1 = 0̄,
therefore
X̄ 2 = −X̄ − 1 = X̄ + 1 = X + 1.
This is because we are in eld F2 where
2 (X + 1) mod 2 = 0
and thus
−X̄ − 1 = X̄ + 1 = X + 1
Also
2
X + 1 = X̄,
and they are inverse each other
X + 1X̄ = 1,
So this is the structure of a eld of four elements. The cardinality of

K = F2 / (X 2 + X + 1) is 4, one writes then K = F4 . Well, this might be
strange at the rst sight, because we only know that K has four elements
and if you write F4 you somehow mean that there is only one eld of four
elements. Well, it is true, there is only one eld of four elements. In fact,
all nite elds of the same cardinality are isomorphic, and we will see it very
shortly (see theorem 3.8).
Arguments for the Example 1.13 can also be found in the following remark
(not part of the lectures)
Remark 1.14 (About quotient eld). Consider an arbitrary eld L (L = F2
at the Example 1.13) and P (X) = X + an−1 X n−1 + an−2 X n−2 + · · · + a0 ∈
n
L [X] is an irreducible polynomial over L. Let α is a root of the polynomial

then
αn + an−1 αn−1 + · · · + a0 = 0,
1.2. ALGEBRAIC ELEMENTS. MINIMAL POLYNOMIAL 15
or
αn = −an−1 αn−1 − · · · − a0 .
Therefore ∀l ≥ n, ∃li ∈ L such that
n−1
X
αk = li αi
i=0
and as result ∀x ∈ L/ (P ) exists xi ∈ L such that

n−1
X
x= xi α i .
i=0
We can say that we have the following basis: {{L}, α, . . . , αn−1 } where {L}
- L basis. For Example 1.13 we have {L} = {F2 } = {0, 1}, α = X̄ and as
result K = F2 / (X 2 + X + 1) will have the following basis {0, 1, X̄}.
1.2.2 Algebraic elements

Denition 1.15 (Algebraic element). Let K ⊂ L and α ∈ L. α is an
algebraic element if ∃P ∈ K [X] such that P (α) = 0. Otherwise the α is
called transcendental.
1.2.3 Minimal polynomial

Lemma 1.16 (About minimal polynomial existence). If α is Algebraic ele-
ment then ∃! P of minimal degree such that P (α) = 0.
unitary polynomial
9
It is irreducible. ∀Q such that Q (α) = 0 is divisible by P
Proof. We know that K [X] is a Principal ideal domain and a polynomial

Q (α) = 0 forms an Ideal: I {Q ∈ K [X] | Q (α) = 0}, so the ideal is gener-
ated by one element: I = (P ). This is an unique (up to constant) polynomial
minimal degree in I .
Lets prove that P is irreducible. If P is not irreducible then ∃Q, R ∈ I
such that P = QR, Q(α) = 0 or R(α) = 0 and degR, Q < degP that
is in contradiction with the denition that P is a polynomial of minimal
degree.
Denition 1.17 (Minimal polynomial). If α is Algebraic element then the

unitary polynomial P of minimal degree such that P (α) = 0 is called minimal
polynomial and denoted by Pmin (α, K).
9 see also About irreducible polynomials (Theorem A.89)
Example 1.18 (Minimal polynomial)

√ .
10
Consider the following minimal
polynomial Pmin (α, Q), where α= 2. In GAP [8] we have
gap> x:=Indeterminate(Rationals,"x");;
gap> alpha:=Sqrt(2);;
gap> MinimalPolynomial(Rationals, alpha);
x^2-2
I.e. Pmin (α, Q) = X 2 − 2.
1.3 Algebraic elements. Algebraic extensions

Denition 1.19. Let K ⊂ L, α ∈ L. The smallest sub-eld contained K
and α denoted by K (α). The smallest sub-ring (or K-algebra) contained K
and α denoted by K [α].
As soon as K [α] is a K-algebra it is a Vector space over K generated by
1, α, α2 , . . . , αn , . . . .
Example 1.20 (C).

C = R (i) = R [i]
C is also a Vector space generated by 1 and i: ∀z ∈ Z it holds z = x + iy
where x, y ∈ R.
Proposition 1.21. The following assignment are equivalent
1. α is algebraic over K
2. K [α] is a nite dimensional Vector space over K

11
3. K [α] = K (α)
Proof.Lets proof that 1 implies 2. If α is algebraic over K then using lemma

Minimal polynomial ∃Pmin (α, K):
Pmin (α, K) = αd + ad−1 αd−1 + a1 α + a0 = 0,
where ak ∈ K . Then
αd = −ad−1 αd−1 − a1 α − a0
10 The example is not a part of the lectures and it uses GAP [8].
11 Contrary to the example 1.8 we see that K -algebra is a eld there.
1.3. ALGEBRAIC ELEMENTS. ALGEBRAIC EXTENSIONS 17
this means that any αn can be represented as a linear combination of -

nite number of powers of α i.e. K [α] generated by 1, α, . . . , αd−1 is a nite
dimensional Vector space.
Lets proof that 2 implies 3. Its enough to prove that K [α] is a eld
because K [α] ⊂ K (α).
Let x 6= 0 ∈ K [α] then lets look at an operation x · K [α] → K [α]. This
is Injection. 12 But the K [α] is nite dimensional Vector space and a Homo-
morphism between 2 vector spaces with the same dimension is Surjection 13
thus ∃y ∈ K [α] such that x · y = 1K[α] . Therefore x is invertable and K [α]
is a Field.
Lets proof that 3 implies 1. Let K [α] is a Field but α is not algebraic.
Thus ∀P ∈ K [X] P (α) 6= 0. The we have an Injection Homomorphism i :
K [X] → K [α] = K (α) which sends P (X) to P (α). 14 But K [X] is not
a eld thus K [α] should not be a eld too that is in contradiction with the
initial conditions. 15
Denition 1.22 (Algebraic extension). L an extension of K is called alge-

braic over K if ∀α ∈ L - α is algebraic over K .
Proposition 1.23. If L is algebraic over K then any K-subalgebra of L is

a Field.
Proof. Let L0 ⊂ L is a subalgebra and let α ∈ L0 . We want to show that α

is invertable. α is algebraic therefore α ∈ K [α] ⊂ L0 ⊂ L and it's invertable.
16
Proposition 1.24. LetK ⊂ L ⊂ M . α ∈ M - algebraic over K then α

algebraic over L and Pmin (α, L) divides Pmin (α, K).
12 If y, z ∈ K [α] and dim K [α] = d < ∞ where d = deg P
min (α, K). Then y =
Pd−1 Pd−1 Pd−1
y
i=0 i α i
and z = z
i=0 i α i
where y ,
i iz ∈ K . We have y − z = i
i=0 (yi − zi ) α 6= 0
if y 6= z (i.e. ∃i : yi 6= zi ) because y − z can be considered as a polynomial of degree
≤ d − 1 < deg Pmin (α, K) and cannot be equal to 0 by minimal polynomial denition.
Continue we have x · (y − z) 6= 0 because it also can be considered as a product of 2
polynomial of degree < d. Thus
x · y 6= x · z
i.e. Injection property is satised.
13 Two vector spaces with same dimension are isomorphic each other (see lemma A.110)
14 And if P (X) 6= 0 then P (α) 6= 0
15 Alternative prove is the following. Let x 6= 0 ∈ K [X] and K [α] is a eld then i(x) is
invertable i.e. ∃y ∈ K [X] : i(x)i(y) = 1 or i(xy) = 1 or nally x - is invertable and K [X]
is a eld but K [X] is a ring and therefore we just got a contradiction.
16 As soon as K [α] = K (α) is a eld then its any element (especially α) is invertable.
17
Proof. It is clear because Pmin (α, K) ∈ L [X].
We can consider the following example as an illustration for proposition

1.24:
Example 1.25. 18
K = R, L = M = C. α = i ∈ M is algebraic over
K =R and therefore using the proposition 1.24 it is algebraic over L = C.
2
Moreover Pmin (α, L) = X − i and it divides Pmin (α, K) = X + 1.
1.4 Finite extensions. Algebraicity and nite-

ness
Denition 1.26 (Finite extension). L is a nite extension of K if dimk L <
∞. dimk L is called as degree of L over K and is denoted by [L : K]
Theorem 1.27 (The multiplicativity formula for degrees). Let K ⊂ L ⊂ M .

Then M is Finite extension over K if and only if M is Finite extension over
L and L is Finite extension over K . In this case
[M : K] = [M : L] [L : K] .
Proof. Let [M : K] < ∞ but any linear independent set of vectors {m1 , m2 , . . . , mn }
over L is also linear independent over K thus
[M : K] < ∞ ⇒ [M : L] < ∞
also L is a vector sub space of M thus if [M : K] < ∞ then [L : K] < ∞.

Let [M : L] < ∞ and [L : K] < ∞ then we have the following basises
• L-basis over M : (e1 , e2 , . . . , en )
• K -basis over L: (ε1 , ε2 , . . . , εd )
Lets proof that ei εj forms a K -basis over M . ∀x ∈ M :

n
X
x= ai ei ,
i=1
17
Thus ∃PL ∈ L [X] such that PL (α) = 0 i.e. α is algebraic over L.
As soon as Pmin (α, K) ∈ L [X] then using About minimal polynomial existence
(Lemma 1.16) one can get that Pmin (α, L) divides Pmin (α, K).
18 The example is not a part of the lectures.
1.4. FINITE EXTENSIONS. ALGEBRAICITY AND FINITENESS 19
where ai ∈ L and can be also written as

d
X
ai = bij εj ,
j=1
where bij ∈ K . Thus

n X
X d
x= bij εj ei ,
i=1 j=1
therefore εj ei = ei εj generates M over K . From the other side we should

check that εj ei linear independent system of vectors. Lets
n d
!
X X X
cij εj ei = cij εj ei ,
i,j i=1 j=1
then ∀i:
d
X
cij εj = 0.
j=1
Thus ∀i, j : cik = 0 that nishes the proof the linear independence. The
number of linear independent vectors is n × d i.e.
[M : K] = [M : L] [L : K] .
Denition 1.28 (K (α1 , . . . , αn )). K (α1 , . . . , αn ) ⊂ L generated by α1 , . . . , αn

is the smallest sub eld of L contained K and αi ∈ L.
Theorem 1.29 (About towers). L is nite over K if and only if L is gen-
erated by a nite number of algebraic elements over K.
Proof. If L is nite then α1 , . . . , αd is a basis. In this case L = K [α1 , . . . , αd ] =
K (α1 , . . . , αd ). Moreover each K [αi ] is nite dimensional thus by proposi-
tion 1.21 αi is algebraic.
From other side if we have a nite set of algebraic elements α1 , . . . , αd
then K [α1 ] is a nite dimensional Vector space over K , K [α1 , α2 ] is a nite
dimensional Vector space over K [α1 ] and so on K [α1 , . . . , αd ] is a nite
dimensional Vector space over K [α1 , . . . , αd−1 ]. All elements are algebraic
thus
K [α1 , . . . , αi ] = K (α1 , . . . , αi )
Then using theorem 1.27 we can conclude that K (α1 , . . . , αd ) has nite di-
mension.
1.5 Algebraicity in towers. An example

Theorem 1.30. K ⊂ L ⊂ M then M Algebraic extension over K if and
only if M algebraic over L and L algebraic over K.
Proof. If α ∈ M is an Algebraic element over K then ∃P ∈ K [X] such that
P (α) = 0 but the polynomial P ∈ K [X] ⊂ L [X] thus α is algebraic over L.
If α ∈ L ⊂ M then α is algebraic over K thus L is algebraic over K .
Let M algebraic over L and L algebraic over K and let α ∈ M . We want
to prove that α is algebraic over K . Lets consider Pmin (α, L) the polynomial
coecients are from L and they (as soon as they count is a nite) generate
a nite extension E over K thus E (α) is nite over E (exists a relation
between powers of α) is nite over K thus α is algebraic over K . 19
√ √
Example 1.31 (Q extension). Q 3 2, 3 algebraic and nite over Q:
√ √ √
3 3
Q⊂Q 2 ⊂Q 2, 3
Minimal polynomial
√
2, Q = x3 − 2.
3
Pmin
√3
√ √ √
2 is generated over Q by 1, 3 2, 3 4 thus Q 3 2 : Q = 3.
Q
√ √ √
But 3 ∈/ Q 3 2 because otherwise Q 3 : Q = 2 must devide
√
Q 3 2 : Q = 3 that is impossible.
2
√3

Therefore x − 3 is irreducible over Q 2 and
√ √
2 = x2 − 3.
3
Pmin 3, Q
h √ √ i
3
Q 2, 3 : Q = 3 · 2 = 6.
Proposition 1.32 (On dimension of extension).
[K (α) : K] = deg (Pmin (α, K)) ,
if α is algebraic.
19 Pd−1
Pmin (α, L) = i=0 li αi where li ∈ L and each li is algebraic over K by algebraic
extension denition 1.22. By theorem 1.29 there are nite number of li and they forms an
algebraic extension E = K (l0 , l1 , . . . , ld−1 ). The E (α) is nite over E and therefore nite
over K . As soon as EP (α) has a nite dimension over K thus there exists a relation for
n
powers of α such that i=0 ki αi = 0 i.e. α is algebraic.
1.6. A DIGRESSION: GAUSS LEMMA, EISENSTEIN CRITERION 21
Proof.If deg (Pmin (α, K)) = d then 1, α, · · · , αd−1 - d independent vectors

and dimension K (α) is d.
Proposition 1.33 (About algebraic closure). If K ⊂ L (L extension of K ).

Consider
L0 = {α ∈ L | α algebraic over K} ,
then L0 sub-eld of L and is called as algebraic closure of K in L.
Proof. We have to prove that if α, β are algebraic then α + β and α · β are

also algebraic. This is trivial because
α + β, α · β ∈ K [α, β]
20
1.6 A digression: Gauss lemma, Eisenstein cri-

terion
What we have seen so far:
• K is a eld, α is an Algebraic element over K if it is a root of a

polynomial P ∈ K [X].
• L is an Algebraic extension over K if ∀α ∈ L: α is an algebraic over K
• L is a Finite extension over K if dimK L < ∞.
• If an extension is nite then it is algebraic
• An extension is nite if and only if it is algebraic and generated by a

nite number of algebraic elements (see theorem 1.29)
• [K [α] : K] = degPmin (α, K) (see proposition 1.32).
How to decide that a polynomial P is irreducible over K ? About polyno-

mial x3 − 2 it is easy to decide that it's irreducible over Q, but what's about
x100 − 2?
20
We also have that K [α, β] is a eld: K [α, β] = K (α, β). Really K [α] = K (α) (see
proposition 1.21). β is algebraic over K and therefore over K (α) thus we can construct
K (α) [β] = K (α, β) by proposition 1.21
Lemma 1.34 (Gauss). Let P ∈ Z [X], i.e. a polynomial with integer coef-
cients, then if P decomposes over Q (P = Q · R, degQ, R < degP ) then it
also decomposes over Z.
Proof. Let P = QR over Q. Then

Q = mQ1 , Q1 ∈ Z [X] ,
R = nR1 , R1 ∈ Z [X] ,
thus
nmP = Q1 R1 .
There exists p that divides mn: p | mn thus in modulo p we have
0 = Q1 R1
but p is prime and the equation is in the eld Fp thus either Q1 = 0 or
R1 = 0. Let Q1 = 0 thus p divides all coecients in Q1 and we can take
Q1
p
= Q2 ∈ Z [X]. Continue for all primes in mn we can get that
P = Qs Rt ,
where Qs , Rt ∈ Z [X].
Example 1.35 (Eisenstein criterion). Lets consider the following polynomial

x100 − 2. It's irreducible. Lets prove it. If it reducible then ∃Q, R ∈ Z [X]
such that
x100 − 2 = QR (1.3)
Lets consider (1.3) modulo 2. In the case we will have
QR ≡ x100 mod 2,
therefore
Q ≡ xk mod 2,
R ≡ xl mod 2,
or
Q = xk + · · · + 2 · m
and
R = xl + · · · + 2 · n
thus
QR = x100 + 4 · nm
that is impossible because n, m ∈ Z and nm 6= − 12 .
1.6. A DIGRESSION: GAUSS LEMMA, EISENSTEIN CRITERION 23
Lemma 1.36 (Eisenstein criterion). Lets P ∈ Z [X] and P = an X n +

an−1 X n−1 + a1 X + a0 . If ∃p - prime such that p - an , p | ai ∀i < n and
p2 - a0 , then P ∈ Z [X] is irreducible.
Proof. the same as for example 1.35.
Note: that both: Gauss (Equation 1.34) and Eisenstein criterion

(Lemma 1.36) are valid by replacing Z with an Unique factorization do-
main R and Q by its factorization eld.
Chapter 2
Stem eld, splitting eld,
algebraic closure
We introduce the notion of a stem eld and a splitting eld (of a polyno-
mial). Using Zorn's lemma, we construct the algebraic closure of a eld and
deduce its unicity (up to an isomorphism) from the theorem on extension of
homomorphisms.
2.1 Stem eld. Some irreducibility criteria

2.1.1 Stem eld
Denition 2.1 (Stem eld). Let P ∈ K [X] is an irreducible Monic polyno-
mial. Field extension E is a stem eld of P if ∃α ∈ E - the root of polynomial
P and E = K [α]. 1
Such things exist, for instance we can take K [X] / (P ). It is a eld

because P is an Irreducible polynomial moreover the root of the P is in the
eld (see example 1.13).
We also can say that for any stem eld E :
K [X] / (P ) ∼
= E.
We can use the following Isomorphism: f : ∀P ∈ K [X] / (P ) → P(α), there

α is a root of polynomial P . 2
To summarize we have the following
1 Comment from Marc Emmanuel: The stem eld is more widely known as simple
extensions [21]
2 In the case we have f (P ) = P (α) = 0 as expected
25
26CHAPTER 2. STEM FIELD, SPLITTING FIELD, ALGEBRAIC CLOSURE
Proposition 2.2 (About stem eld existence). The stem eld exist and if
0
we have 2 stem elds E and E which correspond 2 roots of P : E = K [α],
E = K [α ] then ∃!f : E ∼
0 0 0
= E (Isomorphism of K-algebras) such that f (α) =
0
α.
Proof.Existence: K [X] / (P ) can be took as the stem eld.

Uniquest of the Isomorphism is easy because it is dened by it's value on
argument α: 3
φ : K [X] / (P ) ∼
=x→α E,
ψ : K [X] / (P ) ∼
=x→α0 E 0 ,
thus
ψ ◦ φ−1 : E ∼
=α→x→α0 E 0 .
Remark 2.3 (About stem eld). 1. In particular: If a stem eld contains

2 roots of P then ∃! Automorphism taking one root into another.
2. If E stem eld then [E : K] = deg P
3. If [E : K] = degP and E contains a root of P then E is a stem eld
4
4. If E is not a stem eld but contains root of P then [E : K] > degP
3 First of all if we have an isomorphism f between two K algebras K [α] and K [α0 ] it
should preserve the K -algebra structure, especially ∀k ∈ K : k1K[α] →f k1K[α0 ] . As soon
as k ∈ K [α] we can write the following

f k1K[α] = f (k) f 1K[α] = f (k) 1K[α0 ] .
But from other side

f k1K[α] = kf 1K[α] = k1K[α0 ]
i.e. ∀k ∈ K : f = id.
α forms a basis such that ∀β ∈ E = K [α] we have β = i ki αi where ki ∈ K . We also
P
i i
have f (β) = i ki [f (α)] = i ki [α0 ] . Thus if ∃f 0 isomorphism such that f 0 (α) = α0
P P
i
then f 0 (β) = i ki [α0 ] = f (β) i.e. the isomorphisms are the same.
P
4 Let E 0 is a stem eld of P . In the case we have E 0 ⊂ E as soon as any element of E 0
is an element of E because E contains a root of P . From other side E 6= E 0 as soon as E
is not a stem eld. Thus deg E > deg E 0 = deg P .
2.2. SPLITTING FIELD 27
2.1.2 Some irreducibility criteria

Corollary 2.4. P ∈ K [X] K if and only if it does not
is irreducible over
n
have a root in Field extension L of K such that [L : K] ≤ 2 , where n = degP .
Proof. ⇒: If P is not irreducible then it has a polynomial Q that divides P

and degQ ≤ n2 . 5 The Stem eld L for Q exists and it's degree is degQ ≤ n2 .
L should have a root of Q (as soon as a root of P ) by denition.
⇐: If P has a root α in L then ∃Pmin (α, K) with degree ≤ n2 < n 6 that
divides P (see lemma 1.16) i.e. P become reducible.
Corollary 2.5. P ∈ K [X] irreducible with deg P = n. Let L be an extension

of K such that [L : K] = m. If gcd (n, m) = 1 then P is irreducible over L.
Proof. If it is not a case and ∃Q such that Q | P in L [X]. Let M be a Stem

eld of Q over L.
So we have K ⊂ L ⊂ M = L (α). M is a stem eld of Q therefore
[M : L] = deg Q = d < n. Thus
[M : K] = [M : L] [L : K] = md
Lets K (α) is a stem eld of P over K then [K (α) : K] = deg P = n.

K (α) ⊆ M and therefore n | md 7 thus using gcd(m, n) = 1 one can get
that n | d but this is impossible because d < n.
2.2 Splitting eld

Denition 2.6 (Splitting eld). Let P ∈ K [X]. The splitting eld of P
over K is an extension L where P is split (i.e. is a product of linear factors)
and roots of P generate L
Theorem 2.7 (About splitting elds). 1. Splitting eld L exists and [L : K] ≤

d!, where d = degP .
5
P = RQ and if degQ > n2 then we can take R as Q
6
because [L : K] ≤ n2 (see remark 2.3)
7 K ⊂ K (α) ⊂ M and with The multiplicativity formula for degrees (Theorem 1.27)
we have
md = [M : L] [L : K] = [M : K] = [M : K (α)] [K (α) : K] = [M : K (α)] · n

2. If L and M are 2 splitting elds then ∃φ : L ∼

=M (an Isomorphism).
But the Isomorphism is not necessary to be unique.
Proof. Lets prove by induction on d. The rst case (d = 1) is trivial the K

itself is the splitting eld. Now assume d > 1 and that the theorem is valid
for any polynomial of degree < d over any eld K . Let Q be any irreducible
factor of P . We can create a Stem eld L1 = K (α) for Q that will be also a
Stem eld for P .
Over L1 we have P = (x − α)R, where R is a polynomial with degR =
d − 1. We know (by induction) that there exists a Splitting eld L for R over
L1 and its degree: [L : L1 ] ≤ (d − 1)! We have K ⊂ L1 ⊂ L. The L will be a
splitting eld for original polynomial P . Its degree (by The multiplicativity
formula for degrees (Theorem 1.27) ) is ≤ d · (d − 1)! = d!.
Uniqueness: Let L and M are 2 splitting elds. Let β is a root of Q
(irreducible factor of P ) in M . We have 2 stem elds: L1 = K (α) and
M1 = K (β). Proposition 2.2 says as that
L1 = K (α) ∼
= K (β) = M1 ,
i.e. ∃φ - isomorphism such that φ(α) = β .

Over M1 we have P = (x − β)S , where S = φ (R). 8 M is a splitting eld
for S over K [β] i.e. it is a K [β]-algebra but it's is also a K [α]-algebra 9 and
as result it's a splitting eld for R over K [α] and by induction 10 we have
K [α] isomorphism L ∼ = M and as result K isomorphism L ∼ = M . 11
8 We have φ : K (α) → K (β). Note 3 states that φ |K = id, where φ |K is the φ
restricted to K : φ |K : K → K . Therefore φ (P ) = P because P ∈ K [X]. Thus
P = (X − β)S = φ (P ) = φ ((X − α)R) = (X − β)φ (R)
and S = φ (R).
9 via the existent Isomorphism between K [α] and K [β]
10 Induction steps are the following: we have a polynomial P with deg P = n. For n = 1
the isomorphism exists by proposition 2.2. We suppose that the isomorphism is proved
for polynomial with degree n − 1.
11 Lukas Heger comment about the prove: We can consider another roots: α for R and
2
β2 for S and there is an isomorphism between the 2 stem elds also. Continue in the way
we will get the 2 following chains
K ⊂ L1 ⊂ L2 ⊂ · · · ⊂ Ln ⊂ L
K ⊂ M1 ⊂ M2 ⊂ · · · ⊂ Mn ⊂ M
On each step we have an isomorphism between Li and Mi and as result the isomorphism
between resulting elds L and M (via φ) as Ln algebras and therefore as K algebras.
2.3. AN EXAMPLE. ALGEBRAIC CLOSURE 29
Remark 2.8. The Isomorphism considered in theorem 2.7 is not unique. A

splitting eld can have many Automorphism and this is in fact the subject
of Galois theory.
2.3 An example. Algebraic closure

2.3.1 An example of automorphism
Example 2.9 (X 3 − 2 over Q). X3 − 2
Let we have the following polynomial
√ √
3 3
over Q (see also example 6.10). It has the following roots:
√
2 3 2πi √
3
2, j 2 and
j 2, where j = e . Splitting eld is the following L = Q 2, j . Lets nd
3
2
Automorphisms of the eld. Pmin (j, Q) = X + X + 1 thus using remark 2.3
√
[Q (j) : Q] = 2. Using the same arguments one can get that Q 3 2 : Q =
3. As result the following picture can be got
Q (j)
2 3
√
3

Q Q 2, j = L
3 2
√
3

Q 2
√
3

As soon as L is a stem eld for Q (j) and for Q 2 then 2 types of
automorphism exist:
√
3
2 3
√
1. Q 2Automorphism. We have X + X + 1 as Pmin j, Q 2 . The
2
polynomial has 2 roots: j and j and there is an Automorphism that
12
exchanges the roots. Lets call it τ
√
3
2. Q (j) √
Automorphism. In this case the automorphism of exchanging 2
3 13
and j 2. . Lets call it σ
The group of automorphism of L Aut (L/K) is embedded into permutation

group of 3 elements S3 (see example A.56):
Aut (L/K) ,→ S3 .
12 j → j 2 thus . Therefore
j 2 → j 4 = j√ 2
13
√ √ √ j ↔ j 2√ √ √
3
2 → j 2 produces j 2 → j 2 and j 3 2 → j 3 3 2 = 3 2. This statement
3 3 2 3
corresponds
√ √the fact that
√ the minimal polynomial is X − 2 there and thus we have 3
3
roots: 2, j 2 and j 2
3 3 2 3
It's embedded because the automorphism exchanges the roots of X 3 −2. More-
over
Aut (L/K) = S3 ,
because σ and τ generates S3 because
√ √ √ √
• σ : 3 2 → j 3 2 → j 2 3 2 → 3 2. This is a circle.
√ √ √
• τ - it keeps 3 2 and exchanges j and j 2 : 3 2j ↔ 3 2j 2 (see note 12).
This is a transposition.
2.3.2 Algebraic closure

Denition 2.10 (Algebraically closed eld). K is algebraically closed if any
non constant polynomial P ∈ K [X] has a root in K or in other words if any
P ∈ K [X] splits
Example 2.11 (C). C is an Algebraically closed eld. This will be proved

later.
Denition 2.12 (Algebraic closure). An algebraic closure of K is a eld L

that is Algebraically closed eld and Algebraic extension over K . 14
Theorem 2.13 (About Algebraic closure). Any eld K has an Algebraic

closure
Proof. Lets discuss the strategy of the prove. First construct K1 such that
∀P ∈ K [X] has a root in K1 . There is not a victory because K1 can introduce
new coecients and polynomials that can be irreducible over K1 . Then
construct K2 such that ∀P ∈ K1 [X] has a root in K2 and so forth. As result
we will have
K ⊂ K1 ⊂ K2 ⊂ · · · ⊂ Kn ⊂ . . .
Take K̄ = ∪i Ki and we claim that K̄ is algebraically closed. Really ∀P ∈
K̄ [X] ∃j : P ∈ Kj [X] thus it has a root in Kj+1 and as result in K̄ .
Now howcan we construct K1 . Let S be a set of all irreducible P ∈ K [X].
Let A = K (XP )P ∈S - multi-variable (one variable XP for each P ∈ S )

polynomial ring.
14 If L is algebraic closure of K then the following conditions are valid
• ∀P ∈ L [X] ∃α ∈ L such that P (α) = 0 (see denition of Algebraically closed eld)
• ∀α ∈ L ∃P ∈ K [X] such that P (α) = 0 (see denition of Algebraic extension)
2.3. AN EXAMPLE. ALGEBRAIC CLOSURE 31
Let I ⊂ A is an Ideal generated by a set P (XP ) ∀P ∈ S .15 We claim

that I is a Proper ideal i.e. I 6= A. If not then we can write (see theorem
A.82)
n
X
1A = λi Pi (XPi ) , (2.1)
i
where λi ∈ A and the sum is the nite (see denition A.72). As soon as
the sum
Qn is nite then I can take the product of the polynomials in the sum:
P = i Pi and I can create a Splitting eld L for the polynomial P over K
(see theorem 2.7).
A is a polynomial ring and it's very easy produce a homomorphism be-
tween polynomial algebra and any other algebra. Therefore there is a homo-
morphism between rings A and L such that φ : A → L where XPi → αi 16 if
P = Pi and XPi → 0 otherwise. From (2.1) we have
n
X n
X
φ(1A ) = λi φ (Pi (Xpi )) = λi Pi (αi ) = 0
i i
that is impossible.
Fact: Any Proper ideal I ⊂ A is contained in the Maximal ideal m (see
proposition 2.16 below) and A/m is a eld (see theorem A.94).
Thus I can take K1 = A/m and continue in the same way to construct
K2 , K3 , . . . , Kn , . . . .
2.3.3 Ideals in a ring

The ring is commutative, associative with unity. Any Proper ideal is in a
Maximal ideal. This is a consequence of what one calls Zorn's lemma
Denition 2.14 (Chain). Let P is a partially ordered set (≤ is the order

relation). C ⊂ P is a chain if ∀α, β ∈ C exists a relation between α and β i.e.
α ≤ β or β ≤ α.
Lemma 2.15 (Zorn). If any non-empty Chain C in a non-empty set P has

an upper bound (that is M ∈ P such that M ≥ x, ∀x ∈ C ) then P has a
maximal element.
Proposition 2.16. Any Proper ideal is in a Maximal ideal
15 I = i λi Pi (XPi ), where λi ∈ A
P
16 α is a root of P
i i
Proof. We can use Zorn (Lemma 2.15) to prove that any proper ideal is in
a Maximal ideal.
Let P is the set of proper ideals in A containing I . The set is not empty
because it has at least one element I . Any Chain C = {Iα } 17 has an upper
bound: it's ∪α Iα (exercise that the union is an ideal). So P has a maximal
element m and I ⊂ m.
If we take a Quotient ring by maximal ideal it's always a eld 18 otherwise

it will have a proper ideal: ∃a ∈ A/m such that (a) is a proper ideal and it
pre-image in π : A → A/m should strictly contain m 19 .
2.4 Extension of homomorphisms. Uniqueness

of algebraic closure
Some summary about just proved existence of algebraic closure. There exists
i=1 Ki - algebraic closure of K , where
K̄ = ∪∞
K ⊂ K1 ⊂ K2 ⊂ · · · ⊂ Ki−1 ⊂ Ki ⊂ . . .
Ki is a eld where each polynomial P ∈ Ki−1 has a root. The eld Ki is

Quotient ring of huge polynomial ring Ki−1 [X] by a suitable Maximal ideal
that is got by means of Zorn (Lemma 2.15) .
Another question is the closure unique? The answer is yes. We start the
proof with the following theorem
Theorem 2.17 (About extension of homomorphism). Let K ⊂ L ⊂ M -
Algebraic extension. K ⊂ Ω, where Ω - Algebraic closure of K . ∀φ : L → Ω
extends to φ : M → Ω 20
e
Proof. Apply Zorn (Lemma 2.15) to the following set (of pairs)
E = {(N, ψ) : L ⊂ N ⊂ M, ψ extends φ}
E is non empty because (L, φ) ∈ E .

The set E is partially ordered by the following relation (≤):
(N, ψ) ≤ (N 0 , ψ 0 ) ,
17 The order is the following Iα ≤ Iβ if Iα ⊂ Iβ
18 We refer to it as a theorem with denition provided in A.94. The comments can be
considered as a simple prove of the fact.
19 i.e. m is not a maximal ideal in the case
20 see also example 3.3.
2.4. EXTENSION OF HOMOMORPHISMS. UNIQUENESS OF ALGEBRAIC CLOSURE33
if N ⊆ N 0 and ψ 0 /N = ψ (ψ 0 extends ψ ). Any Chain (Nα , ψα ) has an upper

bound (N, ψ), where N = ∪α Nα - eld, sub extension of M . ψ dened in
the following way: for x ∈ Nα ψ(x) = ψα (x).
Thus E has a maximal element that we denote by (N0 , ψ0 ).
Lets suppose that N0 6= M , i.e. N0 ( M . Now it's very easy to get
a contradiction. Lets take x ∈ M \ N0 and consider Minimal polynomial
Pmin (x, N0 ). It should have a root α ∈ Ω. Now we extend N0 to N0 (x) and
dene ψ 0 on N0 (x) as follows: ∀y ∈ N0 : ψ 0 (y) = ψ0 (y) and ψ 0 (x) = α. Thus
we was able to nd an element of the chain that is greater than maximal.
Therefore our assumption about N0 6= M was incorrect and we can conclude
than N0 = M and therefore φ̃ = ψ0 .
Corollary 2.18 (About algebraic closure isomorphism). If ∆ and ∆0 are 2

algebraic closures of K then they are isomorphic as K-algebras.
Proof.Using theorem 2.17 one can assume L = K , M = ∆0 and Ω = ∆ i.e.

we have
K ⊂ K ⊂ ∆0
in this case homomorphism K → ∆ can be extended to ∆0 → ∆ i.e. there
exists a homomorphism (i.e. Injection) from ∆0 to ∆.
If we assume M = ∆ and Ω = ∆ then there exists a homomorphism (i.e.
Injection) from ∆ to ∆0 . The Injection is also Surjection in another direction:
∆0 → ∆ and as result we have Isomorphism ∆0 → ∆
Chapter 3
Finite elds. Separability, perfect
elds
We recall the construction and basic properties of nite elds. We prove that
the multiplicative group of a nite eld is cyclic, and that the automorphism
group of a nite eld is cyclic generated by the Frobenius map. We intro-
duce the notions of separable (resp. purely inseparable) elements, extensions,
degree. We briey discuss perfect elds.
3.1 An example (of extension)s. Finite elds

Corollary 3.1. Algebraic closure of K is unique up to Isomorphism of K-
1
algebras
Corollary 3.2. Any Algebraic extension of K embeds (see denition A.133)

2
into the Algebraic closure
Example 3.3 (Of extension of homomorphism). Let K = Q and Q is the

Algebraic closure of K . For instance we can consider Q ⊂ C. 3
Let √
2 = Q [X] / X 2 − 2 ,

L=Q
α is a Class of X in L. L has 2 Embeddings into Q

√
1. φ1 : α → 2
1There is a redenition of corollary 2.18.
2i.e. ∀E - algebraic extension of K , ∃φ : E → K̄ - Homomorphism. The statement is
a reformulation of theorem 2.17
3 Really Q = A - the set of all algebraic numbers, i.e. roots of polynomials P ∈ Q [X].
35
36 CHAPTER 3. FINITE FIELDS. SEPARABILITY, PERFECT FIELDS
√
2. φ2 : α → − 2
Let √
2 = Q [Y ] / Y 4 − 2 ,
4

M =Q
β is a Class of Y in M . M has 4 Embeddings into Q

√
1. ψ1 : β → 4 2 (extends φ1 )
√
2. ψ2 : β → − 4 2 (extends φ1 )
√
3. ψ3 : β → i 4 2 (extends φ2 )
√
4. ψ4 : β → −i 4 2 (extends φ2 )
4
This (extends) is because
M = L [Y ] / Y 2 − α

3.1.1 Finite elds

Denition 3.4 (Finite eld). K is a nite eld if it's Characteristic charK =
p, where p - prime number
Remark 3.5 (Fpn ). If K is a nite extension of Fp 5 and n = [K : Fp ] then
number of elements of K : |K| = pn . The following notation is also used for
a nite extension of a nite eld: Fpn 6
Remark 3.6 (Frobenius homomorphism). If charK = p, then exists a Ho-
momorphism Fp : K → K such that Fp (x) = xp . Really if we consider
(x + y)p and (xy)p then we can get (x + y)p = xp + y p 7 and (xy)p = xp y p .
The second property is the truth in the all elds (of course) but the rst one
is the special property of Fp elds.
4 I.e. in our case we have Q ⊂ L ⊂ M . We have φ
1,2 : L → Q which can be extended
(accordingly theorem 2.17) to ψ1,2,3,4 : M → Q
5 i.e. [K : F ] < ∞
p
6 As we know F = Z/pZ. From other side F n 6= Z/pn Z. For example F 6= Z/4Z
p p 4
because Z/4Z is not a eld (2 · 2 = 0 i.e. zero divisors exist). You have to look at example
1.13 to see exact structure of F4 .
7
p p−1
!
p
X p k p−k p p
X
k p−k
(x + y) = x y =x +y +p· ak x y ,
k
k=0 k=1
where ak ∈ Z. I.e.
p
(x + y) ≡ (xp + y p ) mod p
3.2. PROPERTIES OF FINITE FIELDS 37
Remark 3.7. n
Also Fpn : K → K such that Fpn (x) = xp is also homomor-
phism (a power of Frobenius homomorphism (Remark 3.6) . )
3.2 Properties of nite elds

Theorem 3.8. Fp and it's Algebraic closure Fp .
Lets x
n
xp − x has pn elements. Conversely any eld of
The Splitting eld of
n
pn elements is a splitting eld of xp − x. Moreover there is an unique sub
n
extension of Fp with p elements.
n
Proof. Note that Fpn : x → xp is a Homomorphism (see remark 3.7) as
result the following set {x | Fpn (x) = x} is a eld containing Fp 8 i.e.
Fp ⊂ {x | Fpn (x) = x}
or, in other words, the considered set is a Field extension of Fp .

n
If Qn (X) = X p − X then the considered set consists of the root of the
polynomial Qn . The polynomial has no multiple roots because gcd(Qn , Q0n ) =
1. 9 This is because Q0n ≡ 1 mod p. 10 As soon as Qn has no multiple roots
then there are pn dierent roots and therefore the splitting eld is the eld
with pn elements.
Conversely lets |K| = pn and α 6= 0 ∈ K . Using the fact that the
multiplication group of K has pn − 1 elements: |K × | = pn − 1 11 as result the
n n
multiplication of all the elements should give us 1: αp −1 = 1 or αp − α = 0
8 For x ∈ F×
p = Fp \ {0} we have that (see theorem A.40)
×
x|Fp | = xp−1 = 1
and therefore ∀x ∈ Fp : xp = x (x = 0 also satised the equation). We can continue as

follows
2 p
xp = (xp ) = xp = x,
3
2 p
xp = xp = xp = x
n−1 p ...
pn p p
x = x =x =x
and nally get Fpn (x) = x. Thus ∀x ∈ Fp we also have x ∈ {x | Fpn (x) = x}
9 If Q has a multiple root β then it is divisible by (X − β)2 and the Q0 is divisible
n n
by (at least) (X − β) thus the (X − β) should be a part of gcd.
10 Really we have the following one Q0 = pn X pn −1 − 1 ≡ −1 mod p but the sign is not
n
really matter because gcd (Qn , −1) = gcd (Qn , 1) = 1.
11 K × = K \ {0}
(see theorem A.40). Therefore α is a root of Qn . Thus the splitting eld of

Qn consists of elements of K .
The uniqueness 12 of sub-extension of Fp with pn elements is a result of
uniqueness of the splitting eld (see theorem 2.7).
Theorem 3.9. Fpd ⊂ Fpn if and only if d | n.
Proof. Let Fpd ⊂ Fpn in this case Fp ⊂ Fpd ⊂ Fpn and

[Fpn : Fp ] = Fpn : Fpd Fpd : Fp
or n = x · d i.e. d | n
Conversely if d | n then n = x · d or pn = xi=1 pd thus if xp = x then
Q d
Qx Qxi=2 pd Qx
pn i=1 p
d pd pd d
x =x x =x i=2 = · · · = xp = x,
i.e. ∀α ∈ Fpd we also have α ∈ Fpn or in other notation: Fpd ⊂ Fpn .
Theorem 3.10. Fpn is a Stem eld and a Splitting eld of any Irreducible
polynomial P ∈ Fp of degree n.
Proof. Stem eld K has to have degree n over Fp i.e. [K : Fp ] = n (see

remark 2.3) i.e. it should have pn elements (see remark 3.5) and therefore
K = Fpn (see theorem 3.8).
About Splitting eld. Using the just proved result we can say that if α
is a root of P then α ∈ Fpn thus Qn (α) = 0. Therefore P divides Qn 13 and
as result P splits in Fpn .
Corollary 3.11. LetPd is the set of all irreducible, Monic polynomials of

degree d such that Pd ⊂ Fp [X] then
Y Y
Qn = P
d|n P ∈Pd
12 up to Isomorphism
13 as soon as any root of P also a root of Qn
3.2. PROPERTIES OF FINITE FIELDS 39
Proof. As we just seen if P ∈ Pd and d | n then P | Qn . 14 Since all such

polynomials are relatively prime of course 15 16 and Qn have no multiple roots
(as result no multiple factors) then
 
Y Y
 P  | Qn
d|n P ∈Pd
From other side let R is an irreducible factor of Qn . α is a root of R then

Qn (α) = 0 thus Fp (α) ⊂ Fpn . From remark 2.3 we have
[Fp (α) : Fp ] = deg R = d.
From remark 3.5 Fp (α) = Fpd . Theorem 3.9 says that Q d |Qn. As result
R ∈ Pd . Thus the polynomial should be in the product d|n P ∈Pd P .
Example 3.12. Let p = n = 2. The monic irreducible polynomials in F2
2
whose degree divides 2 are: X , X + 1 and X + X + 1. As you can see
X (X + 1) X 2 + X + 1 = X 4 + X = X 4 − X

because 2x = 0 mod 2 or x = −x.

Just another example [3]:
Example 3.13 (Polynomial factorization over a nite eld). 17
In F2 [X],
2n
the irreducible factorization of X − X for n = 1, 2, 3, 4 is as follows
X 2 − X = X (X − 1) ,
X 4 − X = X (X − 1) X 2 + X + 1 ,

X 8 − X = X (X − 1) X 3 + X + 1 X 3 + X 2 + 1 ,

X 16 − X = X (X − 1) X 2 + X + 1

X4 + X + 1 X4 + X3 + 1 X4 + X3 + X2 + X + 1 .

You can compare the example with example 3.12 but you have to take into
consideration the following fact 1 = −1 mod 2
14 Since stem eld is Fpd ⊂ Fpn (see theorem 3.9 and proof at the theorem 3.10)
15 As soon as Fp [X] is Unique factorization domain then any polynomial can be written
as a product of irreducible elements, uniquely up to order and units this means that each
P ∈ Pd (where d | n) should be in the factorization of Qn . It should be only one time
because there is no multiply roots.
16 We also can say that 2 irreducible polynomial P , P ∈ F [X] should not have same
1 2 p
roots. For example if α is the same root - it cannot be in Fp because in the case the
polynomials will be reducible. Thus it can be only in an extension of Fp from other side
gcd(P1 , P2 ) = 1 and therefore with Bezout (Lemma A.85) one can get that ∃Q, R ∈ Fp [X]
such that P1 Q+P2 R = 1 and setting α into the equation leads to fail statement that 0 = 1.
Example 3.14 (Polynomial factorization over a nite eld in GAP). 18 At

the moment GAP provides only methods to factorize polynomials over nite
elds, over subelds of cyclotomic elds, and over algebraic extensions of
these [9].
There is an example of such factorization in F2 [X] similar to Exam-
ple 3.13
gap> x:=Indeterminate(GF(2),"x");;
gap> Factors(x^8-x);
[ x, x+Z(2)^0, x^3+x+Z(2)^0, x^3+x^2+Z(2)^0 ]
gap> x:=Indeterminate(GF(2),"x");;
[ x, x+Z(2)^0 ]
[ x, x+Z(2)^0, x^2+x+Z(2)^0 ]
[ x, x+Z(2)^0, x^3+x+Z(2)^0, x^3+x^2+Z(2)^0 ]
[ x, x+Z(2)^0, x^2+x+Z(2)^0, x^4+x+Z(2)^0, x^4+x^3+Z(2)^0,
x^4+x^3+x^2+x+Z(2)^0 ]
gap>
3.3 Multiplicative group and automorphism group

of a nite eld
Theorem 3.15. LetK be a eld and and G be a nite Subgroup of K× (see
19
denition A.67) then G is a Cyclic group
19 Not every nite group is cyclic. For instance the non-abelian group S3 (see example
A.56) consists of 6 elements but it is not cyclic.
If we want to have a cyclic group it should be Abelian group in the case it can be
represented in the form of a cyclic group or a direct sum of cyclic groups accordingly The
fundamental theorem of nitely generated abelian groups (Theorem A.47) . Another
important requirement is that the orders of the cyclic group have to be coprime.
For example there are 2 groups of order 4: the cyclic one - Z/4Z and V4 - Klein four group
V4 = Z/2Z ⊕ Z/2Z is non-cyclic because the orders are not coprime: gcd(2, 2) = 2 6= 1.
(From the sta) But a product of two cyclic groups is again cyclic, if their orders are
coprime. For example, consider the element (1, 1) ∈ Z/2Z × Z/3Z. It's order is 6, so it
generates the whole group. Thus as it's mentioned Z/N Z is a cyclic group.
3.3. MULTIPLICATIVE GROUP AND AUTOMORPHISM GROUP OF A FINITE FIELD41
Proof. Idea is to compare G and the Cyclic group Z/N Z where N = |G|. 20
Let ψ (d) - is the number of elements of order d ( see also Order
P of element
in group) in G. We need ψ (N ) 6= 0 21 and we know that N = ψ (d).
Let also φ (d) - is the number of elements of order d ( see also Order of
element in group) in Z/N Z. 22 As Z/N Z contains a single (cyclic) subgroup
of order d for each d | N . 23 φ (d) is the number of generators of Z/dZ i.e.
the number of elements between 1 and d − 1 that are prime to d. We know
that φ (N ) 6= 0.
We claim that either ψ (d) = 0 or ψ (d) = φ (d) 24 If no element of order
d in G then ψ (d) = 0 otherwise if x ∈ G has order d then xd = 1 or x is a
root of the following polynomial xd − 1. The roots of the polynomial forms
a cyclic subgroup of G (by Cyclic group denition). So G as well as Z/N Z
has a single cyclic subgroup of order d (which is cyclic) or no such group at
all. 25
If ψ (d) 6= 0 then exists such a subgroup and ψ (d) is equal to the number
of generators of that group or φ (d) 26 In particular
P ψ (d) ≤P φ (d) 27 but there
should be equality because the sum of both ψ (d) = φ (d) = N . In
particular ψ (N ) 6= 0 and we proved the theorem.
Corollary 3.16. If Fp ⊂ K and [K : Fp ] = n then ∃α such that K = Fp (α).

In particular ∃ an Irreducible polynomial of degree n over Fp 28
20 We also will use the fact that any cyclic group of order N is isomorphic to Z/N Z
21 In this case we will have at least one element x of order N i.e. N dierent elements
of G is generated by the x i.e. the G is cyclic.
22 The function φ (d) is also called as Euler's totient function and it counts the positive
integers up to a given integer d that are relatively prime to d
23 The one generated by N/d. Let N = r · d in the case xN = 1 there x is a Z/N Z group
generator. From other side
Yr
xN = xr·d = xd
i=1
thus x = 1 i.e. there

d
is a cyclic subgroup of order d.
24 suces since P ψ (d) = P φ (d) = N
25 Several comments about the subgroup. There is a group because multiplication of
any elements is in the set. It's cyclic because it's generated by one element. All xi where
i ≤ d are dierent (in other case the group should have an order less than d). Each element
of the group xi is a root of xd − 1 because (xi )d = (xd )i = 1i = 1. And the group is unique
as well as we have d dierent roots of xd − 1 in the group.
26 Because the group is cyclic and any cyclic group is isomorphic to Z/dZ and as result
has the same number of generators.
27 because ψ (d) = 0 or ψ (d) = φ (d)
28 The theorem 3.10 and remark 3.5 says that the stem eld for any polynomial of degree
n over Fp exists and there is Fpn and [Fpn : Fp ] = n i.e. K = Fpn . But we had not proved
yet that an irreducible polynomial of degree n exists.
29
Proof. We can take α = generator of K ×
Corollary 3.17. The group of automorphism of Fpn over Fp is cyclic and

p
generated by Frobenius map: Fp : x → x (see remark 3.6 where we showed
that the Frobenius map is a eld automorphism)
As we know from theorem 3.8: ∀x ∈ Fpn : xp = x so 30 Fpn = id. As

n
Proof.
result the order of hFp i is no greater than n. Lets prove that the ordFp = n.
Really if m < n then xp − x = 0 has pm < pn roots and 31 Fpm cannot be
m
identity. Finally (from corollary 3.16 ) we have Fpn = Fp (α) where α is a

root of an irreducible polynomial of degree n. I.e. there cannot be more than
n automorphism 32 so
|Aut (Fpn /Fp )| ≤ n
33
and as we have n of them (Automorphisms) then
|Aut (Fpn /Fp )| = n
and the group is cyclic generated by Fp .
3.4 Separable elements

Let E is a Splitting eld of an irreducible polynomial P . We would like to
say that it has many Automorphisms. What does this mean? This means
the following thing: Let α and β be 2 roots of P then we have 2 extensions
K (α) ⊂ E and K (β) ⊂ E .
There exists an Isomorphism (see proposition 2.2) over K
φ : K (α) → K (β)
29 This is because from theorem 3.15 K × = hαi i.e. any element of K except 0 can be
got as a power of α. Moreover α ∈ / Fp (in other case we will got K = Fp ) i.e. we really got
K = Fp (α). α is an Algebraic element because we can consider 1, α, . . . , αn−1 as a basis
and αn can be represented via the basis. I.e. ∃P ∈ Fp [X] such that P (α) = 0. By lemma
1.16 there exists an irreducible polynomial Pmin (α, Fp ).
30 because
n n−1 n−1 n
(Fp ) (x) = (Fp ) (Fp (x)) = (Fp ) (xp ) = · · · = xp = x
31 because operates only with pm elements i.e. not of all elements of Fpn .
32 Each automorphism converts the root α into another one of n roots of the irreducible
polynomial
33 We have n dierent elements of cyclic group hF i. The generator of the group is an
p
automorphism and as result each of n elements is also an automorphism.
3.4. SEPARABLE ELEMENTS 43
that is also extended to an Automorphism on E (see theorem A.135).

There is one problem with it: is that truth that an irreducible polynomial
of degree n has many i.e. exactly n ( it cannot have more than n) roots.
The answer is yes if charK = 0, but not always if charK = p (where p is
a prime number). P can have multiple roots in the case i.e. gcd(P, P 0 ) 6= 1.
Why it's not a case for charK = 0 - it is because deg P 0 < deg P and
P - P 0 for P 0 6= 0 (non constant polynomial) 34
But for charK = p there can be a case when P 0 = 0 for a non constant
polynomial thus P | P 0 and as result gcd(P, PP 0
) = P . The P 0 = 0 i.e. it
vanishes P is a polynomial in X p . I.e. if P = ai xi and p | i or ai = 0. In
h
that case (P = 0) let r = max h such that P is a polynomial in X p that is
0
ai = 0 whenever ph - i. See the following example

Example 3.18. 35
Let p = 2. The polynomial P (X) = X 16 + 1 has the
0
required property (P = 0). The polynomial can be present in the following
form
4
P (X) = X 2 + 1 = Q(Y )
where Y = X 16 and Q(Y ) = Y + 1. In the case r = 4, pr = 16 | 16
For polynomial P (X) = X 12 + 1 we have
2 3
P (X) = X 2 + 1 = Q(Y )
where Y = X4 and Q(Y ) = Y 3 + 1. 2

In the case r = 2, p = 4 | 12 because
h 3
h = 3 > 2 does not t into the requirements: p = 2 = 8 - 12.
Proposition 3.19. r
Q0 6= 0 i.e. gcd(Q, Q0 ) = 1

Let P (X) = Q X p and
r
then Q does not have multiple roots but all roots of P have multiplicity p .
r
Proof. If λ is a root of P then λ: P (X) = (X − λ)R Thus µ = λp is the
root of Q 36 as result Q(Y ) = (Y − λp )S(Y ) therefore
r
r r r r r
P (X) = X p − λp S X p = (X − λ)p S X p
and λ is not a root of S X p . 37 Thus we just got that multiplicity of λ is

r
pr .
34 Let P has multiply roots. As soon as it's irreducible a multiply root is in an extension
of K . In this case the root should be also a root for P 0 thus by lemma 1.16 (or theorem
A.89) one can get that P | P 0 in K [X] but that is impossible because deg P 0 < deg P and
can be only possible if P 0 = 0.
36 Q (µ) = Q λpr = P (λ) = 0

37 This is because Q does not have multiply roots and as result µ = λpr is not a root of
r
S or in other words S X p X=λ 6= 0
Denition 3.20 (Separable polynomial). P ∈ K [X] irreducible polynomial

is called separable if gcd (P, P 0 ) = 1
Denition 3.21 (Degree of separability). dsep (P ) = deg Q (as above) 38

Denition 3.22 (Degree of inseparability). di (P ) = deg P
deg Q
( = pr in proposi-
tion 3.19)
Denition 3.23 (Pure inseparable polynomial). P is pure inseparable if

di = deg P . Then P = X p − a 39
r
Denition 3.24 (Separable element). Let L be an Algebraic extension of

K then α ∈ L is called separable(inseparable) if it's Minimal polynomial
Pmin (α, K) has the property. Note: the separable element is also Algebraic
element because it has minimal polynomial.
Proposition 3.25 (On number of homomorphisms). If α is separable on K

then the number of Homomorphisms over K from K to K̄

HomK K (α) , K̄ = deg Pmin (α, K)
in general

HomK K (α) , K̄ = dsep Pmin (α, K)
Proof. It's obvious because dsep is the number of distinct roots.
3.5 Separable degree, separable extensions

We want to generalize the proposition 3.25 for any eld extension (not nec-
essary K (α)). Let L be a nite extension of K
Denition 3.26 (Separable degree). [L : K]sep = HomK L, K̄

As we know if L = K (α) then Separable degree is a number of distinct

roots of minimal polynomial Pmin (α, K)
Denition 3.27 (Separable extension). L is separable over K if [L : K]sep =

[L : K]
38 It requires some explanation compare to that one was got on the lecture video. If P
is a Separable polynomial then dsep (P ) = deg P . In other case P should be represented as
P (X) = q1 (X p ). If q1 (Y ) is separable than Q = q1 otherwise we continue and represent
q2 (X ). We should stop on some qr for which we will have Q = qr and P (X) =
p
q1 (X) =
pr
Q X . In the case dsep (P ) = deg Q.
39 In the case deg Q = 1 i.e. Q (Y ) = Y − a or P = X pr − a.
3.5. SEPARABLE DEGREE, SEPARABLE EXTENSIONS 45
Denition 3.28 (Inseparable degree).

[L : K]
[L : K]i =
[L : K]sep
Theorem 3.29 (About separable extensions). 1. If K ⊂ L ⊂ M then

[M : K]sep = [M : L]sep [L : K]sep and M is Separable extension over
K if and only if M is separable over L and L is separable over K
2. The following things are equivalent
(a) L is separable over K

(b) ∀α ∈ L α Separable element over K
(c) L is generated over K by a nite number of Separable elements
i.e. L = K (α1 , α2 , . . . , αn ), there αi is separable over K
(d) L = K (α1 , α2 , . . . , αn ), there αi is separable over K (α1 , α2 , . . . , αi−1 )
Remark 3.30. That holds if we replace separability with pure inseparability.

Proof. About 1st part: If we have a Homomorphism φ : L → K̄ then it
is extended to φ̃ : M → K̄ (by extension theorem 2.17) it can be done
with one way
per each homomorphism from L to M i.e. it can be done by
HomL M, K̄ ways but we have

HomL M, K̄ = HomL M, L̄ = [M : L]
sep
because K̄ is also L̄ (Algebraic closure over L) thus for the total number of
homomorphisms one can get the following equations

[M : K]sep = HomK M, K̄ = HomK L, K̄ HomL M, K̄ =

HomK L, K̄ HomL M, L̄ = [M : L] [L : K]
sep sep
40
We have the following inequality
[E : K]sep ≤ [E : K] . (3.1)
40The inequality can be proved by induction using the fact that it's true for K (α)
because from general case of proposition 3.25

HomK K (α) , K̄ = dsep Pmin (α, K) ≤ deg Pmin (α, K) = [K (α) : K]
Then let it was proved for E = K (α1 , . . . , αn−1 ) and we want to prove it for
K (α1 , . . . , αn−1 , αn ) = E (αn ). It's easy because Ē = K̄ and we can use the same
approach as for the rst induction step.
With the inequality (3.1) we also have
[M : K]sep = [M : L]sep [L : K]sep ≤ [M : L] [L : K] = [M : K]

The equality is possible if [M : L]sep = [M : L] and [L : K]sep = [L : K] i.e.
if M is separable over L and L is separable over K . This nishes the proof
of the rst part.
About 2d part:
2a ⇒ 2b: Part 1 implies that a separable sub extension K (α) or a sepa-
rable extension L is separable. 41
2b ⇒ 2c: obvious 42
2c ⇒ 2d: We know that Pmin (αi , K (α1 , . . . , αi−1 )) divides Pmin (αi , K).
43
Thus if Pmin (αi , K) is separable (i.e. have distinct roots) then it's di-
visor Pmin (αi , K (α1 , . . . , αi−1 )) also should have distinct roots i.e. αi is a
Separable element over K (α1 , . . . , αi−1 )
2d ⇒ 2a: Induction as above 44
What's about not nite extension? For that case we can dene separable
extension as follows.
Denition 3.31 (Separable closure). If L over K not necessary nite (but
algebraic over K ) we can dene
Lsep = {x|x separable over K}

45
Lsep is a sub extension called separable closure of K over L
L is pure inseparable over Lsep .
Remark 3.32. 1. If charK = 0 then any extension of K is separable
2. If charK = p then pure inseparable extension has degree pr and always

degree of inseparability [L : K]i = pr
41 I.e. in the case we have K ⊂ K (α) ⊂ L and if L is separable then K (α) is separable
and as result α is a Separable element because Pmin (α, K) is separable.
42 We consider nite extensions (see remark 3.5) i.e. which consists of nite number of
elements
43 Let K (α , . . . , α
1 i−1 ) = L then K ⊂ L and Pmin (αi , K) ∈ L [X] From other side
Pmin (αi , L) is the minimal irreducible polynomial in L [X] and any other polynomial with
αi as root has to by divisible by it (see also lemma 1.16).
44 The rst induction step is trivial: L = K (α) where α is separable over K in this
case K (α) is also separable. Now we have that ∀k < n: if L = K (α1 , α2 , . . . , αk ),
there αi is separable over K (α1 , α2 , . . . , αi−1 ) then L is separable over K . Thus we have
K (α1 , α2 , . . . , αn−1 ) separable and αn is separable over K (α1 , α2 , . . . , αn−1 ) thus using
the rst part of the theorem we can conclude that K (α1 , α2 , . . . , αn ) is also separable over
K
45 K ⊂ Lsep ⊂ L
3.6. PERFECT FIELDS 47
3.6 Perfect elds

Denition 3.33 (Perfect eld). Let K is a eld and charK = p > 0. K is
perfect if Frobenius homomorphism (Remark 3.6) is a Surjection
Example 3.34. 1. Finite eld is perfect because an Injection of a set into

itself is always a Surjection
2. Algebraically closed elds are perfect because Xp − a has a root α for

46
any a particularly a = Fp (α)
3. Not perfect eld example. Let K = Fp (X) be a eld of rational fractions

f (X)
in 1 variable over Fp . I.e. elements of the eld are g(X) where f, g ∈
Fp [X]. It's not perfect because Im (Fp ) = Fp (X p ) 6= Fp (X)
Theorem 3.35. K is a Perfect eld if and only if all irreducible polynomial

over K are separable or in other words all Algebraic extensions of K are
separable.
Proof. Let K is perfect and P ∈ K [X] is an irreducible polynomial. Let also

r r i
X
P (X) = Q X p = ai X p
i
47
but as soon as my eld is perfect then I can extract p-root of ai and do it
r
repeatedly. I.e. ∃bi ∈ K such that bpi = ai . Therefore
!p r
X pr pr
i X r
i p
X
P (X) = bi X = bi X = bi X i .
i i i
The polynomial is not irreducible unless r = 0 48 so irreducible means sepa-

rable.
If K is not perfect but all irreducible polynomial are separable. K is not
perfect means that ∃a ∈ / Im (Fp ) and lets consider the following polynomial:
r
X p − a. It is irreducible and not separable.
r
About separability: in fact all roots are in K̄ are the same x with xp = a
49
/ K . 50
r−1
and of course xp ∈
46 αp − a = 0 as soon as α is a root of X p − a. Thus a = αp = Fp (α).
47 The root bi is a root of the following equation X p − ai i.e. bpi − ai = 0 or ai = Fp (bi ).
48 In other case each root will have at least multiplicity pr .
49 We have xpr = a thus polynomial X pr − a can be written as X pr − a = X pr − xpr =
pr
(X − x) thus x has multiplicity pr
50 as soon as any power of x (little x but not the big one X )
About the polynomial is irreducible. We have already seen that in the

case [K (x) : K] = pr so the polynomial is irreducible 51 and this nishes 52
the proof.
51 Corollary 3.16 says that there exists an irreducible polynomial of degree pr with x as
r
the root. Theorem A.89 says that the polynomial should divide our polynomial X p − a
as soon as they have the same root. The two polynomial have same degree and as result
they are the same (up to a constant). Therefore the considered polynomial is irreducible.
52 Because we found an irreducible polynomial that is not separable because has a root
of multiplicity pr
Chapter 4
Tensor product. Structure of
nite K-algebras
This is a digression on commutative algebra. We introduce and study the
notion of tensor product of modules over a ring. We prove a structure the-
orem for nite algebras over a eld (a version of the well-known "Chinese
remainder theorem").
4.1 Denition of tensor product

4.1.1 Summary for previous lectures
We considered nite Field extension L i.e [L : K] < ∞. We also saw that if
L is generated by a nite number of Separable elements α1 , . . . , αr then the
number of Homomorphisms over K from L to K̄ denoted by HomK L, K̄

is equal to [L : K]. In general

[L : K]sep = HomK L, K̄ ≤ [L : K] .
For L = K (α) it is clear because the number of homomorphisms is equal

to the number of roots of the Minimal polynomial Pmin (α, K). In general
one can use induction and multiplicativity of the degree [L : K] and number
of homomorphisms (see theorem About separable extensions (Remark 3.29)
). Thus separable extension was exactly an extension which had the right
number of homomorphisms into the algebraic closure.
Our next goal is to characterize the separability in the terms of tensor
product.
49
50CHAPTER 4. TENSOR PRODUCT. STRUCTURE OF FINITE K-ALGEBRAS
4.1.2 Tensor product

Denition 4.1 (Tensor product). Let A is a ring, N , M are A-Modules. The
tensor product M ⊗A N is another A-Module together with an A-bilinear map
φ : M × N → M ⊗A N which has Universal property dened below
Denition 4.2 (Universal property). A-bilinear map φ : M × N → M ⊗A N
has universal property if ∀P - A-Module and for A-bilinear f : M ×N → P (
i.e. ∀m, fm : N −−−−−−→ P and ∀n, fn : M −−−−−−→ P are Homomorphisms
n→f (m,n) m→f (m,n)
of A-modules ), then ∃!f˜ - homomorphism of A-modules such that f = f˜ ◦ φ
1
f
M ×N P
φ f˜
M ⊗A N
The property
characterize the pair (φ, M ⊗ N ). Really if have another
pair φ, M ⊗ N like this one then by denition we have mutually inverse
homomorphisms of A-modules between them
Lemma 4.3 (About uniqueness of object dened by universal property).
2
If we have two objects(φ, M ⊗ N ) and φ, M ⊗ N which both satises
Universal property than there is an unique Isomorphism between them:
(φ, M ⊗ N ) ∼

= φ, M ⊗ N
Proof.Let P = M ⊗ N and f = φ. In the case we can consider the following

diagram
M ⊗A N
φ
˜
g=φ
φ
M ×N M ⊗A N
φ
ḡ = φ̃
M ⊗A N
1That means that we have a Commutative diagram there

2It is out of the lecture video and can be considered as an explanation for the claim
about having mutually inverse homomorphisms of A-modules. The proof was taken from
[6].
4.1. DEFINITION OF TENSOR PRODUCT 51
As soon as we xed M ⊗A N we 2 unique homomorphisms (which are

dened by the xed M ⊗A N ) - g : M ⊗A N → M ⊗A N and ḡ : M ⊗A N →
M ⊗A N . Both g and ḡ are linear and, as mentioned above, the pair is unique
(if we x g we will have only one ḡ that corresponds to g ). The composition
g ◦ ḡ maps M ⊗A N to itself. I.e. ∀x ∈ M × N we have ḡ (g (φ (x))) = φ (x).
If y = φ (x) then ḡ (g (y)) = y . The last equation holds ∀y ∈ Im (φ) i.e.
∀y ∈ M ⊗A N . Therefore ḡ = g −1 . 3
Thus we have an Isomorphism and the isomorphism is unique as soon as
the function g is unique due the Universal property.
We just prove an isomorphism existence between M ⊗ N and M ⊗ N but
the tensor product is characterized not only by the module M ⊗ N but also a
bilinear map φ. Let P = M ⊗ N thus we can get that φ̄ = φ̄˜ ◦φ is determined
by the unique relation φ → φ̄ as soon as φ̄˜ is unique. Analogues one can get
the unique relation φ̄ → φ.
The uniqueness does not mean existence and we should proof that such
object exists.
Lemma 4.4 (About tensor product existence). Tensor product dened via
Universal property exists
Proof. Lets consider E the maps (functions) from M × N to A as sets which

are 0 almost everywhere (i.e. outside of a nite set). For example we can
consider delta functions:
δm,n : M × N → A
such that
δm,n (m, n) = 1,
δm,n (m , n ) = 0 if (m, n) 6= (m0 , n0 )
0 0
Then E is a A-Free module with basis δm,n . Thus we have a map of sets
M × N → E such that (m, n) → δm,n which is not bilinear but we can make
it bilinear by means of changing E .
Let F ⊂ E a submodule generated by δm+m0 ,n − δm,n − δm0 ,n , δm,n+n0 −
δm,n − δm,n0 , δam,n − aδm,n , δm,an − aδm,n . 4
3 Another explanation is the following: thus if we x g and choose ḡ = g −1 we will get
g ◦ ḡ = idM ⊗A N that satised all requirements. The choice is nal because we don't have
a possibility to choose any other ḡ (it should be unique).
4 The basis is chosen to be a bilinear mod F , for instance δ
m+m0 ,n = δm,n + δm0 ,n
mod F
It can be shown that M × N → E/F is bilinear 5 and has the desired

Universal property.
Really lets we have the following bilinear map: f : M × N → P . Then
we can consider the following linear map (Homomorphism) f 0 : E → P that
sends δm,n to f (n, m). Using the fact that f is bilinear we can get
f 0 (δm+m0 ,n ) = f (m + m0 , n) = f (m, n) + f (m0 , n) =

= f 0 (δm,n ) + f 0 (δm0 ,n ).
With the same approach one can get the following relations
f 0 (δm,n+n0 ) = f 0 (δm,n ) + f 0 (δm,n0 ),

f 0 (δam,n ) = af 0 (δm,n ),
f 0 (δm,an ) = af 0 (δm,n )
with the f 0 linearity we have
f 0 (δm+m0 ,n ) = f 0 (δm,n + δm0 ,n ),

f 0 (δm,n+n0 ) = f 0 (δm,n + δm,n0 ),
f 0 (δam,n ) = af 0 (δm,n ),
f 0 (δm,an ) = af 0 (δm,n )
The kernel ker f 0 = F thus if we want to have a homomorphism to P we

have to replace E with E/F that is also denoted by M ⊗A N . In the case we
will replace f 0 with f˜ (δm,n mod F) = f (m, n). As soon as the images for
the basis is xed the mapping is unique.
We will denote φ (m, n) = δm,n mod F as m ⊗ n. I.e our tensor product

can be considered as the (⊗, M ⊗A N ) pair.
Remark 4.5. Wrong idea is to dene M ⊗A N as a set of m ⊗ n. I.e.

M ⊗A N 6= {mPk n}. The M ⊗A N is generated by m ⊗ n i.e. ∀x ∈ M ⊗A N
⊗
we have x = i=1 mi ⊗ ni i.e. each element is a nite sum of m ⊗ n and I
cannot reduce these further 6 .
5 Follows from the basis choice
6 i.e. ∃x ∈ M ⊗A NPsuch that ∃!m ∈ M, n ∈ N : x = m ⊗ n but ∃m1 , . . . , mk ∈
k
M, n1 , . . . , nk ∈ N : x = i=1 mi ⊗ ni
4.2. TENSOR PRODUCT OF MODULES 53
4.2 Tensor product of modules

4.2.1 Advantages of the universal property
Now, you can ask why haven't I just dened the tensor product by this
construction? Why am I talking of this universal property? And the answer
is because it is easier to prove things this way. So advantages of the universal
property is as follows: the proofs become easy.
4.2.2 Several examples of universal property usage

Example 4.6 (Commutativity proof). We want to prove that
M ⊗A N ∼
= N ⊗A M
We have the following bilinear map: M × N → N ⊗A M for which the

pair (m, n) is mapped to n ⊗ m. Thus from Universal property we have that
there is a linear map (homomorphism) α : M ⊗A N → N ⊗A M :
(m, n) → n ⊗ m
M ×N N ⊗A M
(m, n) → m ⊗ n α
M ⊗A N
With the same construction we can also get the inverse map a−1 that
sends N ⊗A M to M ⊗A N :
(m, n) → m ⊗ n
M ×N M ⊗A N
(m, n) → n ⊗ m α−1
N ⊗A M
Also
Corollary 4.7.
A ⊗A M ∼
=M
Proof.For the proof 7 lets look at A. Really A can be considered as A-module

because all requirements from denition A.99 are satised. The following
7 The proof is missed in the lectures
diagrams shows that there exist 2 homomorphisms: α : A ⊗A M → M and

α−1 : M → A ⊗A M as result there is a homomorphism A ⊗A M ∼ = M:
(a, m) → a · m (a, m) → a ⊗A m
A×M M A×M A ⊗A M
(a, m) → a ⊗A m α (a, m) → a · m α−1
A ⊗A M M
In the diagrams m ∈ M , as usual, and a ∈ A.
If we have that M is generated by e1 , e2 , . . . and N is generated by

1 , 2 , . . . than M ⊗A N is generated by pairs ei ⊗ j . It's obvious.
More complex fact is the following
Proposition 4.8. Let M and N are Free modules with corresponding basises
e1 , e2 , . . . , en and 1 , 2 , . . . , m than M ⊗A N is also free module with basis
ei ⊗ j where 1 ≤ i ≤ n and 1 ≤ j ≤ m.
Lets dene fi0 ,j0 : M × N → A as a map that sends ( ai ei , bj j )
P P
Proof.
to ai0 bj0 . It's bilinear 8 so it factors through the tensor product f˜i0 ,j0 :
M ⊗A N → A. The map f˜i0 ,j0 sends ei0 ⊗ j0 to 1 and all others to 0. 9 So if
X
αij ei ⊗ j = 0
10
then applying f˜i0 ,j0 for all indices one can get that ∀i, j : αij = 0.
In particular for the Vector space the tensor product is dened in the
same way (as just proved in the proposition 4.8): the tensor product of 2
vector spaces with basises e1 , e2 , . . . , en and 1 , 2 , . . . , m is another vector
space with the following basis ei ⊗ j i.e. the denition does not take into
consideration the Universal property.
8 for example ( (ai + a0i )ei , bj j ) is sent to (aj0 + a0j0 )bj0 .
P P
9 Because fi0 ,j0 = f˜i0 ,j0 φ i.e.
X X
ai0 bj0 = fi0 ,j0 ai ei , bj j =
X X
= f˜i0 ,j0 φ ai ei , bj j =
X X X
= f˜i0 ,j0 ai ei ⊗ bj j = ai bj f˜i0 ,j0 (ei ⊗ j ).
i,j
10 because f˜ should be linear.

4.3. BASE CHANGE 55
Proposition 4.9 (Associative).
(M1 ⊗A M2 ) ⊗A M3 ∼
= M1 ⊗A (M2 ⊗A M3 )
Proof. There is just a scratch of the proof. Introduce M1 ⊗A M2 ⊗A M3 as

a universal object for 3-linear maps and show that 2 considered parts are
isomorphic each other.
4.3 Base change

Let A is a Ring and B is A-algebra. Let also M is an A-Module and N is
B -module.
I can of course make N into A-module (just forgetting the additional
A-algebra structure). But we can also make B -module on M (that is not
a trivial thing) by considering B ⊗A M 11 . We can introduce B -module
structure on B ⊗A M by 12
b · (b0 ⊗ m) = (b · b0 ) ⊗ m
Example 4.10 (The complexication of a real vector space). We can make

R2n from Cn by forgetting the complex structure. 13 The Cn has the following
2n
basis e1 , . . . , en . The R has the following one e1 , . . . , en , ie1 , . . . , ien . Now
√
we forgot about multiplication rules for i = −1 and denote iei as vi . In the
2n
case the basis for R is the following one: e1 , . . . , en , v1 , . . . , vn .
But we can also do the following constructions
Rn → Cn = C ⊗ Rn → R2n
for the Cn basis we have 1C ⊗ e1 , . . . , 1C ⊗ en and for R2n - 1 ⊗ e1 , . . . , 1 ⊗

en , i ⊗ e1 , . . . , i ⊗ en . 14
11 In other words we can make a B -module from A-module M
12 I.e. we introduced B -algebra operations for objects from B ⊗A M . See also denition
1.1.
13 In the case we have ring A = R and B = C - A algebra. A - module is the following
vector space M = Rn and B - module is N = Cn .
14 some additional clarication: ∀x ∈ C ⊗ Rn we have
n
X n
X n
X
x= ci ⊗ ri ei = ci ri 1C ⊗ ei = c0i 1C ⊗ ei ,
i=1 i=1 i=1
where ci , c0i = ci ri ∈ C, ri ∈ R. Thus we just got Cn . From other side we can write ci as
Proposition 4.11. In general we have the following. If M - free A -

module with basis e1 , . . . , e n then B ⊗A M is a free B module with basis
1B ⊗ e1 , . . . , 1B ⊗ en .
Proof. The proof is the same as at proposition 4.8. Again we construct

certain bilinear maps and say that those factor over the tensor product and
this implies that certain families are linearly independent.
Really lets dene bilinear map fi0 : B × M → A such that
n
!
X
fi0 b, mi ei = bmi0 ei0
i=1
so there exists a linear map f˜i0 (homomorphism) such that fi0 = f˜i0 φ or
n
! n
!!
X X
fi 0 b, mi ei = f˜i0 φ b, mi ei
i=1 i=1
n
! n
!
X X
= f˜i0 b⊗ mi ei = bf˜i0 1B ⊗ mi ei = bmi0
i=1 i=1
i.e.
P it sends 1B ⊗ ei0 to 1 and all others 1B ⊗ ei to 0. Thus the following sum
αi 1B ⊗ ei is equal to 0 if and only if αi = 0 i.e. αi 1B ⊗ ei forms a basis.
Remark 4.12. We have the following maps.
• For A - modules: α : M −−−−−−−→ B ⊗A M which makes a B -module

m→1B ⊗A m
from an A-module.
• For B - modules: µ : B ⊗A N −−−−−→ N .

b⊗n→bn
Theorem 4.13 (Base-change). Let A is a Ring and B is A-algebra. Let also

M is an A-Module and N is B -module.
HomA (M, N ) ↔ HomB (B ⊗A M, N )

follows: ci = ai + ibi , where ai , bi ∈ R. Therefore
n
X n
X
x= ai ri 1 ⊗ ei + bi ri i ⊗ ei .
i=1 i=1
I.e. x ∈ R2n and the basis in R2n is formed by 1 ⊗ e1 , . . . , 1 ⊗ en , i ⊗ e1 , . . . , i ⊗ en .

4.3. BASE CHANGE 57
15
I.e. the homomorphisms are the same or in other words the corresponding
groups of homomorphisms are isomorphic:
HomA (M, N ) ∼
= HomB (B ⊗A M, N )
Proof. First of all we have 16 Homomorphism f : B ⊗A M → N . We also have

the following map (see remark 4.12): α : M → B ⊗A M . Thus f · α : M → N
i.e. we can set the following relation
fˆ : HomB (B ⊗A M, N ) → HomA (M, N )
such that fˆ (f ) = f α.
In other direction we have g : M → N thus idB ⊗ g : B ⊗A M → B ⊗A N
but ( see remark 4.12) we have µ : B ⊗A N → N i.e. we have the following
relation
ĝ : HomA (M, N ) → HomB (B ⊗A M, N )
such that
ĝ(g) = µ · (idB ⊗ g).
And we can check that those maps (fˆ and ĝ ) are mutually inverse. For the
proof 17 the fact consider the following diagram
M fα N
g
α f µ
B ⊗A M idB ⊗ g B ⊗A N
One can conclude, as soon as the diagram commutes
M g N
α µ
B ⊗A M idB ⊗ g B ⊗A N
fˆ (ĝ (g)) = µ · (idB ⊗ g) · α = g.

18
I. e. fˆ ◦ ĝ = id or in other words fˆ and ĝ are mutually inverse.
15 A-homomorphisms between A modules HomA (M, N ) are the same as B -
homomorphisms between B modules HomB (B ⊗A M, N ).
16 One homomorphism from Hom (B ⊗ M, N )
B A
17 It's missed in the lectures
18 Operation ◦ is dened as follows (â ◦ b̂)(x) = â(b̂(x)) where â, b̂ are 2 maps acting on
a set X and x ∈ X .
4.4 Examples. Tensor product of algebras

Proposition 4.14. If I ⊂ A - is an Ideal so my B - A algebra will be
B = A/I then
A/I ⊗A M ∼
= M/IM
where IM is a sub-module of M.
Proof. We have map α : M → B ⊗A M = A/I ⊗A M (see remark 4.12) which
sends m to 1̄ ⊗ m. 19 The map sends IM to 0 because ∀i ∈ I, m ∈ M : im →
1̄ ⊗ im = ī ⊗ m because the tensor product is over A and everything is A
linear and as result 1̄ ⊗ im = ī ⊗ m, but ī ⊗ m = 0̄ ⊗ m = 0. 20 Thus α sends
IM to 0. So α induces ᾱ : M/IM → A/I ⊗A M such that ᾱ (m̄) = 1̄ ⊗ m.
For other direction we apply Base-change (Theorem 4.13) . The following
map (projection) of A-modules
m→m̄
M −−−→ M/IM
21
gives us the following map of B -modules
β̄ : B ⊗A M → M/IM
i.e.
β̄ : A/I ⊗A M → M/IM
22
that sends ā ⊗ m to am
¯ Ones check again that this inverse to ᾱ.
Several examples:
Example 4.15. Let Z/2Z ⊗Z Z/3Z what will we obtain?
Z/2Z⊗Z ∼
=Z/3Z /(2)·Z/3Z
23
but 2 is invertible: 2−1 = −1 mod 3 thus (2)Z/3Z = Z/3Z and as result
Z/2Z⊗Z ∼
=Z/3Z /Z/3Z = 0
19 1̄ = 1A + I
20 because ī = 0 mod I
21 the Base-change (Theorem 4.13) says that homomorphism of A-modules M → N
is isomorphic to homomorphism of B modules B ⊗A M → N . Using N = M/IM we can
get that B ⊗A M → M/IM is isomorphic (i.e. the same) to M → M/IM . We also can
get the same result just ignore B : B ⊗A M → M → M/IM .
22 For example β̄ (ᾱ (m̄)) = β̄ (1̄ ⊗ m) = 1 · m = m̄
23 I.e. there exist an invertible element 2−1 ∈ Z/3Z = F therefore 1
3 F3 ∈ 2 · F3 or
2 · F3 = F3 and as result (2) · F3 = F3 . I.e. (2) is not a Proper ideal in F3 because it is
equal to F3 .
4.4. EXAMPLES. TENSOR PRODUCT OF ALGEBRAS 59
Example 4.16. Another obvious example

24
B ⊗A A [X] ∼
= B [X]
and more interesting one
B ⊗A A [X] /(P ) ∼
= B [X] /(P ),
there (P ) becomes an ideal generated by P in B [X].
4.4.1 Tensor product of A-algebras

Let B, C are A-algebras. The following maps form an algebra structure on
A:
α:A→B
β:A→C
25
New A-algebra B ⊗A C : is a ring with respect to the following operation
(b ⊗ c) · (b0 ⊗ c0 ) = (b · b0 ) ⊗ (c · c0 ) (4.1)
The tensor product has the following
Denition 4.17 (Universal property). Let we have the following maps
α : A → B,
β : A → C,
φ : B −−−−−→ B ⊗A C,
b→b⊗1C
ψ : C −−−−−→ B ⊗A C
c→1B ⊗c
Then for any A-algebra D one has
HomA (B ⊗A C, D) ↔ HomA (B, D) × HomA (C, D)

24 B ⊗A A [X] has the following B -basis: {1B ⊗ X i } thus ∀b ∈ B ⊗A A [X] we can get
X
b= bi · 1B ⊗ X i
and there is an obvious isomorphism
f : B ⊗A A [X] −−−P
−−−−→
i
B [X]
b→ bi X
25 that makes it A-algebra (see K-algebra)

i.e. if I have some Homomorphism h ∈ HomA (B ⊗A C, D) this is the same

as giving 2 homomorphisms f ∈ HomA (B, D) and g ∈ HomA (C, D) such
that all maps in the following diagram commute (see Commutative diagram).
h
D B⊗C
f φ ψ
g
B C
α β
A
Thus if we have h then we can dene f = h · φ and g = h · ψ . And
conversely if I have f and g then I can dene h by the following rule:
h (b ⊗ c) = f (b) · g(c)
The main point for us is that the tensor product of the A-algebras is
itself an A-algebra by this very simple rule, component-wise multiplication
(see (4.1)).
Let consider next example. We will start with the following
C∼ = R [X]/ X 2 + 1

therefore with result from example 4.16 one can get

C ⊗R C ∼ = C ⊗R R [X]/ X 2 + 1 ∼
C[X]
= /(X 2 +1)
but by Chinese remainder (Theorem 4.20)
C[X]
=C[X] /(X+i) ×C[X] /(X−i) ∼
/(X 2 +1) ∼ =C×C
As result we have that C ⊗R C is not a eld because it has zero divisors.
How we can get the zero divisors? The element X + i is a zero divisor in
C[X]
/(X 2 +1) because
(X + i) (X − i) ≡ 0 mod X 2 + 1

.
Another proof (not a part of the lecture) of the fact that C ⊗R C is not
a eld consider the following one
C ⊗R C ∼
=C[X] /(X 2 +1)
but the polynomial X 2 + 1 = (X + i) (X − i) is reducible in C [X] i.e. is not
a Maximal ideal (see theorem A.90) and with the theorem A.94 the quotient
by the polynomial is not a eld (see also claim 1.11).
4.5. RELATIVELY PRIME IDEALS. CHINESE REMAINDER THEOREM61
4.5 Relatively prime ideals. Chinese remainder

theorem
Denition 4.18 (Relatively prime ideals). Let A - Ring and I, J are Ideals.
I and J are relatively prime if I + J = A.
Lemma 4.19. 1. If I, J are relatively prime then IJ = I ∩ J

Qk
2. If I1 , . . . , Ik relatively prime with J then i=1 Ii = I1 . . . Ik is also rel-
atively prime with J.
3. If I, J relatively prime then Ik and Jl are also relatively prime for any
l and k.
Proof. 1. The following one IJ ⊂ I ∩ J is clear 26 If I and J are relatively

prime then 1A = i + j for some i ∈ I and j ∈ J . Thus ∀x ∈ I ∩ J we
have the following ones: xi ∈ IJ and xj ∈ IJ and as result
x = xi + xj ∈ IJ
i.e. I ∩ J ⊂ IJ .
2. Suppose for simplicity that k = 2. In the case we have 1A = i1 + j1 =

i2 + j2 where i1 ∈ I1 , i2 ∈ I2 and j1 , j2 ∈ J . we also have
1A = (i1 + j1 ) (i2 + j2 ) = i1 i2 + (j1 i2 + j2 i1 + j1 j2 ) ∈ I1 I2 + J
thus ∀x ∈ A we have
x = 1A x = i1 i2 x + (j1 i2 + j2 i1 + j1 j2 ) x ∈ I1 I2 + J,
i.e. I1 I2 + J = A therefore I1 I2 and J are relatively prime.
27
3. is obvious
26 Assuming that I and J commute we have if x ∈ IJ then x ∈ I and if x ∈ JI then

x ∈ J i.e. x ∈ I ∩ J .
27 It follows from the 2 because we can assume I = I and will get that ∀k, I k is relatively
i
prime with J . From other side we can assume Ii = J and J = I k and conclude that J l is
relatively prime with I k .
Theorem 4.20 (Chinese remainder). Let I1 , . . . , In - ideals and map π :

A → A/I1 × · · · × A/In dened as follows
π(a) = (a mod I1 , . . . , a mod In ) (4.2)
The kernel ker π = I1 ∩ · · · ∩ In .

The π is Surjection if and only if I1 , . . . , In are pairwise relatively prime.
In that case
A/ ∩ Ik ∼ Ik ∼
Y Y
= A/ = (A/Ik )
28
Proof. Let π is Surjection. In the case ∃ai ∈ A such that
π(ai ) = (0, . . . , 1( in i-th place ), 0, . . . , 0)
i.e. ai mod Ij = 0 or ai ∈ Ij for i 6= j . We also have ai mod Ii = 1 thus

1 − ai = kIi i.e. 1 − ai ∈ Ii . Thus ∀j, ∃ai ∈ Ij , ak ∈ Ii such that 1 = ai + ak
thus A = Ij + Ii i.e. Ii relatively prime with any Ij .
Conversely if Ii is relatively prime with any Ij where Q j 6= i then it also
Q with the product (see lemma 4.19) j6=i Ij . In the case
relatively prime
∃xi ∈ Ii , yi ∈ j6=i Ij such that 1 = xi + yi in the case
π(yi ) = (0, . . . , 1( in i-th place ), 0, . . . , 0)
and ∀bi ∈ A/Ii !

n
X
π bi y i = (b1 , . . . , bn )
i=1
i.e. π is surjective.
Let K is a eld and A is a nite (nite dimensional vector space) K -

algebra.
Proposition 4.21. 1. If A is an Integral domain then A is a eld.
2. (replacing the rst one) Any Prime ideal of A is a Maximal ideal
28
See First isomorphism (Theorem A.134) where G = A, H = A/I1 × · · · × A/In ,
φ = π and with lemma 4.19 (as soon as Ik pairwise relatively prime)
ker φ = ker π = I1 ∩ · · · ∩ In = I1 . . . In .
4.6. STRUCTURE OF FINITE ALGEBRAS OVER A FIELD. EXAMPLES63
Proof. Well, I shall prove only the rst part, the second part is just a conse-
quence of denitions. In fact, a factor over a prime ideal, a quotient over a
prime ideal is an integral domain, and a quotient over a maximal ideal is a
eld. 29 If you don't know this, please look it up in any book.
Lets prove the rst part. Integral domain means that there is no zero
divisors i.e. ∀a ∈ A 30 multiplication by a is Injection. A is nite dimen-
sional Vector space (see above) that implies that ×a is an Isomorphism, 31
in particular Surjection i.e. ∃b ∈ A such that b × a = 1 i.e. a is invertible
therefore A is eld.
4.6 Structure of nite algebras over a eld. Ex-

amples
Remark 4.22. 32
Let A is a K -algebra and m is a maximal ideal of A.
Then A/m is also K -algebra.
Proof. Lemma 1.5 says that there is a ring homomorphism between K and
A: fK : K → A. There is also a canonical homomorphism fA : A −−→ A/m.
a→ā
Therefore we can dene f = fA fK : K → A/m and therefore with lemma
about K-algebra and homomorphism (Lemma 1.5) can conclude that A/m
is a K -algebra. 33
Note that if A is a eld then fA is injection, m = {0K } and A = A/m.
Theorem 4.23 (Structure of nite K -algebra). Let A be a nite K -algebra

i.e. dimK A < ∞. Then
1. There are only nitely many Maximal ideals m1 , . . . , mr in A

34
2. Let J = m1 ∩ · · · ∩ mr = m1 . . . mr . Then Jn = 0 for some n
3. A∼
= A/mn1 1 × · · · × A/mnr r for some n1 , . . . , nr .
29 i.e. prime ideal is a maximal ideal

30 a 6= 0A
31 ×a sends a vector space into another vector space with the same dimension. But
with About vector space isomorphism (Lemma A.110) one can get that the spaces are
isomorphic each others and as result the operation ×a is an Isomorphism.
32 The remark is not a part of the lectures but it is important to understand the below
content.
33 Thanks Zonglin Jiang for the proof.
34 Since the ideals are relatively prime the intersection is the same as the product of the
ideals
Proof. 1. Let m1 , . . . , mi are maximal ideals. By Chinese remainder (The-

orem 4.20) we have 35
A/m1 . . . mi ∼
= A/m1 × · · · × A/mi .
We know that A as well as A/m1 . . . mi and A/mk are nite dimensional

K -Vector space 36 thus we have the following relations
i
X
dimK A ≥ dimK A/m1 . . . mi = dimK A/mj ≥ i.
j=1
Therefore if N the number of maximal ideals then dimK A ≥ N i.e.

the number of maximal ideal is limited by the vector space dimension.
2. J = m1 ∩ · · · ∩ mr = m1 . . . mr is nite dimensional vector space over

K 37 as well as its powers J k . We have the following sequence 38
· · · ⊆ J k ⊆ · · · ⊆ J 2 ⊆ J.
and the sequence should stop somewhere 39 i.e. ∃n such that J n = J n+1 .
We claim that J n = 0 in the case. Indeed if not we have the following
basis of J n : e1 , . . . , es . And as soon as J n = JJ n we can write a vector
ei ∈ J n as a vector from J n multiplied on an object from J i.e.
X
ei = λij ej ,
there ej ∈ J n , λij ∈ J . Thus if M = id − λij

 
e1
M ·  ...  = 0.
 
es
35 Maximal ideals are relatively prime because in a commutative ring with unity, every
Maximal ideal is a Prime ideal see also proposition 4.21.
36 See remark 4.22 and take into consideration the theorem statement (see above) that
says dimK A < ∞. A/m1 is a projection i.e. dimK A/m1 < dimK A < ∞.
37 We have m ⊂ A i.e. dim m ≤ dim A < ∞. From other side J = m . . . m =
i K i K 1 r
(m1 ∩ · · · ∩ mr ) ⊂ mi , ∀i ∈ {1, . . . , r}. Thus dimK J ≤ dimK mi ≤ dimK A < ∞ ≤
38 Let j ∈ J ⊂ A and x ∈ JJ . Then ∃j ∈ J such that x = jj but jj ∈ J because
∀y ∈ J : jy ∈ J . As result J 2 ⊆ J .
39 On each step we should decrease the dimension if we don't stop. The dimension is
limited and as result the sequence should stop.
4.6. STRUCTURE OF FINITE ALGEBRAS OVER A FIELD. EXAMPLES65
It's possible over ring to nd a matrix M̃ such that M̃ M = det M · id,
40
i.e.  
e1
 ..
det M ·  .  = 0. (4.3)

es
But det M = 1 + λ where λ ∈ J . 41 Since J = m1 ∩ · · · ∩ mr then
∀i : λ ∈ mi so @i such that 1 + λ ∈ mi 42 thus 1 + λ is invertable 43
therefore e1 = · · · = es = 0 44
3. Using part 2 ∃n1 , . . . , nr such that mn1 1 . . . mnr r = 0 (for example we can
assume ni = n). Then by Chinese remainder (Theorem 4.20)
A∼
= A/mn1 1 × · · · × A/mnr r .
We used the following facts:

45
• A = A/mn1 1 . . . mnr r
46
• mni i are pairwise relatively prime
Remark 4.24. The ni s are not uniquely dened. For example (see also
example 5.1)
A =K[X] /(X 2 (X+1)3 ) .
We have 2 ideals there: m1 = (X) and m2 = (X + 1). We of course have
A∼
= A/m21 × A/m32
40 If M is a matrix over a eld we can consider the following 2 cases:
(a) det M = 0: we can assume that M̃ = 0
(b) det M 6= 0: at the case ∃M −1 and therefore M̃ = det M · M −1 will work.
For more common case we have to look at Adjugate matrix [60]
41 Because the det consists of the following items Q(1 − λ ) = 1 + (−1)s Q λ and
ii ii
λij . The sum of the items (det) consists of 1 and another sumP inQ
which all items are
Q
from J . Thus the second sum is an element of J i.e. det M = 1 + λij = 1 + λ.
42 We have that m is a Maximal ideal and therefore (by its denition) it is a Proper
i
ideal i.e. 1 ∈/ mi . From other side if 1 + λ ∈ mi then 1 + λ − λ ∈ mi as soon as λ ∈ mi .
I.e. we have a contradiction.
43 0 ∈ m thus if 1 + λ ∈/ mi then 1 + λ 6= 0.
i
44 Because det M = 1 + λ 6= 0
45 Because A = A/{0}. For example if I = {0} and x ∈ A then x̄ ∈ A/I if x̄ = x + I .
In our case x̄ = x + {0} = x i.e. ∀x ∈ A we have x ∈ A/{0}. (See also Quotient ring)
46 As soon as {m } - Maximal ideals and as result Prime ideals then with lemma 4.19
i
n
one can get that ∀i 6= j mni i is relatively prime with mj j .
but also we have

A∼
= A/m31 × A/m32
47
as soon as m21 = m31 in A: (X)2 ⊂ (X)3 but also (X)3 ⊂ (X)2
Several examples:
C ⊗R C = C × C.
Another example
√ √ √ √
Q 2 ⊗Q Q 3 =Q 2, 3
And you see that those algebras are Cartesian products of elds. So all ni 's
may be taken equal to 1 48 . In other words, we don't have Nilpotent elements
in our algebra 49 . So, it is a reduced algebras. Reduced, by denition, is with-
out nilpotents. It's general phenomena because the presence of nilpotents is
due to the inseparability of extensions come from inseparable extensions.
47 (X)3 ⊂ (X)2 - this is true for any polynomial P (X) because if P (X) ∈ (X)3 then
P (X) = X 3 P 0 (X) = X 2 P̄ (X) ∈ (X)2 where P̄ (X) = XP 0 (X)
(X)2 ⊂ (X)3 is the more complex one and it does not true for any polynomial ring
2
but this is true for our A. Let P (X) ∈ (X) ⊂ K [X] then P (X) = X 2 P 0 (X) but
X 2 P 0 (X) ≡ 0 mod X 2 . From other side X 4 P 0 (X) ≡ 0 mod X 2 therefore
P (X) ≡ X 3 XP 0 (X) mod X 2
i.e. P (X) ∈ (X)3 .

48 Because A/mn (where m is a maximal ideal) is a eld if n = 1
49 Cartesian products of elds has no nilpotent elements except 0 [17].
Chapter 5
Structure of nite K-algebras
continued
We apply the discussion from the last lecture to the case of eld extensions.
We show that the separable extensions remain reduced after a base change:
the inseparability is responsible for eventual nilpotents. As our next subject,
we introduce normal and Galois extensions and prove Artin's theorem on
invariants.
5.1 Structure of nite K-algebras, examples (cont'd)

Last time we have seen that a nite K -algebra A ([A : K] < ∞) has only
nitely many maximal ideals m1 , . . . , mr and the following equation holds
(see theorem 4.23):
A∼= A/mk1 × · · · × A/mkr
1 r
This is a general form of Chinese remainder (Theorem 4.20) .
Example 5.1. Let
A = K [X] / (F )
And the polynomial F

is not necessary irreducible so let's decompose into a
k k
product of irreducible factors: F = P1 1 . . . Pr r . Then by the Chinese remain-
1
der (Theorem 4.20) one can get
A∼ k k
= K [X] / (P1 ) 1 × · · · × K [X] / (Pr ) r ,
1 See also remark 4.24 and theorem 4.23
67
68CHAPTER 5. STRUCTURE OF FINITE K-ALGEBRAS CONTINUED
2
where K [X] / (Pi )ki = A/mki i and mi = (Pi mod F ) - an ideal.
Denition 5.2 (Nilpotent element). Let A is a Ring than x ∈ A is nilpotent

3
if x 6= 0 but ∃k : x = 0.
k
Denition 5.3 (reduced). K -algebra A is reduced if it has no Nilpotent

elements. Or in other words 4 if in the decomposition
A∼
= A/mk11 × · · · × A/mkr r
5
∀i : ki = 1. Or if A is a product of elds.
Denition 5.4 (local). Ring A is called local if it has only one Maximal
∼
ideal i.e. A = A/mk .
If A is local then all elements of A are nilpotents i.e. any element of A is

a identity,zero or nilpotent 6 7 .
Most of our last examples were examples of reduced K -algebras such as
C ⊗R C = C × C
2Using denition A.83 one can get that Pi ∈ K [X] corresponds to Pi mod F in
A = K [X] / (F ) therefore we have (Pi ) is a Maximal ideal for K [X] and
ki k
A/ (Pi ) = K [X] / (Pi mod F ) i .
but Pi mod F = Pi and as result
K [X] / (F ) ∼
k k
= K [X] / (P1 ) 1 × · · · × K [X] / (Pr ) r .
3 Alternative denition from [50]: An element, x, of a ring, R, is called nilpotent if

there exists some positive integer, n, such that xn = 0.
4 Let we have an i-th element of the product Q A/mki with k > 1 and m = (p) when
i i i
p ∈ A/mki i and p 6= 0 = pki i.e. p is a nilpotent.
5 A/m is a eld as soon as m is a Maximal ideal
i i
6 As it was mentioned in [47], a nonzero ring in which every element is either a unit or
nilpotent is a local ring, but not reverse, as it was pointed on the lectures. There is also
an example A ∼ = A/{0} where A is a Field and {0} is the only maximal ideal for the elds
(see example A.77). In this case there are many non nilpotents dierent from identity and
zero but the ring (K -algebra) is local.
7 Comment from Sta on the issue: It looks confusing because Katya immediately
applies this denition to the case of nite algebras without explicitly mentioning.
You are right that it is not true in general. For example, discrete valuation rings are
local and have no zero divisors at all. (Also your counterexample is not quite correct,
because a eld is exactly the case when every element is either invertible or nilpotent).
But if a nite algebra has only one maximal ideal, then by the structural theorem it
consists of nilpotents.
5.2. SEPARABILITY AND BASE CHANGE 69
or √ √
Q 2 ⊗Q Q (i) = Q i, 2
that is a eld and if we start producing similar examples then mostly they
are reduced. Well, why? Because in fact the presence of nilpotents has to do
with inseparability. The presence of nilpotents reects inseparability.
So let me give you one more example: tensor product of extensions which
is not reduced. Let K be a eld of characteristic p, for instance Fp . Consider
a eld of rational functions over K 8 : K (X). We will consider K (X) as
an extension of K (X p ) (or with new variable Y = X p - K (Y ). We will be
interested in K (X) ⊗K(Y ) K (X) where X is a pth root of Y so 9
K (X) ⊗K(Y ) K (X) ∼ =

∼
= K (X) ⊗K(Y ) K [T ] / (T − Y ) ∼
p
=
∼ p
= K (X) [T ] / (T − Y ) =
= K (X) [T ] / (T − X p ) = K (X) [T ] / (T − X)p
p
where T is another variable. As result we have got a ring with nilpotents for
example T − X and of course the reason is that our extension K (X) is pure
inseparable extension (see denition 3.28) of K (Y ).
5.2 Separability and base change

What is the reason for such a mysterious connection between presence of
nilpotents and separability? IfL is separable over K then the number of
Homomorphisms HomK L, K̄ is maximal and equal to degree [L : K] but

in general it is less or equal to the degree. This is of course clear, because if
we have a polynomial with distinct roots, then it's stem eld for instance has
exactly this number of homomorphisms into the-algebraic closure and this
number is equal to the number of roots. So if some roots coincide, then the
number of homomorphisms diminishes.
Lets also recall Base-change (Theorem 4.13) . If L and E are extensions
of K and L is nite over K then
HomK (L, E) ∼
= HomE (L ⊗K E, E) .
In the formula, L ⊗K E is a nite E -algebra denoted as A below.

8
As it was shown in example 3.34 (part 3) it's not a Perfect eld and as result of
theorem 3.35 is not separable.
9 as soon as K (X) = K [T ] / (T p − Y )
Remark 5.5. 10
We have that A = L ⊗K E ∼
= E ⊗K L is a free E module
as soon as L is a free K module and with proposition 4.11 we have that
[A : E] < ∞ (as soon as [L : K] < ∞) and as result with theorem 4.23 one
can get that there are nitely many maximal ideals mi and
A∼
= A/mk11 × · · · × A/mkr r
Denition 5.6. With Chinese remainder (Theorem 4.20) we have
A∼
= A/mk11 × · · · × A/mkr r
Reduced algebra Ared is dened by the following equation
Ared = A/m1 × · · · × A/mr
11
We have that
Ared = A/η (A)
where η (A) is an Ideal of nilpotents in A.
It is clear that
HomE (A, E) = HomE (Ared , E)

because all homomorphism into a eld must be zero on all nilpotents 12 .
So again, we see that if there are nilpotents in the tensor product, then
there is somehow fewer space for homomorphisms. Because if A is not re-
duced, then the dimension
[Ared : E] < [A : E] .
content.
11 i.e. nilpotents become zeros in the A
red .
12 It requires some clarication. Consider a homomorphism φ ∈ Hom (A, E). ∀x, y ∈
E
A, φ(xy) = φ(x)φ(y). Let x ∈ η (A) i.e. x is a nilpotent then x 6= 0A , xk = 0A . We have
0E = φ(xk ) = φ(x)k
i.e. φ(x) = 0E . Therefore all nilpotents go to zero and, instead of A (as the set the φ acts
on), we can consider Ared . As result, we will get that φ(0Ared ) = 0E and all other properties
of homomorphism are also hold, for instance ∀x̄, ȳ ∈ Ared : φ(x̄ + ȳ) = φ(x̄) + φ(ȳ). Really
x̄ = x + η(A), x̄ = x + η(A) and
φ(x̄ + ȳ) = φ(x + y) = φ(x) + φ(y) = φ(x̄) + φ(ȳ)
as soon as φ(η(A)) = 0E .
So the maximal number of homomorphisms, so let's say the slogan Maximal

number of homomorphisms is attained when A is reduced and all quotients
A/mi ∼
=E (5.1)
because those quotients are of course extensions of E . 13 In general, those

quotients are extensions of E (see remark 4.22). We also have
A∼
= A/m1 × · · · × A/mr
but Hom (A/mi , E) = {0} if [A/mi : E] > 1. This is because a eld homo-
morphism is always injective. A eld homomorphism a homomorphism of
elds which are extensions of E an E -homomorphism is injective. So you
cannot map an E -vector space of dimension greater than 1 into an E -vector
space of dimension 1.
Lets take E = K̄ then automatically we will get A/mi ∼ = E because an
algebraically closed eld does not have a non trivial nite extension.
So what have we had (see also example 5.9)?
A = L ⊗K K̄,
r
Y
Ared = K̄. (5.2)
i=1
The
following
14 one A = Ared is the same to r is maximal and equal to [L : K] =
A : K̄ . In the case

r = HomK̄ A, K̄ = HomK L, K̄
13 as it was mentioned above A = L ⊗K E is a nite E -algebra i.e. A is a E -extension.
The A/m is also E -algebra (see remark 4.22) i.e. also E -extension.
From other hand we consider Hom (A/mi , E) i.e. there is a homomorphism (injection)
from A/mi → E and as result A/mi ∼ = E . These arguments are also used in the text
below for the fact explanation.
14 It is because there are [L : K] = r roots of a polynomial and all the roots are in K̄ .
Each root αi forms a polynomial X − αi which creates an ideal mi = (X − αi ). We also
have L = K (α1 , . . . , αr ) and
A = L ⊗K K̄ ∼
= K̄ ⊗K L = K̄ ⊗K K (α1 , . . . , αr ) .
It expands to (see example 4.16)

K [X]
A∼
= K̄ ⊗K ∼
=
(X − α1 ) . . . (X − αr )
∼ K̄ [X] K̄ [X] ∼
= × ··· × = K̄ × · · · × K̄.
(X − α1 ) (X − αr )
So this explains why seperability is the same thing as the absence of nilpo-
tents. So let me formulate it as a theorem.
Theorem 5.7. Let L is a nite extension over K then
1. L is separable if and only if L ⊗K K̄ is reduced. L is pure inseparable

if and only if L ⊗K K̄ is local
2. L is separable if and only if for all algebraic extension Ω, L ⊗K Ω is

reduced.L is pure inseparable if and only if for all algebraic extension
Ω, L ⊗K Ω is local.
3. If L is separable then the map
φ : L ⊗K K̄ → K̄ n
which sends
φ (l ⊗ k) = (kφ1 (l) , . . . , kφn (l))
where φi are distinct homomorphisms from L to K̄ , is an isomorphism.
Proof. 1. L separable is the same thing that the algebra A = L ⊗K K̄ has

[L : K] factors 15 K̄ which is the same as A is reduced since dimK̄ A =
[L : K]. 16
L is pure inseparable: this means that exists only one homomorphism
of L into K̄ i.e. A has only one K̄ -homomorphism into K̄ thus only
one factor and as result A is local.
17
2. If Ω is an algebraic extension then
L ⊗K Ω ,→ L ⊗K Ω̄ = L ⊗K K̄.
There is a sub-ring and so one easily checks, that a sub-ring of a reduced
algebra is reduced and same for local.
15 If we have L = K (α) (see theorem 3.29) then there exists a minimal polynomial
Pmin (α, K) of degree r = [L : K]. The polynomial splits and has r roots: α1 , . . . , αr .
Thus we have r maximal ideals m1 = (X − α1 ) , . . . , mr = (X − αr ). For each maximal
ideal we have A/mi ∼= A (see claim 1.12) thus with theorem 4.23 and (5.1) we have
r
A∼
Y
= K̄
i=1
16 Qr
From (5.2), if A = Ared , one can get dimK̄ A = dimK̄ Ared = dimK̄ i=1 K̄ = r =
[L : K].
17 i.e. we can setup an Injection and structure preserving map f : L ⊗ Ω → L ⊗ Ω̄
K K
(see denition A.133)
18
3. Leave as an excises
Remark 5.8. In general for modules M , N and P over a ring R not true
that if M ,→ N (M is a sub module of N ) then M ⊗R P ,→ N ⊗R P . But
this become the truth if R is a eld and as result M, N, P are Vector spaces.
So, for my eld extensions, I can say that if I have an extension and then I
take a base change, then it remains an extension, but you should not think
that the same thing is true for arbitrary modules over a ring.
Example
√ 5.9. 19
The example was taken from [5]. Consider extension
Q 2 over Q. Since
√
2 ∼= Q [X] / X 2 − 2

Q
tensoring with Q gives

√ √
Q ∼
2 ⊗Q Q̄ = Q̄ ⊗Q Q 2 ∼ =
= Q̄ ⊗Q Q [X] / X 2 − 2 ∼
∼

=
√ √
∼
= Q̄ [X] / X − 2 X + 2 ∼ =
√ √
∼
= Q̄ [X] / X − 2 × Q̄ [X] / X − 2 ∼ = Q̄ × Q̄
18
Accordingly [13] we have the following proof. Let l1 , l2 , . . . , ln is a K basis in L. Then
we can get the following elements of K̄ n : vi = (φ1 (li ), φ2 (li ), . . . , φn (li )). We can construct
the following matrix if we put the vi vectors as columns:
 
φ1 (l1 ) φ1 (l2 ) · · · φ1 (ln )
 φ2 (l1 ) φ2 (l2 ) · · · φ2 (ln ) 
 .. .. . .. 
 
 . . . . . 
φn (l1 ) φn (l2 ) · · · φn (ln )
Linear dependence relation between rows looks like
X
ck φk (li ) = 0. (5.3)
k=1
Pn
Thus ∀l ∈ L∗ , ∃{ki } ⊂ K : l = i=1 ki li (as soon as {li } is the K basis of L). Thus after
multiplication (5.3) by ki and summation we have
X X X
ki ck φk (li ) = ck φk (l) = 0
i=1 k=1 k=1
that should hold for any ki and as result for any l. By theorem A.98 we can conclude that
∀k : ck = 0. As result we have the set {vi } as linear independent. The size of the set is
n = dim K̄ n i.e. {vi } is the K̄ basis of K̄ n .
We used the following fact (see claim 1.12)
√

Q̄ [X] / X ± 2 ∼= Q̄
5.3 Primitive element theorem

Denition 5.10 (Idempotent). The element x is called idempotent if x · x =
x
Theorem 5.11 (Primitive element). Let L is a nite Separable extension of

K then it has only nitely many sub extensions i.e. E such that K ⊂ E ⊂ L.
Proof. So, let's base change to K̄ 20 : E ⊗K K̄ ,→ L ⊗K K̄ 21

. We also have
(see also equation (5.2))
E ⊗K K̄ ∼ = K̄ m
and
L ⊗K K̄ ∼
= K̄ n
are reduced K̄ sub-algebras generated by Idempotents namely by
ei = (0, 0, . . . , 1, . . . , 0) ,
where 1 is in i-th place 22 .

On the other hand L ⊗K K̄ ∼ = K̄ n has only nitely many Idempotents
because (a1 , . . . , ai , . . . , an ) is an idempotent if and only if all ai are 0 or 1
and therefore there are only nitely many ways to choose m idempotents out
of them, 23 so there is only nitely many ways to generate a subalgebra.
20 see proof of theorem 5.7 (second part of it).

21 See also proof of theorem 5.7.
22 this is because the typical element k ∈ K̄ n has the following form
n
X
k = (k1 , k2 , . . . , kn ) = ki ei ,
i=1
where ki ∈ K̄ .
23 we have the following equation
K̄ m ,→ K̄ n
5.4. EXAMPLES. NORMAL EXTENSIONS 75
Corollary 5.12 (Primitive element theorem). ∃α ∈ L such that L = K (α)

whenever L is nite and separable.
Proof. And this is easy to see, of course, because if L and K are innite, then
L cannot be a union, a nite union of proper subextension. A vector space
over an innite eld is not a nite union of proper subspaces. For instance a
plane is not a nite union of lines. 24
If L and K are Finite elds, then we have already described this situation
completely. We have described all nite extensions and have seen that they
are generated by one element. 25
5.4 Examples. Normal extensions

5.4.1 Examples
Example 5.13 (Primitive element).
√ √ √ √
Q 2, 3 = Q 2+ 3 .
√ √
Q 2, 3 : Q = 4 so all sub-extensions are quadratic. 26 As no

We have
√ √ √ √
quadratic polynomial has α = 2+ 3 for a root 27 , α generates Q 2, 3 .
24 It will require some additional explanations. Took α ∈ L such that P
min (α, K) has
maximal degree. If K (α) = L we complete and found the primitive element. If not then
let β ∈ L \ K (α). Consider the following element γa = α + aβ where a ∈ K . For any
a ∈ K exists K (γa ) such that K ⊂ K (γa ) ⊂ L. We have |K| = ∞ but the number
of sub-extensions is limited by theorem 5.11 therefore ∃a, b ∈ K such that a 6= b and
K (γa ) = K (γb ) = K (γ) where γ = γa .
We have γa − γb = (a − b)β ∈ K (γ), i.e. β ∈ K (γ). Therefore α = γa − aβ =
γ − aβ ∈ K (γ). We also have (as soon as β ∈ / K (α)) K ⊂ K (α) ( K (γ) ⊂ L. Thus
[K (γ) : K] > [K (α) : K]. Therefore (see proposition 1.32)
deg (Pmin (γ, K)) > deg (Pmin (α, K))
that is in contradiction with α choose.

Note that [2] has another, more known, proof for the fact and prove that if L = K (α, β)
then ∃λ ∈ K such that γ = α + λβ is the primitive element i.e. L = K (γ) = K (α, β).
25 As it was mentioned in the proof of corollary 3.16 we can take α = generator of K × .
For more info see corollary 3.16.
26 As soon as extension has degree 4 = 2 · 2 then a sub-extension should have degree 2
and the minimal polynomial should be quadratic.
27 Quadratic polynomials have very simple formula for roots with only one square (dis-
criminant) and it is not possible to get 2 squares with it
This must be a primitive element, generates our eld. It is not contained

in any proper subextension.
√ is another proof (not part of√the√lectures) that shows that β =

√ There
2+ 3 √ is a primitive element i.e.
√ 2, 3 are generated by β . Really
2
β = 5 + 2 2 3 i.e.
√ √ β2 − 5
2 3= .
2
From other side
√ √ √ √ √ √ β2 − 1
2β = 2 2+ 3 =2+ 2 3= .
2
Therefore
√ β2 − 1
2=
2β
and
√ β2 − 1
3=β− .
2β
Example 5.14 (Extension which cannot be generated by a single element).

So, take Kequal to Fp and consider K (x, y) as an extension of K (xp , y p ).
2 28
It has degree p i.e.
[K (x, y) : K (xp , y p )] = p2 .
We have ∀α ∈ K (x, y) \ K (xp , y p ) is of degree p over K (xp , y p ). This is

because αp ∈ K (xp , y p ) 29 . So, no element like these can generate our
extension.
28 K ⊂ K (xp ) ⊂ K (xp , y p ) and [K (xp ) : K] = p as well as [K (xp , y p ) : K (xp )] thus

with theorem 1.27
[K (xp , y p ) : K] = p2 .
29 α = k1 x + k2 y , where k1 , k2 ∈ K = Fp . Using remark 3.6 we can get
αp = k1p xp + k2p y p ∈ K (xp , y p ) .

5.4. EXAMPLES. NORMAL EXTENSIONS 77
5.4.2 Normal extensions

Denition 5.15 (Normal extension). A normal extension of K is a Splitting
eld of a family of polynomials 30 in K [X]. 31
Remark 5.16 (Normal extension). So, take a bunch of polynomials in K and

we adjoin all their roots to K , and this is what is called a normal extension.
For instance, a Splitting eld of one polynomial is also a normal extension.
Theorem 5.17. The following conditions are equivalent for an extension L

of K:
1. ∀x ∈ L Pmin (x, K) splits in L.
2. L is Normal extension
32
3. All Homomorphisms from L to K̄ have the same image.
30 There is a set of polynomials (can be only one polynomial in the set) and the polyno-
mials not necessary to be irreducible.
Example Q/Q - is a normal extension because there
is a set of polynomials split in it : X − 1, X 2 − 1 . Note that any irreducible polynomial
(another denition below) that has a root in it also splits, for instance X − a, where a ∈ Q
is an irreducible, has a root a ∈ Q and splits in it. From other side an arbitrary polynomial
( for example X 3 − 1 ) can have a root in Q but does not necessary split in it.
31 Another good denition of a normal extension [22] can also be used. Normal extension
E/K is such algebraic extension in which every irreducible polynomial P (X) ∈ K [X] that
has a single root in E splits in E .
32 And the image is L as soon as id is also a Homomorphism
There are 2 sta's comments about the claim:
• You can't naturally identify L with a particular subeld of K (unless L is given as
a subeld of K a priori). But from the lecture 2.5 (see theorem About extension
of homomorphism (Theorem 2.17) ) we know, that there exists an embedding
L ,→ K since L is algebraic over K . And the theorem says that images of all such
embeddings are equal if L is normal over K .
• My previous answer was a little bit misleading. I meant, if you have an arbitrary
algebraic extension of K , there is no canonical embedding into K .
If we identify roots of polynomials which dene L with some elements of K (up to
a permutation for each polynomial), then the image is L itself.
As conclusion [1] (p. 54), we can say that if L is a splitting eld of a polynomial
P (just another denition of normal extension [1]) with roots α1 , α2 , . . . , αn , then L =
K (α1 , α2 , . . . , αn ) and as soon as the homomorphism φ permutes the roots of P (if L ⊂ K̄
or if we identify roots of polynomials which dene L with some elements of K ) :
φ (L) = φ (K (α1 , α2 , . . . , αn )) =
K (φ (α1 ) , φ (α2 ) , . . . φ (αn )) = K (α1 , α2 , . . . , αn ) = L
4. The Group of Automorphisms Aut (L/K) acts transitively (see deni-

tion A.31) on this set of homomorphisms HomK L, K̄ .
Proof. 1 implies 2: Take (Pi )i∈I = {Pmin (x, K) | x ∈ L} - the set of polyno-
mials (i.e. the family of polynomials). L will be a splitting eld of the set
(Pi )i∈I and therefore (by denition 5.15) L is normal.
2 implies 3: Let S = {roots of Pi , i ∈ I in L} and S 0 = roots of Pi , i ∈ I in K̄

33
then any homomorphism φ : L → K̄ sends S to S 0 , but S generates L over
K , so φ (S) determines φ (L) 34 .
3 implies 4: Let j, j 0 ∈ HomK L, K̄ (i.e. we took 2 homomorphisms)

then they send L isomorphically to its image L0 . So, these are isomorphisms
from L to L0 . So
L0
j0 j −1
L j −1 j 0 L
take j −1 · j 0 ∈ Aut (L/K) and it sends j to j 0 35 .
4 implies 1: I have this Transitive group action and I have to prove that
any minimal polynomial splits. Consider Pmin (x, K). α1 , . . . , αn - roots in
K̄ . Then I have map K (x) → K (αi ) that extends to ji : L −−−→ K̄ . This is
x→αi
by About extension of homomorphism (Theorem 2.17) . ∃θi ∈ Aut (L/K)
such that j1 θi = ji 36 thus αi ∈ j1 (L) 37 or all roots are in j1 (L) and the
polynomial Pmin x, K̄ splits over j1 (L) but this means that it splits over L
38
33 S and S 0 are 2 sets of roots

34 As soon as φ(S) = S 0 then we have one set of roots and as result one image (that
is generated by the set) P fork homomorphisms. Or more concrete, let l ∈ L, S = {si }, S =
0
{s }. We have l =
0
lik si where lik ∈ K . For the homomorphism φ we have φ(l) =
Pi k k
lik s0 i i.e. the image φ(L) consists of elements of the following form: lik s0 i . For
P
any homomorphism we will have the same form for the image or in other words - any
homomorphism has the same image.
35 We have got that ∀j, j 0 ∈ Hom
K L, K̄ , ∃g ∈ Aut (L/K) such that g(j) = j , for
0

instance g = j j and
−1 0
g(j) ≡ jg = jj −1 j 0 = j 0 . (5.4)
Therefore group automorphisms acts transitively by the denition A.31

36 as it was mentioned at the equation (5.4)
37 θ : L → L thus j θ : L → j (L)
i 1 i 1
38 If x = α then j = id and j (L) = L
1 1 1
5.5. GALOIS EXTENSIONS 79
5.5 Galois extensions

Now we are ready to give a denition for central object of Galois theory
Denition 5.18 (Galois extension). A Galois extension is a Normal exten-

sion and Separable extension.
Theorem 5.19. Let L be a nite over K then the number of automorphisms

Aut (L/K) is less or equal to degree [L : K]:
|Aut (L/K)| ≤ [L : K] .
The equality holds if and only if L is Galois extension.
Proof. We know that the group of automorphisms

39 Aut (L/K) acts freely (see
denition A.32) on the set HomK L, K̄ , so the number of automorphisms
|Aut (L/K)| is equal to the number of Orbit of this actionwhich is less or
equal 40 to the cardinality of the set it self: HomK L, K̄ . The equality
holds whenever (if and only if) Action is Transitive group action. 41 We just
seen in theorem 5.17 that this means that L is normal over K . So we have

|Aut (L/K)| ≤ HomK L, K̄ ≤ [L : K] .
The rst inequality become equality if L is normal and the second one if L
is separable 42 , thus
|Aut (L/K)| ≤ [L : K]
and equality holds if L is both normal and separable i.e. if it's Galois exten-
sion.
Denition 5.20 (Set of invariants). If group G acts on a set X then X G =

{x ∈ X such that gx = x, ∀g ∈ G} - the set of invariants.
Remark 5.21 (on normal extensions). If L is normal over K then
39 Note that the number of distinct roots is equal to the number of distinct homo-
morphisms (see proposition 3.25). Thus we can set an association between roots and
homomorphisms. Proposition 2.2 says that the stem eld isomorphism (that translated
to the considered automorphism) is unique determined by it's value on the roots that it
exchanges. I.e.
the the action of Aut (L/K) on the set of roots (and as result on the set
HomK L, K̄ ) is exactly that one dened for Free group action.
See also rst and second remarks at 5.21.
40 in ideal case each element has an unique orbit and the number of elements and the
number of orbits are equal.
41 This follows directly from the denition of Transitive group action i.e. ∀h , h ∈
1 2
HomK L, K̄ ∃a ∈ Aut (L/K) such that a(h1 ) = h2 .

42 see denitions 3.26 and 3.27.
1. If we have an Isomorphism of sub-extensions (K ⊂ L1 , L2 ⊂ L) φ :

L1 ∼= L2 then it extends to an Automorphism of L. To see this, we
embed L into an algebraic closure K̄ . And remark that φ extends to
a map from L1 to K̄ , 43 but all those maps have the same image (see
theorem 5.17), namely L 44 .
2. The group of automorphisms Aut (L/K) acts transitively on the roots

of any irreducible polynomial P ∈ K [X]. Again, an isomorphism of
stem elds extends to an automorphism of L 45 .
3. If the group Aut (L/K) xes (see denition A.28) some element x ∈ /K
then x is pure inseparable (see denitions 3.23 and 3.24) i.e. Pmin (x, K)
has a single root x. Thus if L is Galois extension (i.e. is separable)
then the set of elements which are xed by the automorphisms of L
over K is just K itself: LAut(L/K) = K (see denition 5.20). 46
Denition 5.22 (Galois group). If L is Galois extension then Galois group
G = Gal (L/K) is the group of automorphisms Aut (L/K).
Thus we can write
LGal(L/K) = K. (5.5)
5.6 Artin's theorem

Motivated by (5.5) let formulate and proof the important theorem
Theorem 5.23 (Artin). L is a eld and G ⊂ Aut (L)
1. If G acts with nite orbits, so, I mean all orbits of G are nite (i.e.
G
∀x ∈ L : |Orb(x)| < ∞), then L is a Galois extension of L .
43 Ekaterina really said that but may be more better to say that φ extends to a map
from L to K̄ ???
44 Consider L ⊂ L ⊂ L ⊂ K̄ we have a homomorphism φ : L → L (our isomor-
1 2 1 1 2
phism) that accordingly theorem 2.17, can be extended to homomorphism L1 → K̄ and
the last one can be extended to φ˜1 : L → K̄ . We can also consider L2 ⊂ L1 ⊂ L ⊂ K̄ and
homomorphism φ2 : L2 → L1 that can extended (using the same approach) to φ˜2 : L → K̄ .
Both φ˜1,2 has the same image L as soon as L - normal extension (see theorem 5.17). Thus
we have an Automorphism.
45 In the case we have Transitive group action i.e. for any roots x, y we have an isomor-
phism of stem elds φ such that φ(x) = y (see proposition 2.2). This isomorphism extends
to an automorphism accordingly the prev. remark.
46 Another explanation is the following. Let x ∈ L \ K then g ∈ Aut (L/K) such that
g 6= id permutes roots of irreducible Pmin (x, K) and g(x) = x0 6= x. The only thing that
works for ∀g ∈ Aut (L/K) is K i.e. ∀k ∈ K and ∀g ∈ Aut (L/K) we have g(k) = k . This
is because Pmin (x, K) = X − k i.e. only one root possible.
5.6. ARTIN'S THEOREM 81

2. If |G| = n < ∞ then L : LG = n and G is a Galois group
Remark 5.24. Well notice, that acting with nite orbits and being nite
is not the same thing. So, a short remark before giving a proof: notice
that nite orbits does not mean niteness because it's typical for Galois
groups to act with nite orbits. If we have some G, which is Galois of L
over K : G = Gal (L/K), and x ∈ L, then x is a root of a polynomial of
some nite degree and it's splitting eld is nite over K , so, the orbit of x
is also nite because it's always sent to another root of the same polynomial
and so consists of roots of the Pmin (x, K). But of course the Galois group
itself Gal (L/K) can be innite when L is not nite over K . For instance, if
K = Fp and L = Fp . It is very easy to compute all the Galois groups, and
in fact we shall see shortly what is exactly this Galois group of L over K is
innite. 47
Proof. 1. Let me take x, well say, x1 ∈ L which is not G-invariant:

x1 ∈ L \ LG and Q G-Orbit of x Orb(x) = {x1 , x2 , . . . , xk }. The poly-
nomial P (X) = ki=1 (X − xi ) is G-invariant 48 . G just permutes
the xi , it permutes the factors of these polynomial, so the polynomial
is G-invariant. Therefore its coecients are G-invariant and as result
P ∈ LG [X] by denition 5.20. LG is a eld of G invariants, and it is
separable. P is separable, because all xi are distinct ( there are distinct
elements of the orbit ). And L is splitting eld of P , therefore L is a
Galois extension over LG by the Galois extension denition.
2. We have |G| = n then L : |Orb (y)| ≤ n. Take x as

∀y ∈ 49 above (x ∈
L \ L ) L (x) : L ≤ n.
G G G
Claim that this implies L : L ≤ n.
G
If I knew already, that L is nite over LG , this would be very easy,

this would be just a direct consequence of Primitive element (Theo-
rem 5.11) . I would say that L is generated by one element. I take this
one element as my x and I see that L is of degree at most n over LG .
But I don't know yet that L is nite so I have to do some trick. So,
proof of the claim: take x such that LG (x) : LG is maximal then take
y ∈ L. LG (x, y) is nite over L 50 and I can apply Primitive element
47 Section 6.4.2 shows that Gal F̄p /Fp is not cyclic, but theorem 3.15 says that if G is

nite than it should be cyclic. Thus we can conclude that Gal F̄p /Fp is not nite
48 I.e. ∀g ∈ G : g (P (X)) = P (X)
49 x is a root of the following polynomial P (X) = Qk (X − x ), where x ∈ Orb(x).
i=1 i i
The polynomial has degree and G

deg P = |Orb(x)| ≤ n deg P min x, L ≤ deg P ≤ n
therefore L (x) : L ≤ n.
G G

50 because x, y are algebraic elements and using theorem 1.29 and proposition 1.21 we
can conclude that LG (x, y) is nite over L.
(Theorem 5.11) . Therefore LG (x, y) = LG (z). But

G
L (x) : LG ≥ LG (z) : LG

thus LG (x) = LG (z) so y ∈ LG (x) and since I can do this for any y ,
I eventually conclude that L = LG (x), and in particular, L : LG ≤
n. Well, now if this is strictly less than n , then L cannot have n
automorphisms over L G
but G ⊂ Aut L/L G
so this is a contradiction.
Therefore L : L = n and G = Aut L/L (see theorem 5.19).
G
G
Chapter 6
Galois correspondence and rst
examples
We state and prove the main theorem of these lectures: the Galois corre-
spondence. Then we start doing examples (low degree, discriminant, nite
elds, roots of unity).
6.1 Some further remarks on normal extension.

Fixed eld
Some denitions from previous lecture. L over K is Galois extension if and
only if it is a Separable extension and Normal extension or in other words L is
a Splitting eld of a family of separable irreducible polynomials over K . We
also seen (see theorem 5.19) that in the case of nite extension [L : K] < ∞
the number of automorphisms |Aut (L/K)| = [L : K].
There are several remarks on Normal extensions which show that the
extensions behave sometimes dierently compare to other types of extensions.
Especially we have seen for that an extension L over M over K was nite
or algebraic or separable or purely inseparable if, and only if, it was true for
L over M and M over K . So, for a normal extensions, this is not the case
anymore.
Remark 6.1. Let we have a tower of extensions K ⊂ L ⊂ M . If M is
normal over K then of course the M is normal over L. It is clear because
if M is a splitting eld of a family of polynomials over K the one can just
consider them as being polynomials over L and say that M is a splitting eld
of a family of polynomials over L.
But L does not have to be normal over K (see example 6.2).
83
84CHAPTER 6. GALOIS CORRESPONDENCE AND FIRST EXAMPLES
Example 6.2. Consider

√ √
4 4
Q⊂Q 2 ⊂Q 2, i
√ √
We have Q 4
2, i to be a splitting eld for polynomial X 4 −2 but Q 4 2
is just a Stem eld (not Splitting eld) for this polynomial. And as result
√
4

Q 2 is not a normal over Q.
Remark 6.3. A quadratic extension 1 is normal. This is by formula for roots

2
of a quadratic equation.
If P quadratic over K has 1 root in L then its another root is also in L.
Remark 6.4. One often has K ⊂ L ⊂ M with L normal over K , M normal
over L but M not normal over K (see example 6.5).
Example 6.5. Consider

√ √
4
Q⊂Q 2 ⊂Q 2
√ √ √
We have Q 2 normal over Q Q 4 2 normal over Q 2
as well as
√
4

because they both are quadratic extensions. But Q 2 is not normal over
Q (as it was mentioned in example 6.2)
We also seen at last lecture the following denition (see denition 5.20):
Denition 6.6 (Fixed eld). If L is a eld and G ⊂ Aut (L) then
LG = {x ∈ L | ∀g ∈ G : gx = x}
is a xed eld
If we have a sub-eld K ⊂ L then we can consider the following group of

automorphisms of L over K : Aut (L/K) in the case if L is normal because
otherwise the group will be too small to give information about L. But in
the normal case it makes sense to consider the group of automorphisms of L
over K .
We have seen (see (5.5)) that if L is separable over K then
LAut(L/K) = K
1Extensions with degree 2.
2Because a if we have 2 roots x1 , x2 ∈ L ⊃ K then there exists the following relation
x2 = k1 + k2 x1 where k1 , k2 ∈ K . I.e. if we know one root then the other is easy computed
and located in the same extension as the rst one.
6.2. THE GALOIS CORRESPONDENCE 85
This is because of the group of automorphisms was permuting the roots over
the minimal polynomial of x over K (see item 3 in remark 5.21).
We also have seen (see
theorem 5.23) that if G is nite the L is Galois
extension over L and L : L = |G|.
G G

And now we are going to summarize all these in a theorem which is in

fact the main subject of this lecture course and this theorem is called the
Galois correspondence.
6.2 The Galois correspondence

Let L over K be a Galois extension. By denition the group automorphisms
Aut (L/K) is called Galois group and denoted as Gal (L/K)
Theorem 6.7 (Galois correspondence). There are several statements:
1. If L is nite over K then there is a Bijection between a sub-extension

F (K ⊂ F ⊂ L) and a subgroup H ⊂ Gal (L/K). The correspondence
is the following
F → Gal (L/F )
LH ← H
2. The following statement are equivalent (if and only if )
(a) F is Galois over K

(b) ∀g ∈ Gal (L/K) g (F ) = F
(c) Gal (L/F ) is a Normal subgroup in Gal (L/K)
g goes to g restricted to F : g → g|F this

In this case is a Surjection
3
Gal (L/K) Gal (F/K) and the kernel is Gal (L/F ).
Proof. 1. Most work have been done before. What have we got by now?
4
LGal(L/F ) = F (see (5.5)). We also have H ⊂ Gal L/L H
. Artin
(Theorem 5.23) gives us L :
LH = |H| but with theorem 5.19 we also

have L : LH = Gal L/LH so one must have H = Gal L/LH as

3 The fact is explained at the end of the proof of the theorem and is used in the claim
8.16 validation.
4 Ekaterina called the fact obvious but let try to explain it. From (5.5) we have
Gal(L/LH )
thus it should also satisfy therefore H ⊂
H H

L = L ∀h ∈ H h ∈ Gal L/L
Gal L/L . H

soon as the subset H has the same cardinality as the set Gal L/LH

itself.
This means that the maps that we have in the theorem : F → Gal (L/F )
and LH ← H are mutually inverse 5 and if a map is invertible then there
is a Bijection.
2. We should proof equivalence of the following statements:
(a) F is Galois over K

6
(b) ∀g ∈ Gal (L/K) , g (F ) = F
(c) Gal (L/F ) / Gal (L/K)
Lets show that 2a implies 2b. Fix x ∈ F then the minimal polynomial
Pmin (x, K) splits in L but it has a root in F , thus it should have all roots in F
by normality. I. e. as soon as F is a Normal extension, Pmin (x, K) splits in F
(see theorem 5.17). This means, of course, that any map from Galois group
preserves F since it premutes the roots: ∀g ∈ Gal (L/K) g permutes the
roots of Pmin (x, K) and that is the true for any x ∈ F therefore g (F ) ⊂ F
since F is generated (consists of) such roots. 7
Lets show that 2b implies 2a. If g (F ) ⊂ F then all roots of Pmin (x, K),
x ∈ F , are in F since g permutes those roots or, in other words, since Galois
group acts transitively on roots of an irreducible polynomial 8 therefore F is
normal by denition. 9
Lets show that 2a and 2b are equivalent to 2c i.e. let g ∈ G, g (F ) ⊂ L
then if h ∈ Gal (L/F ) is such that h|F = id then ghg −1 |g(F ) = id. 10 This
5 We have the following maps:
F → Gal (L/F ) → LGal(L/F ) = F
I.e. ∀F such that K ⊂ F ⊂ L we can construct H = Gal (L/F ) then we can construct
LH such that (by theorem 5.23) K ⊂ LH ⊂ L.
6 g ∈ Gal (L/K) operates on L ⊃ F .
7 We have the following fact: x ∈ F then also g(x) ∈ F i.e. g(F ) ⊂ F . For equality
g (F ) = F we have to proof that ∀g ∈ Gal (L/K) : F ⊂ g (F ). Really let g ∈ Gal (L/K)
then g −1 ∈ Gal (L/K) and g −1 (F ) ⊂ F therefore F = id(F ) = g g −1 (F ) ⊂ g(F ) i.e.
F ⊂ g(F ).
8 see also 2d remark 5.21 that says that if F is normal then Gal(L/F ) acts transitively
on the roots of any irreducible polynomial P ∈ F [X].
9 More clear explanation is below. We have to prove that F is normal and separable.
Normality: ∀Pmin (x, K) splits in F thus the theorem 5.17 gives us the required normality.
About separability. We have K ⊂ F ⊂ L. L is separable over K as soon as L is Galois
extension. Thus theorem 3.29 gives us F separability.
10 The map g −1 acts as follows g −1 : g(F ) → F i.e. h in ghg −1 operates on F only

g(F )
6.3. FIRST EXAMPLES (POLYNOMIALS OF DEGREE 2 AND 3) 87
means that ghg −1 ∈ Gal (L/g (F )) (see (5.5)) so the statement g (F ) = F is

the same to say
gGal (L/F ) g −1 = Gal (L/F )
So apply this to all g ∈ Gal (L/K) one can get that Gal (L/F ) is a Normal
subgroup of Gal (L/K) by the denition of Normal subgroup.
Finally if all this statements (2a ⇐⇒ 2b ⇐⇒ 2c ) are true then we can
consider map (make sense by 2b)
φ : Gal (L/K) −−−−→ Gal (L/F ) .
g→ g|F
This is a Surjection by theorem 2.17 11 and the kernel ker φ = Gal (L/F ) by
denition because the kernel consists of things which are identity on F .
Remark 6.8. If L over K is not nite then Galois correspondence is not
Bijection i.e. the maps which are in the theorem still make sense, but they
will not be mutually inverse bijections and we shall see an example (see
section 6.4.2 about it).
6.3 First examples (polynomials of degree 2 and

3)
Example 6.9 (Degree 2). Let [L : K] = 2 and charK 6= 2. The extension L
is generated by a root of quadratic polynomial i.e. x ∈ L \ K then Pmin (x, K)
is quadratic and if we look at the formula for the root of the equation we
will
see that the extension is generated by a root of discriminant ∆: L =
√
K ∆ .
What can we say about the Gal (L/K). It consists of 2 elements and there
12
is only one cyclic group of 2 elements : Z/2Z. Therefore
Gal (L/K) ∼
= Z/2Z.
The elements of the group is identity and an element that exchanges the 2
√ √
roots i.e. permutes ∆ and − ∆.
therefore
ghg −1 g(F ) = ghF g −1 g(F ) = gg −1 g(F ) = id

11 We have that ∀gF ∈ Gal (L/F ) gF is a homomorphism that operates as follows

gF : F → K̄ (note that by the theorem 5.17 all homomorphisms on a normal extension
F have the same image namely F ) and therefore it can be extended (by theorem 2.17) to
homomorphism g : L → K̄ i.e. ∀gF ∈ Gal (L/F ) , ∃g ∈ Gal (L/K) such that φ(g) = gF .
Or in other words φ is a Surjection.
12 There is only one group of 2 elements [64]
Example 6.10 (Degree 3). Let [L : K] = 3 and charK 6= 3 (separable

extensions) then L is generated by x - root of degree 3 polynomial P and
there are 2 cases
1. P splits in L therefore L is a Galois extension and Gal (L/K) is the

13
Galois group but the Galois group of 3 elements must be cyclic i.e.
Gal (L/K) ∼
= Z/3Z - cyclic group of order 3.
2. P does not split in L then there exists M = K (x1 , x2 , x3 ) - splitting eld

where x1,2,3 are roots of P and L = K (x1 ). M is Galois extension and
the Galois group is embeded into a group of permutation of 3 elements
(because Galois group permutes the roots): Gal (M/K) ,→ S3 .
As soon L ( M then [M : K] > 3 so Gal (M/K) = S3 . In particular
[M : K] = |S3 | = 6
If you see a cubic polynomial how will you decide is its Galois group is
cyclic or S3 ? This is determined by a discriminant of polynomial which is a
subject of next section (see example 6.13 and proposition 6.12).
6.4 Discriminant. Degree 3 (cont'd). Finite

elds
6.4.1 Discriminant
Denition 6.11 (discriminant). Let P ∈ K [X]. The polynomial has the
following roots in K̄ : x1 , x2 , . . . , xn . The following product is called discrim-
inant: Y
∆= (xi − xj )2
i<j
If we take group G = Gal (P ) then G ⊂ Sn and any permutation preserves

∆ and as result we have ∆ ∈ K (see (5.5)).
Lets take a root of discriminant (we have to choose some roots order for
this operation) then
√ Y
∆= (xi − xj )
i<j
13 There is only one group of 3 elements and it is cyclic [64].

6.4. DISCRIMINANT. DEGREE 3 (CONT'D). FINITE FIELDS 89
14
this quantity is preserved only by even (and not by odd) permutations.
Proposition
√ 6.12. Let G = Gal (P ) - Galois group then G ⊂ An 15
if and
only if ∆ ∈ K.
Proof. Since if the Galois group is even then, this will be preserved by an
element of Galois group and so will be in K and conversely, if it is an element
of K , then it must be preserved by the Galois group, but we know it is
preserved only by even permutations.
If we return to our example 6.10 we can get the following one
Example 6.13 (Discriminant of polynomial degree 3). Lets compute the

discriminant for the following polynomial: X 3 + pX + q . 16
17
The discriminant easy to compute: ∆ = −4p3 − 27q 2 . So if ∆ is a
square in = A3 (cyclic of order 3) 18 . If not then Gal (P ) ∼
K then Gal (P ) ∼ =
S3 (non commutative group of 6 elements).
14 You can see from denition A.50 and example A.51 that for even permutations each
insertion is equivalent to a root exchange. If the number of such exchange is even then we
can return the changed root of a discriminant to its original form with event steps. Each
step changes the sign and as result the even steps will not change the sign.
15 A is a group of even permutations
n
16 X 2 element can be always hidden via a variable change. Thus the polynomial can be
considered as a common case for cubic polynomials.
17 The below explanation was taken from [67]. The discriminant is a polynomial of
degree 6. p can be represented as a polynomial of degree 2. q = −X 3 − pX is a polynomial
of degree 3. The discriminant therefore can be represented as follows ∆ = ap3 + bq 2 where
a, b are numbers. Let p = −1, q = 0 then the polynomial X 3 − X has 3 roots: −1, 0, 1. It
is easy to compute discriminant ∆ = (0 + 1)2 (0 − 1)2 (1 + 1)2 = 4 therefore a = −4. For
the case p = 0, q = −1 we will get the polynomial X 3 − 1 with roots 1, ζ, ζ 2 . We also have
ζ2 + ζ + 1 = 0
therefore
2
(ζ − 1) = ζ 2 − 2ζ + 1 = −3ζ
and
2
(ζ + 1) = ζ 2 + 2ζ + 1 = ζ
thus
2 2 2
∆ = (ζ − 1) ζ2 − 1 ζ2 − ζ =
2 2 2 2 2
= (ζ − 1) (ζ − 1) (ζ + 1) ζ (ζ − 1) =
3
= (−3ζ) ζ 3 = −27.
As result b = −27.
18 See example A.56 about groups S and Alternating group A .
3 3
What can we say about sub-extensions for the two cases? Let M is a
splitting eld of P over K then for the rst case there is no any sub-extension.
For the second case there are several sub-extensions (they are determined by
sub-groups of the Galois group: S3 for our case). Especially we have 3 sub-
19
extension of degree 3: K (x1 ), K (x2 ) and K (x3 ), xed by non-normal
20
sub-groups of order 2 - because M is degree 2 over K (x1,2,3 ). And we have
√
one quadratic sub-extension (of degree 2) xed by A3 ⊂ S3 this is K ∆ .
21
Galois correspondence (Theorem 6.7) says us that there are no other

sub-extensions. Because those sub extensions correspond objectively to sub-
groups of the Galois group. And in this case, it does not have so many sub-
groups. These are just three subgroups of order 2 generated by transpositions,
and one subgroup of order 3 generated by a three cycle.
6.4.2 Finite elds. An innite degree example

We have seen that theory of nite elds is easy. Especially all Galois groups
are cyclic (see corollary 3.17). I.e. we have the eld Fpn over Fp . The Galois
group is cyclic and generated by Frobenius map (see remark 3.6) which is
F p : x → xp .
More interesting are innite extensions of a nite eld, for instance the
Algebraic closure. Thus consider F̄p as an extension of Fp . If we take an
invariant generated by Frobenius Fp 22 then 23
F̄hF
p
pi
= Fp
but
Gal F̄p /Fp =6 hFp i
19
because x1 , x2 , x3 are roots of a cubic polynomial
20
K (xi ) = M Gal(M/K(xi )) and Gal (M/K (xi )) has order 2, as soon as [M : K (xi )] = 2,
i.e. Gal (M/K (xi )) ∼ = Z/2Z - transposition of roots
21 The extension should be K (α) where α is a root of a quadratic polynomial and
√ √
α∈ /√K but α ∈ M . √ ∆ is a good choice for the sub-extension generator because ∆ ∈ /K
but ∆ ∈ M and ∆ - is a root of a quadratic polynomial i.e. it generates a quadratic
√ √ √
extension. In the case K ∆ = M Gal(M/K ( ∆)) = M A3 i.e. Gal M/K ∆ = A3 .
22 The group generated by one element F is denoted as hF i.
p p
23 It required some explanation. hFp i consists of powers of Frobenius map:
F n
Fp , Fp2 , . . . , Fpn , . . . , we also have that F̄p p = Fpn (This is because ∀x ∈ Fpn : Fpn (x) =
n
xp = x). Therefore
F̄hF
p
pi
= ∩n Fpn = Fp
6.4. DISCRIMINANT. DEGREE 3 (CONT'D). FINITE FIELDS 91
therefore there is no bijective correspondence between sub-elds and sub-

groups. In particular the Galois correspondence (Theorem 6.7) is not
Bijection (as it was mentioned at remark 6.8)
So how to see that the Galois group is not cyclic: Gal F̄p /Fp 6= hFp i?

Really a smaller group is not cyclic. 24 Lets look at the following:

F p ⊂ F p 2 ⊂ . . . F p 2n ⊂ . . .
and let
L = ∪Fp2n
We claim that Gal (L/Fp ) is not cyclic. Consider the following number an =
1 + 2 + 4 + · · · + 2n then ∀x ∈ Fp2n
Fpan (x) = Fpam (x) (6.1)
for any m > n. 25 This is because the Frobenius map Fp sends x to xp is
is identity on Fp2n ∀l ≥ 0. 26 This implies
n+l
an identity on Fp therefore Fp2
that there exists an automorphism φ : L → L such that ∀n ≥ 0 27
φ|F n
= F an (6.2)
p2
6 φ. 28 One can look at φ as φ = Fp1+2+4+···+2 +... but this is,

n
but ∀k ∈ Z Fpk =
of course, very informal. 29 The rigorous conclusions we can draw from this
24 We have a theorem that a subgroup of a cyclic group is cyclic (see Fundamental
theorem of cyclic groups (Theorem A.24) ). This means that if the whole group is cyclic
then the smaller group has to be a cyclic too.
25 If F 2n+l = id (see below in the text) then
p
n+1 n+2 m
Fpam (x) = Fpan +2 +2 +···+2 (x) =
n+1 n+2 m

= Fpan Fp2 +2 +···+2 (x) = Fpan (id (x)) = Fpan (x) .
26 n+1
For example consider Fp2 and let x ∈ x ∈ Fp2n . We have
n+1 n n n
Fp2 (x) = Fp2 ·2 (x) = Fp2 +2 (x) =
n
n
n
n n n
= Fp2 Fp2 (x) = Fp2 x2 = Fp2 (x) = x2 = x.
27 ??? We can say that φ maps L to L. If equation (6.2) holds for an arbitrary large N
(i.e. informally Fp2N very close to L) then it also holds ∀n < N because (6.1)
28 i.e. φ ∈
/ hFp i
29 Using (6.2) informally one can get
n
φ = φ|L = lim φ|F n
= lim F an = F 1+2+4+···+2 +...
n→∞ p2 n→∞
is that our Galois group

is not a cyclical group generated by the Frobenius
map i.e. Gal F̄p /Fp 6= hFp i. And also, that we don't have a bijective
Galois correspondents like we have for nite eld extensions i.e no bijective
correspondents between sub-groups of the Galois group and sub-extensions.
Indeed the xed eld of the Fp and and the whole Galois group coincide.
6.5 Roots of unity: cyclotomic polynomials

Consider a number n that is prime to Characteristic of K : (n, char (K)) = 1
and consider the polynomial Pn = X n − 1 (if (n, char (K)) = 1 then the
polynomial has no multiple roots) 30 . Thus the polynomial has exactly n
roots which form a cyclic (see denitionA.23) multiplicative subgroup of K̄ ×
(see denition A.67) 31 denoted by µn . So µn is just the group of n roots of
unity in K̄ × .
Denition 6.14 (Primitive roots of unity). There are root of unity of degree
n such that not root of unity of degree d < n.
The set of Primitive roots of unity is denoted as µ∗n . All elements of µn
are powers of a single one: ∀x ∈ µn , ∃a ∈ N : x = ζ a for some ζ ∈ µn . And
primitive roots of unity form the following set {ζ a } where (a, n) = 1. 32
The number of such primitive roots is determined by Euler's totient function:
|µ∗n | = φ (n). 33
Denition 6.15 (n-th cyclotomic polynomial). The polynomial
Y
Φn = (X − α) ∈ K̄ [X]
α∈µ∗n
is called n-th cyclotomic polynomial.

Example 6.16 (n-th cyclotomic polynomial).
Φ1 = X − 1
30 if (n, char (K)) = 1 then P 0 = nX n−1 6≡ 0 mod char (K) and as result
n
gcd (Pn , Pn0 ) = 1 i.e. Pn does not have multiple roots.
31 ∃x ∈ µ such that xn = 1 i.e. x is a root of P .
n n
32 The fact require a proof. Consider case (a, n) = k > 1. Thus n , a ∈ Z and we can
k k
get n a n
(ζ a ) k = ζ k = 1

n
therefore ζ a is a root of X k − 1 and as soon as nk < n we conclude that ζ a ∈ / µ∗n (but
a
ζ ∈ / µn ) i.e. ζ is not a Primitive roots of unity.
a
33 That is by denition of Euler's totient function that counts the positive integers up
to a given integer n that are relatively prime to n
6.5. ROOTS OF UNITY: CYCLOTOMIC POLYNOMIALS 93
X2 − 1
Φ2 = =X +1
X −1
X3 − 1
Φ3 = = X2 + X + 1
X −1
X4 − 1
Φ4 = = X2 + 1
(X − 1) (X + 1)
Φ5 = X 4 + X 3 + X 2 + 1
Example 6.17 (Cyclotomic polynomial). 34 There are several ways to get a

n-th cyclotomic polynomial in GAP [8]. Consider 7th cyclotomic polynomial.
We can use the fact that Φ7 = Pmin (ζ7 , Q):
gap> zeta:=E(7);;
gap> MinimalPolynomial(Rationals, zeta);
x^6+x^5+x^4+x^3+x^2+x+1
gap>
I.e. Φ7 = Pmin (ζ7 , Q) = X 6 + X 5 + X 4 + X 3 + X 2 + X + 1.

Another way to get the same answer is
gap> CyclotomicPolynomial(Rationals,7);
x^6+x^5+x^4+x^3+x^2+x+1
Proposition 6.18.
Q
1. Pn = d|n Φd
2. Φn has coecients in Prime elds: Q if charK = 0 or Fp if charK = p
3. If charK = 0 then Φn ∈ Z [X]. If charK = p then Φn is the reduction

mod p of the n-th cyclotomic polynomial over Z.
Proof. The proof was marked as an exercise in the lectures

The rst item proof is the following. The X n − 1 has n roots and lets ζ is
one of the root. Let d = ord (ζ) i.e. ζ d = 1 from other side ζ n = 1 i.e. d | n.
ThereforeQ ζ ∈ µd i.e. ζ is a root of Φd . Thus any
∗
root X n − 1 will also be a
root of d|n Φd . From other side for any root ζ 0 of d|n Φd exists d such that
Q
ζ 0 will be a root of Φd and as soon as d | n then ∃l ∈ Z : n = d · l therefore

l
(ζ 0 )n = (ζ 0 )d = 1 i.e. ζ 0 is also a root of X n − 1. The both polynomial are
Monic polynomials and therefore are the same.
Lets consider case char(K) = 0 (for the second and third parts) and lets
proof that in the case Φd has coecients in Z. Lets proof by induction. The
case n is trivial because Φ1 = X − 1. Let for any d < n we have that Φd has
integer coecients. Then using
Y
Pn = Φd
d|n
we can get
Xn − 1
Φn = Q
d|n,d6=n Φd
using the fact that d|n,d6=n Φd has integer coecients (by induction hypothe-
Q
sis) we can conclude that Φn has coecients in Q. Then using Gauss (Equa-
tion 1.34) we can conclude that Φn ∈ Z [X].
Finally using
Y
Pn = X n − 1 = P n mod p = (Φd mod p)
d|n
one can conclude that if charK = p then Φn is the reduction mod p of the
n-th cyclotomic polynomial over Z.
6.6 Irreducibility of cyclotomic polynomial.The

Galois group
Theorem 6.19. Let char(K) = 0, then Φn is irreducible in Q [X] (it amounts
to the same to say that this is irreducible in Z [X] as we know).
Proof. We have to prove that all Primitive roots of unity have the same
minimal polynomial over Q. It must be Φn by degree reason.
Let x one primitive root ζ and all others have the form ζ a where a is
prime to n: (a, n) = 1. 35 If we can show these for a prime then we can also
35 as it was mentioned at the section 6.5.
6.6. IRREDUCIBILITY OF CYCLOTOMIC POLYNOMIAL.THE GALOIS GROUP95
36
show this for all a. Thus we may assume that a is a prime number l and
suppose
Pmin (ζ, Q) 6= Pmin ζ l , Q .

Then Φn = f · g where f has ζ as a root and g has ζ l as aroot. This is true

37
in Q [X] but also as we seen in Z [X]. So we have g ζ = 0 thus
l
we can
dene gl (X) = g X then gl will have ζ as a root. But gl ≡ g mod l. 38
l l
Thus in modulo l Φn has ζ as a multiple root. This is impossible because Φn

divides Pn and this does not have multiple roots whenever l prime to n.
Remark 6.20. Statements of theorem 6.19 are not true if char(K) > 0. I.e.
over Fp Φn is not always irreducible.
For instance Φ8 = X 4 + 1 is reducible over Fp for any p. For instance
p = 2 39 In fact it splits in Fp2 . This is because if p is odd then 8 | p2 − 1 so
the Multiplicative group F× p2 contains a cyclic subgroup of order 8 which is
exactly the group of 8 roots of unity.
The main theorem about cyclotomic extensions is the following
Theorem 6.21. The splitting eld L of Pn over K is K (ζ), where ζ is a

root of Φn .
∀g ∈ Gal (L/K) acts by g : ζ → ζ ag where (ag , n) = 1.
Gal (L/K) ,→ (Z/nZ)× and this is an isomorphism whenever Φn is irre-
ducible over K (for example K = Q).
Proof. 1. All n-th roots of unity are powers of ζ so they are in K (ζ).
2. Thus any g ∈ Gal (L/K) induces an automorphism of µn ⊂ L and all

such automorphisms are raising a root to a power that is prime to n.
36 Let a = l · k where l and k are prime. We have (as it will be proved later) that
ν = ζ l and ζ are roots of the same minimal polynomial. ν k will also be a root of that
minimal polynomial (if ν is a root and k is prime than ν k will also be a root of the same
polynomial), but ν k = ζ k·l = ζ a .
37 Φ ∈ Z [X] (see proposition 6.18). Gauss (Equation 1.34) says that if Φ is reducible
n n
than they factors are also in Z [X].
38 Using g (X) = a X n + · · · + a X + a we can get
n 1 0
l
g l (X) = (an X n + · · · + a1 X + a0 ) =
ln
= an X + · · · + a1 X l + a0 + l · (. . . ) ≡ gl mod l
39 1 is a root because 14 + 1 = 2 ≡ 0 mod 2. Therefore Φ8 = X 4 + 1 is reducible over

F2
40
3. Gal (L/K) ,→ Aut (µn ) ∼ = (Z/nZ) . That 41 is because if a g is an

×
identity on µn then it is also identity on L because µn generates L over

K 42
4. If Φn is irreducible then there is an isomorphism because of cardinality:

[L : K] = deg Φn = φ (n). But φ (n) = (Z/nZ)× . So in the case the

embedding must be isomorphism.
40 This is because ∀g ∈ Gal (L/K) g permutes roots of the irreducible polynomial Φn .

The roots a generated by means of ζ → ζ ag , where (ag , n) = 1. Therefore we can associate
a root permutation (i.e. g ) with the following map ζ → ζ ag , where (ag , n) = 1.
41 Gal (L/K) ,→ Aut (µ )
n
42 We have to proof there that Gal (L/K) ,→ Aut (µ ) is an Embedding i.e. Injection
n
that preserves the group structure. The injection was proved before because for g1 , g2 ∈
Gal (L/K) such that g1 6= g2 we have two dierent root exchange ζ → ζ ag1 and ζ → ζ ag2 .
About structure preserving if f : g → (ζ → ζa ) then easy check that f (g1 g2 ) = f (g1 )f ( g2 )
and that there is a homomorphism. By the way any embedding should preserve the identity
element that was shown by Ekaterina.
Comments from sta about the identity preserving: The proof of injectivity in the
lecture goes as follows: take arbitrary element g ∈ Gal(L/K). We prove, that if the image
of g is the identity in (Z/n)∗, then g is itself the identity in Gal(L/K).
Chapter 7
Galois correspondence and rst
examples. Examples continued
We continue to study the examples: cyclotomic extensions (roots of unity),
cyclic extensions (Kummer and Artin-Schreier extensions). We introduce
the notion of the composite extension and make remarks on its Galois group
(when it is Galois), in the case when the composed extensions are in some
sense independent and one or both of them is Galois. The notion of inde-
pendence is also given a precise sense ("linearly disjoint extensions").
7.1 Cyclotomic extensions (cont'd). Examples

over Q
Last time we discussed cyclotomic extensions which are splitting elds of Φn
(generated by n-th roots (Primitive roots of unity) of 1). And we got a very
precise description of those extensions in the case when Φn was irreducible,
for instance, over Q.
We have seen (see theorem 6.21) that Q (ζn ) is a Galois extension of Galois
2πi
group (Z/nZ)× (see example A.68) where ζn = e n . So it acts as ga : ζn → ζna
where a ∈ (Z/nZ)× that can be considered as a number relatively prime to
n: (a, n) = 1. 1
Lets consider several examples
Example 7.1 (n = 8). Lets considern = 8. In the case
(Z/8Z)× = 4

1 ×
(Z/nZ) consists of elements which are invertible in Z/nZ. The numbers which are
prime to n are invertible.
97
98CHAPTER 7. GALOIS CORRESPONDENCE AND FIRST EXAMPLES. EXAMPLES C
i.e. the group has 4 elements there are
(Z/8Z)× = {1, 3, 5, 7} .
2
So our Galois group also has 4 elements :
Gal : id, ζ8 → ζ83 , ζ8 → ζ85 , ζ8 → ζ87 = {id, σ3 , σ5 , σ7 } .

We can note that σ7 = ζ8 → ζ87 is something very simple - it is complex

σ7
conjugation: σ7 = ζ8 → ζ̄8 . It's Fixed eld Q (ζ8 ) is determined by the
3
following expression
√
Q (ζ8 )σ7 = Q (ζ8 ) ∩ R = Q ζ8 + ζ̄8 = Q

2
4
i.e. there is a quadratic extension.
Our Galois group has 3 subgroups of order 2 so we have 3 quadratic sub-
√
extensions. One o them we have already found (Q 2 ) lets nd 2 others.
5
√
Q (ζ8 )σ3 = Q ζ8 + ζ83 = Q i 2 .

2 There is a well known Klein four group V4 - the only non-cyclic group of order 4
(really there are 2 groups of order 4: the rst one is the Klein four group, the second one
is the cyclic group of order 4). We will also see the group when we will investigate the
solvability of group S4 at example 8.7.
3 First of all by the Galois correspondence (Theorem 6.7) for any normal sub-
group of the Galois group (V4 in our case) there exists a sub-extension that is xed by
the normal sub-group. For our V4 (see Klein four group) we have 3 normal subgroups:
{id, σ3 }, {id, σ5 }, {id, σ7 }. Lets consider the last one i.e. lets nd the extension that cor-
responds to {id, σ7 }. The extension has the following form (we are using triviality of id:
Lid = L)
{id,σ7 } σ
Q (ζ8 ) = Q (ζ8 ) 7
σ
To calculate Q (ζ8 ) 7 we have to nd a Primitive element (Example 5.13) ν such that
ν∈/ Q, ν ∈ Q (ζ8 ) and σ7 (ν) = ν (i.e. σ7 xes Q (ν)). Using the fact that ζ̄8 = ζ87 , it can
be easy to check that
σ7 ζ8 + ζ̄8 = σ7 ζ8 + ζ87 =

= ζ87 + ζ849 = ζ87 + ζ86·8+1 = ζ87 + ζ8 = ζ̄8 + ζ8 ,
i.e. we can take ν = ζ8 + ζ̄8 as an element that generates the required extension.
4 In the expression we used the following fact

Q (ζ8 ) ∩ R = Q (Re (ζ8 )) = Q ζ8 + ζ̄8 .
5 The following equations were used
σ3 ζ8 + ζ83 = ζ83 + ζ89 = ζ83 + ζ8

7.2. KUMMER EXTENSIONS 99
and nally (with note ζ85 = −ζ8 , ζ86 = −i) 6
Q (ζ8 )σ5 = Q ζ8 · ζ85 = Q ζ86 = Q (i) .

Example 7.2 (n = 5). Q (ζ5 )

2πi
where ζ5 = e 5 . The Galois group is the
following:
Gal ∼
= (Z/5Z)
×
7 8
that is a Cyclic group of order 4. It is generated by ζ5 → ζ52 and it has
only one Proper subgroup ∼
= Z/2Z (see theorem A.25) so our eld Q (ζ5 ) has
only one sub-eld dierent from Q of course and this going to be a real part
all of the complex conjugation which are part of Galois group. Now this is
2π

the same as the real part Q (ζ5 ) ∩ R = Q ζ5 + ζ̄5 = Q cos 5 .
So these were the examples of cyclotomic extensions of Q and of course the

picture is exactly the same as long as the cyclotomic polynomial is irreducible.
If it is not reducible, which can happen as we have seen, the Galois group
becomes smaller.
7.2 Kummer extensions

Consider a eld K such that the Characteristic of K is prime to a certain
number n: (char(K), n) = 1 9 and such that X n − 1 splits in K . So K
contains
√ all roots of unity. Consider an element a of K : a ∈ K and let
α = a (i.e. a root of X n − a). Take
n
d = min i | αi ∈ K . (7.1)

and
σ5 ζ8 · ζ85 = ζ85 · ζ825 = ζ85 · ζ83·8+1 = ζ85 · ζ8 .

6 we cannot choose ζ8 + ζ85 = 0 and have chosen ζ8 · ζ85 instead of it.

7 As it was mentioned above there are only 2 nite group of order 4. The rst one
V4 (see Klein four group) was considered at example 7.1. There is the second one that
isomorphic to the cyclic group of order 4:
×
(Z/5Z) = {1, 2, 3, 4}
8 Gal = hζ5 i and the group action on an element is just a multiplication by ζ5 i.e.
id → ζ5 , ζ5 → ζ52 , ζ52 → ζ53 , ζ53 → ζ54 , ζ54 → id
9 case char(K) = 0 is also included because it is used only one time to prove separability
(see note 11)
Proposition 7.3. d | n, minimal polynomial of α is X d − αd and K (α) is

a Galois extension with cyclic Galois group of order d.
Proof. It's clear that K (α) is Galois because all the n-th roots of unity are
in K . So K (α) contains all roots of X n − a. 10 Therefore K (α) is a splitting
eld of X n − a thus it's normal. The extension is also separable because
(char(K), n) = 1. 11 Thus K (α) is Galois extension.
Lets dene a Homomorphism f : Gal (K (α) /K) −−−−→ µn . This is
g(α)
g→ α
g(α)
correct because g sends α to another root of X n − a thus the quotient α
is
a root of unity: 12 n
g (α)
= 1.
α
The homomorphism f is Injection because g (α) determines g . 13 What's
the image? It should be a Subgroup of a Cyclic group µn but the subgroup
should be also cyclic. 14 Letδ is the order of the image and we want to show
that δ = d. Consider g αδ = f (g)δ · αδ = αδ because f (g) is a root of 1
√
(f (g) = δ 1). 15 Thus αδ ∈ K ( see (5.5)). And αi ∈ / K for i < δ since
otherwise deg Pmin (α, K) = i < δ . But this is impossible because
[K (α) : K] = |Gal (K (α) /K)| = δ.

16
Thus only possible option is d = δ . Thus Pmin (α, K) = X d − αd .
10 Consider αk = αζ k where k = 0, 1, . . . , n − 1. All such αk are roots of X n − a because
k
αkn − a = αn (ζ n ) − a = a − a = 0.
We also have that ∀k1 6= k2 : αk1 6= αk2 because ζ k1 6= ζ k2 . Therefore we have n distinct
roots i.e. all roots are in K (α)
11 If P = X n − a then P 0 = nX n−1 6= 0 as soon as (char(K), n) = 1 and as result
n n
(Pn , Pn0 ) = 1 and Pn does not have multiple roots and therefore it is separable. As result
the extension K (α) is also separable.
The case char(K) = 0 is obvious because such extensions are always separable (see
section 3.4).
12 g n (α) − αn = g n (α) − a = 0
13 If we have g 6= g then g (α) 6= g (α) because in the case g (α) and g (α) are 2
1 2 1 2 1 2
dierent roots of X n − a. As result the homomorphism f is Injection.
14 see Fundamental theorem of cyclic groups (Theorem A.24)
15 Using g (α) g (α) = g (α · α) = g α2 :

δ g δ (α) g αδ
f (g) = = .
αδ αδ
16 d was chosen accordingly (7.1) and therefore δ ≥ d.

7.3. ARTIN-SCHREIER EXTENSIONS 101
Proposition 7.4. And conversely (to 7.3) for all cyclic extension of degree
√
n such that (char(K), n) = 1 is generated by n
a for some a ∈ K.
Proof. Consider L is an extension of K . Gal (L/K) = hσi then we have

σ = id. Linear algebra says that σ is Diagonalizable map. 17 Now, let us
n
show that all eigenspaces

have dimension 1. Indeed if x, y are in the same
Eigenspace then σ xy = xy 18 because x and y are multiplied by the same
number. Therefore xy ∈ K . And this is exactly means that dimension of
the eigenspace is 1, x, y are proportional over K . 19 Thus all roots of 1 are
eigenvalues of σ . 20 Then take α such that σ (α) = ζα where ζ is a Primitive
roots of unity. Then hσi - orbit of α has n elements therefore [K (α) : K] = n
(see explanation below) and αn ∈ K since σ (αn ) = ζ n · αn = αn . We see
that α is a root of X n − a. This is irreducible by degree reason.
Maybe I should have said here, why it follows from the formula, hσi -
orbit of α has n elements that the degree of the extension is exactly n. While
this is easy because either degree of the extension was less than n, then also,
α would have to be xed by some non-trivial subgroup of Galois group by
Galois correspondence. And then its orbit would have less than n elements.
21
7.3 Artin-Schreier extensions

Denition 7.5 (Artin-Schreier extension). Let L is an extension of a eld K
such that charK = p, where p is a prime number. If degree of the extension
n = [L : K] is equal to p, i.e. n = charK then the extension is called as
Artin-Schreier extension.
17
Apply theorem A.124 to the diagonalizable σ n = id. Really a map is diagonalizable
if it has n distinct eigenvalues (see theorem A.123). This fact will be proved below.
18 Lhσi = K i.e. Lσ = K .
19 I.e. if L is the eigenspace then we have that ∃x ∈ L such that ∀y ∈ L : y = kx, where
k ∈ K . This exactly means that dim L = 1
20 Let x - eigenvector of σ and ν is the eigenvalue. We have σ (x) = νx, using σ n = id
and σ ◦ σ(x) = σ(νx) = ν 2 x, one can get σ n (x) = ν n x = id(x) = x. Thus ν n = 1 i.e. ν is
a root of unity.
21 If we have K ⊂ K (α) ⊂ L and hσi is a Galois group L/K and [K (α) :] = d < n then
by Galois correspondence (Theorem 6.7) there exists a subgroup H ⊂ hσi (cyclic as soon
as hσi) that xes the K (α) and especially ∀h ∈ H : h (α) = α therefore σ k (α) < n i.e.
the orbit contains less than n elements.
Another explanation as follows. If [K (α) :] = d < n then Pmin (α, K) has d roots and
the orbit Orb (α) consists of roots of Pmin (α, K) and the number of roots is d < n.
Denition 7.6 (Cyclic extension). The Galois extension is called cyclic ex-
tension if the corresponding Galois group is cyclic.
Theorem 7.7 (Artin-Schreier). p = char(K) and let P = X p − X − a ∈

Let
K [X]. Then P is irreducible or splits over K . Let α be a root. If P is
irreducible then K (α) is Cyclic extension of K of degree p.
Conversely any cyclic extension of degree p is like this: L/K, ∃α ∈ K
p
such that L = K (α), α - root of X − X − a for some a ∈ K .
Proof. First of all notice that roots of P are α + k where k ∈ Fp (k is an

element of prime eld). 22
If P is irreducible then Galois group should be transitive on the roots (see
remark 5.21) then ∃σ ∈ Gal (K (α) /K) such that σ (α) = α + 1 (because
roots of P are α + k ). The Order of element in group for σ is p = [K (α) : K]
23
so the σ must generate the Galois group: Gal (K (α) /K) = hσi.
We have to show that if P is not irreducible then P splits i.e. α ∈ K .
Leave it for an exercise 24
Now we will prove the converse statement. Let L is a Cyclic extension
of K of degree p. We want to nd α such that σ (α) = α + 1 where σ is a
generator of Gal (L/K) (we know that the Galois group is cyclic i.e. must
have the following form Gal (L/K) = hσi).
22 In Fp we have k p = k and therefore
p
(α + k) − (α + k) − a =
= αp + k p − α − k − a = αp + k − α − k − a =
= αp − α − a = 0,
as soon as α is a root of X p − X − a.
23
σ p = σ(σ(σ(. . . (σ(α)) . . . ))) = α + p = α

i.e. σ p = id and order of hσi is p.
24 Let α is a root then we can get p dierent roots as α + k where k ∈ F . Suppose that
p
α∈ / K and Q is a factor of P (it should exist as soon as P is reducible). Q splits in K (α)
as soon as P splits there and its factors have the form X − (α − i). Let d = deg Q, i.e.
Y
Q= (X − (α − i)) = X d + a1 X d−1 + · · · + ad .
i
Consider the coecient a1 . It should

P have the form a1 = i (α − i) = dα − j (where
P
i is taken from d factors of Q and i i = j - another integer). a1 ∈ K (and as result
Q ∈ K [X]) only if d = 0 mod p. Therefore P has no nontrivial factors that is in the
contradiction with statement that P is reducible.
Thanks Ben Petschel for the hint.
7.3. ARTIN-SCHREIER EXTENSIONS 103
Set f = σ − id, K = ker f 25 and the Rank rgf = p − 1. 26 We have

(σ − id)p = 0 27 so the Kernel must be included into Image: K = ker f ⊂
Im f because otherwise L = ker f ⊕ Im f (L is a direct sum 28 of kernel and
image) and f k is never zero (but we have f p = 0) 29 .
So as soon as K is in the image of f then ∃α ∈ L such that f (α) = 1 30
but this means that σ (α) = α + 1. Now consider σ (αp − α) = (α + 1)p −
(α + 1) = αp −α (because we are in the eld of Characteristic p). This means
that αp − α ∈ K because the eld is xed by Galois group (see (5.5)). So
αp − α = a ∈ K and α is a root of X p − X − a and this nished the proof of
the theorem.
Remark 7.8. 31 Theorem 7.7 shows that Artin-Schreier extension is a Cyclic

extension as well as Kummer extensions (section 7.2) . The dierent be-
25 this is because ∀x ∈ K : σ (x) = x (see (5.5)).

26 In lectures we can hear about range (Image) not a Rank but by future content we
spoke about the rank but not about range (image). In any way the equation rgf = p − 1
requires some explanation. As soon as f p−1 6= 0 ∃x ∈ L such that f p−1 (x) 6= 0. Therefore
f k (x) 6= 0 for all k < p − 1 because in other case
f p−1 (x) = f . . . f f k (x) . . . = f (. . . f (0) . . . ) = 0.

Lets show that f (x) , f 2 (x) . . . , f p−1 (x) is linearly independent. For k ∈ K, x ∈ L we

can get
f (k · x) = σ (k · x) − k · x = σ (k) · σ (x) − k · x = k · σ (x) − k · x = k · f (x) .
Thus for ki ∈ K such that:

p−1
X
ki f i (x) = 0
i=1
applying f p−1
one can get that k1 = 0, applying f p−2 gives us k2 = 0. Continue the way
we get that all ki = 0. That proves the linear independence. Therefore rg (f ) ≥ p − 1. The
vector y = f p−1 (x) , f p−1 (x) . . . , f p−1 (x) is not zero but f (y) = 0 therefore y ∈ ker f .
This means that dim ker f ≥ 1. Using Ranknullity theorem (Theorem A.115) one can
conclude that only possible choice there is dim ker f = 1 and rg (f ) = p − 1.
27 In F we have
p
p
(σ − id) = σ p − id = 0
as soon as hσi has order p.
28 see also denition A.104 and example A.105
29 As soon as f 6= 0 exists x ∈ L such that y = f p−1 (x) 6= 0 but f (y) = f p (x) = 0,
therefore y ∈ ker f . Using the fact that dim ker f = 1 (follows from rg(f ) =p − 1 and
Ranknullity theorem (Theorem A.115) ) and y ∈ Im f (y = f f p−2 (x) ) one can
conclude that the Kernel must be included into Image.
30 This is because 1 ∈ K but K ⊂ ker (f ) ⊂ Im (f ) therefore 1 ∈ Im (f ).
content.
tween both is in relation between n and p. Kummer extensions (section 7.2)

plays a role in solvability for elds over Q (see Extension solvable by radicals),
Artin-Schreier extension plays the same role for nite elds.
7.4 Composite extensions. Properties

Denition 7.9 (Composite extension). Let L1 and L2 are extensions of
K both contained in some extension L (for instance the Algebraic closure
K̄ ). The composite extension L1 L2 is the extension they generate: L1 L2 =
L2 L1 = K (L1 ∪ L2 ). I.e. the composite extension is the smallest extension
that contains both L1 and L2 .
Another way to view this: consider the tensor product L1 ⊗K L2 - there is

a K -algebra. By Universal property there is a map from the tensor product
to L: j : L1 ⊗K L2 → L such that j (l1 ⊗ l2 ) = l1 l2 32
f : (l1 , l2 ) → l1 l2
L1 × L2 L
φ f˜ = j
L1 ⊗K L2
The Image Im j is a sub-algebra of L. If L is algebraic then any sub alge-

bra is a sub eld (see proposition 1.23) and this is exactly the eld generated
by L1 L2 . In general we can take its fraction eld (to obtain a eld from a
ring (an algebra)) but in our case, as it was mentioned above, as soon as L
is algebraic then L1,2 are elds.
Property 7.10. If L1 is separable (pure inseparable, normal, nite of degree

n) over K then L1 L2 is also separable (pure inseparable, normal, nite of
degree ≤ n) over L2
Proof. Let x ∈ L1 (L1 L2 is generated by L1 over L2 ) 33 then it's minimal

polynomial Pmin (x, L2 ) is a divisor of Pmin (x, K) in L2 [X] (see proposi-
tion 1.24). Therefore Pmin (x, L2 ) has a degree ≤ n where n is degree of
Pmin (x, K).
32 By the Universal property j is a Field homomorphism as soon as L1 , L2 , L1 ⊗K
L2 are the elds (see also theorem 7.12), but by lemma About homomorphism of elds
(Lemma 1.6) the homomorphism is injection.
33 We have K ⊂ L ⊂ L L as the case there
2 1 2
7.4. COMPOSITE EXTENSIONS. PROPERTIES 105
So if Pmin (x, K) is separable (pure inseparable) then Pmin (x, L2 ) is sep-

arable (pure inseparable). 34
The same is true for splitting so the normality is preserved. 35
About dimensions (nite extension of degree in the property formula-
tion). By the Base-change (Theorem 4.13) 36 :
dimK L1 = dimL2 (L1 ⊗K L2 )

37
and as soon as L1 L2 is the Im j :
dimL2 (L1 ⊗K L2 ) ≥ dimL2 (L1 L2 )
i.e.
dimL2 (L1 L2 ) ≤ dimK L1 = n.
Property 7.11. L1 , L2 are separable (pure inseparable,

If normal, nite of
degree n and m) over K then L1 L2 is also separable (pure inseparable, nor-
mal, nite of degree ≤ nm) over K
Proof. We have the following towers:
K ,→ L1 ,→ L1 L2
34 From proposition 1.24) we know that Pmin (x, L2 ) is a divisor of Pmin (x, K). There-
fore if Pmin (x, K) does not have multiply roots then Pmin (x, L2 ) (its divisor) will also
have only non-multiply roots.
The inseparability is obvious because if the Pmin (x, K) has only one root the same will
be the truth for Pmin (x, L).
35 i.e. the polynomial splits in L if it splits in K ⊃ L
2 2
36 From Base-change (Theorem 4.13) one can get

HomK L1 , K̄ = HomL L2 ⊗K L1 , K̄
2
But proposition 3.25 says that

HomK L1 , K̄ = deg (Pmin (x, K)) = [L1 : K] = dimK (L1 )
and

HomL
2
L2 ⊗K L1 , K̄ = [L2 ⊗K L1 : L2 ] = dimL2 (L1 ⊗K L2 )
37Using Ranknullity theorem (Theorem A.115) one can conclude that for j : L1 ⊗K
L2 → L1 L2 , dim Im j ≤ dim L1 ⊗K L2 (the equal sign is when dim ker j = 0).
and all properties except normality are preserved in the towers so follows
from property 7.10. 38
Normality is obvious because if L1 is a splitting eld of the family poly-
nomials {Pi }i∈I and L2 is a splitting eld of the family polynomials {Qj }j∈J
then L1 L2 is a splitting eld of the union of those families {Pi , Qj }i∈I,j∈J . So
normality is obviously preserved.
7.5 Linearly disjoint extensions. Examples

Theorem 7.12. The following statements are equivalent (for algebraic ex-
tensions)
1. L1 ⊗K L2 is a eld
2. j is Injection
3. if we have x1 , x2 , . . . , xn ∈ L1 linearly independent over K then they

are linearly independent over L2
4. if we have two families: x1 , x2 , . . . , xn ∈ L1 linearly independent over

K and y1 , y2 , . . . , ym ∈ L2 linearly independent over K then xi yj are
also linearly independent over K
When L1 nite over K then all the statements are equivalent to [L1 L2 : L2 ] =
[L1 : K] or in other words [L1 L2 : K] = [L1 : K] [L2 : K] 39
Denition 7.13 (Linearly disjoint extensions). In the case L1 and L2 are
called linearly disjoint extensions
40
Proof. Equivalence 1 and 2 is clear because we have that L1 L2 = Im j .
38 From property 7.10 follows that if L1 is separable over K then L1 L2 is separable
over L2 . If L2 is separable over K then theorem 3.29 about separability says the L1 L2 is
separable over K as soon as K ⊂ L2 ⊂ L1 L2 .
About degrees: if n = [L1 : K] and m = [L2 : K] then from property 7.10 follows that
[L1 L2 : L2 ] ≤ n. Using theorem 1.27 [L1 L2 : K] = [L1 L2 : L2 ] [L2 : K] ≤ nm.
39 Using theorem 1.27 we have [L L : K] = [L L : L ] [L : K].
1 2 1 2 2 2
40 By the Universal property j is a Homomorphism, but by lemma About homomorphism
of elds (Lemma 1.6) the homomorphism is injection if L1 ⊗K L2 is a eld i.e. if the
homomorphism is a Field homomorphism.
If j is injection then from the fact L1 L2 = Im j we can conclude that for any x ∈
L1 ⊗K L2 such that x 6= 0 there exists y ∈ L1 ⊗K L2 such that j(x)j(y) = 1 as soon as
L1 L2 = K (L1 ∪ L2 ) is a eld (both L1 and L2 are algebraic). Therefore xy = 1 and for
any non zero element of L1 ⊗K L2 we can the the inverse one. This means that L1 ⊗K L2
is a eld.
7.5. LINEARLY DISJOINT EXTENSIONS. EXAMPLES 107
Then 2 implies 3: we have x1 ⊗ 1, . . . , xn ⊗ 1 are linearly independent over

L2 by base change property (see proposition 4.11). If j is injective then their
images x1 , . . . , xn are also linearly independent over L2 . This is because an
injective map transforms a linearly independent set of vectors into a linearly
independent set. 41
3 implies 4: if we have some relation i,jP aij xi yj = 0, aij ∈ K then since
P
xi linearly independent over K one can get j aij yj = 0 but as soon as yj
linearly independent we will get aij = 0.
Next 4 implies 2 (remember that P 2 is injectivity of j ). Take
P z ∈ L1 ⊗K L2
such that j (z) = 0. We have z = aij xi ⊗ yj and j (z) = aij xi yj = 0 i.e.
aij = 0 and therefore z = 0. I.e. j is Injection.
The part about nite degrees follows from the 4 properties. 42
Example 7.14. First of all, the extensions which have relatively prime de-
grees are always linearly disjoint.
[L1 : K] = n, [L2 : K] = m and (m, n) = 1 then L1 and L2 are
I.e. if
linearly disjoint. Indeedm and n must divide [L1 L2 : K] ≤ mn. With our
conditions [L1 L2 : K] = mn but this is one of denition of linearly disjoint
43
extensions (see denition 7.13).
√5
√
In particular Q 2 and Q 5 1 are linearly disjoint extensions because
√ √
2 : Q = 5 and Q 5 1 : Q = 4. 44
5

the degrees are Q
√5 2πi
From the other side with 1 = e 5 the following extensions are not
√5
2πi √
linearly disjoint: Q 2 and Q e 5 · 5 2 . Indeed in both cases L1 L2 is a
splitting eld of X 5 − 2 and (for the rst case) [L1 L2 : Q] = 4 · 5 = 20. In
the second case both [L1,2 : Q] = 5 and 5 · 5 6= 20.
So, you see that the dierence is rather subtle. Well, the obvious reason
in the second case is that those extensions are generated by rules of the same
polynomial, but, still some eort is needed to formalize why this is not linearly
41 Let x1 , . . . , xn are not linearly independent over L2 then there exists α1 , . . . , αn ∈

Pn
L2Psuch that ∃αk 6= 0 butP i=1 αi xi = 0. From other side xi = j (xi ⊗ 1) therefore
n n
j ( i=1 αi xi ⊗ 1) = 0 but i=1 αi xi ⊗ 1 6= 0 as soon as x1 ⊗ 1, . . . , xn ⊗ 1 are linearly
independent over L2 . Therefore we just got a contradiction: j cannot be injection.
42 Let L and L are linearly disjoint extensions over K with nite degree: [L : K] =
1 2 1
n, [L2 : K] = m. From 3 we can conclude that [L1 L2 : L2 ] = n and using theorem 1.27 we
will obtain [L1 L2 : K] = [L1 L2 : L2 ] [L2 : K] = nm.
43 The claim is the following: if (m, n) = 1 then L and L are linearly disjoint. From
1 2
property 7.11 [L1 L2 : K] ≤ mn. But from theorem 1.27 n | [L1 L2 : K], as soon as K ⊂
L1 ⊂ L1 L2 , and m | [L1 L2 : K], as soon as K ⊂ L2 ⊂ L1 L2 and therefore using (m, n) = 1
one can conclude that [L1 L2 : K] = nm and thus L1 and L2 are linearly disjoint by last
statement of√theorem 7.12. √
44 P − 2 and deg Pmin 5 2, Q = 5.
5
5

min 2, Q = X
√ √
Pmin 5 1, Q = X 4 + X 3 + X 2 + X + 1 and deg Pmin 5 1, Q = 4.

disjoint case. In particular we see that L1 ∩ L2 = Q does not imply that L1

and L2 are linearly disjoint over Q. It's exactly what's happen in the second
case.
7.6 Linearly disjoint extensions in the Galois

case
Theorem 7.15. Let L1 , L2 ⊂ K̄ - extensions of K . L1 is Galois extension
over K . Let K 0 = L1 ∩ L2 . Then L1 L2 is Galois over L2 . Gal (L1 L2 /L2 ) sta-
bilizes L1 . φ : g → g|L is an injective map of Gal (L1 L2 /L2 ) → Gal (L1 /K)
1
0 0
with image Gal (L1 /K ) and L1 , L2 are linearly disjoint over K .
Proof. The proof that L1 L2 is Galois over L2 is obvious. 45

The next statement is that Gal (L1 L2 /L2 ) stabilizes L1 . 46 Let x ∈ L1
and g ∈ Gal (L1 L2 /L2 ) then g (x) is a root of Pmin (x, L2 ). 47 It is also a
root of Pmin (x, K). 48 But all such roots are in L1 because L1 is a Galois
extension.
Therefore the map φ is well dened. The map is injective because if we
have some σ such that σ|L1 = σ|L2 = id then it should be σ = id. This is
because our extension is generated by L1 and L2 , so if it happened to be in
identity on them both, it must be an identity. 49
45 We have L1 is Galois extension and therefore normal and separable over K . By
property 7.10 this means that L1 L2 is normal and separable over L2 or in other words
L1 L2 is Galois over L2 .
46 I.e. ∀g ∈ Gal (L L /L ) and for any x ∈ L we have g(x) ∈ L or in other words
1 2 2 1 1
g(L1 ) = L1 (see also denition A.29).
47 This is because g permutes roots and the image of the permutation is the set of roots
of the following polynomial Pmin (x, L2 ). Note that x is also one of the roots.
48 This is because proposition 1.24 i.e. because P
min (x, L2 ) divides Pmin (x, K) i.e. all
roots of Pmin (x, L2 ) are also roots of Pmin (x, K).
49 We have that φ(id) = id i.e. φ is Injection. Because if g , g 6= id then the equality
1 2
φ(g1 ) = φ(g2 ) holds if g1 g2−1 = id i.e. if g1 = g2 that is accordingly with the injectivity
denition.
Another proof is the following. Accordingly the sta comment on last section of lecture
6 (see note 42). We have to take an arbitrary element g ∈ L then if its image is identity
(φ(g) = id) then g = id. We got the required property if we assume
(
φ(g) (x) = g (x) , if x ∈ L1
φ(g) (x) = id (x) = x, if x ∈ L2 .
In the case if φ(g) = id then g|L1 = id and g|L2 = id or as it was mentioned in the lectures
g = id. This nishes the proof that φ is injective.
See also theorem A.136.
7.7. ON THE GALOIS GROUP OF THE COMPOSITE 109
So now lets nd the image of φ. If g (x) = x, ∀g ∈ Gal (L1 L2 /L2 ) then
x ∈ L2 by Galois correspondence (Theorem 6.7) . So if also x ∈ L1 then
it should be x ∈ K 0 = L1 ∩ L2 . So if L1 is nite over K then by Galois
correspondence (Theorem 6.7) we can conclude that Im φ = Gal (L1 /K 0 )
50
because the Fixed eld is K 0 .
In general 51 we have to nd nite sub-extension of L1 : let denote it as
L01 . We also have a nite Galois sub extension of L1 that contains L01 . We
denote this Galois sub-extension as L001 . 52
We have L001 and L2 are linearly disjoint over K 0 53 then it follows that L1
and L2 are also linearly disjoint over K 0 (see theorem 7.12 point 3).
Several additional comments about the Im φ. Let γ ∈ Gal (L1 /K 0 ) then
exists an element in Gal (L1 L2 /L2 ) which is sent by φ to γ . 54 We have
j : L1 ⊗K L2 ∼ = L1 L2 and we can take j · (γ ⊗ idL2 ) · j −1 - this will be the
element of the required Galois group. 55
7.7 On the Galois group of the composite

From the theorem 7.15 follows the following proposition
Proposition 7.16. 1. L1 andL2 are both Galois
over K and linearly dis-
joint then the following map g → g|L , g|L denes the isomorphism
1 2
50 In the lectures Ekaterina marked it as Im φ = Gal (L/K 0 ) but really we should have
L1 instead of L there.
51 not nite L over K
1
52 ??? The claim require a proof. Ekaterina provided only initial ideas about how the
proof can look like: You can take the union of all images of L01 by all automorphisms. And
this will be a nite union since L01 was nite, so there are nitely many possible roots of
minimal polynomials so there are not really many possibilities for the images of this L01 .
So I shall leave it as an exercise , but the solution is more or less what I just have told
you.
53 As soon as j is injection, the theorem 7.12 gives us that L00 and L are linearly disjoint
1 2
over K 0 .
54 The element exists if Im φ = Gal (L /K 0 ).
1
55 I.e. there is an element g ∈ Gal (L L /L ) such that φ(g) = γ . Let x ∈ L (but
1 2 2 1
x∈/ L2 i.e. x ∈ / K 0 because for such x and ∀g ∈ Gal (L1 L2 /L2 ) it will be g(x) = x as
well as γ(x) = x i.e. obviously the required property is satised) then, using equality
x = x · 1L2 , one can get
j −1 (x) = x ⊗ 1L2
and therefore
φ(g)(x) = j · (γ ⊗ idL2 ) · j −1 (x) =

= j · (γ(x) ⊗ idL2 (1)) = γ(x) · 1 = γ(x).
I.e. φ(g) = γ .
56
Gal (L1 L2 /K) ∼

= Gal (L1 /K) × Gal (L2 /K)
2. conversely to the rst part: if Gal (L/K) = G1 × G2 then L = LG1 LG2

57
which are linearly disjoint over the intersection.
Proof. The rst part is very sure because the injectivity of this map
is clear: if something is trivial both on L1 and L2 , then it's trivial on
the composite, 58 so I only have to prove the surjectivity. I will use the
same trick as before: L1 ⊗K L2 ∼ =j L1 L2 then j · (g1 ⊗ g2 ) · j −1 goes to
59
(g1 , g2 ).
The second part. LG1 and LG2 are both Galois 60 because G1 and G2
are normal in the product (see property A.35): G1 , G2 / G1 × G2 . What
56 Gal (L1 /K)×Gal (L2 /K) is a Direct product of 2 groups Gal (L1 /K) and Gal (L2 /K).
57 i.e. LG1 and LG2 are linearly disjoint over LG1 ∩ LG2
58 If we take an arbitrary element g ∈ Gal (L L /K) and the image of the element
1 2
is: id|L1 , id|L2 ∈ Gal (L1 /K) × Gal (L2 /K) then the taken element is the identity too:

g = id. This proves the injectivity i.e. the proof is the same as in the note 49.
See also theorem A.136.
59 Let x ∈ L then using equality x = x · 1 , one can get
1 L2
j −1 (x) = x ⊗ 1L2
and therefore
j · g|L1 ⊗ g|L2 · j −1 (x) = j · g|L1 (x) ⊗ g|L2 (1) =

= g|L1 (x) · 1 = g1 (x).
If x ∈ L2 then using equality x = 1L1 · x, one can get
j −1 (x) = 1L1 ⊗ x
and therefore
j · g|L1 ⊗ g|L2 · j −1 (x) = j · g|L1 (1) ⊗ g|L2 (x) =

= 1 · g|L2 (x) = g2 (x).
Thus we have the following construction

(
j · (g1 ⊗ g2 ) · j −1 → g1 = gL1 , ∀x ∈ L1
j · (g1 ⊗ g2 ) · j −1 → g2 = gL2 , ∀x ∈ L2
In the obvious case x ∈ L1 ∩ L2 we have g1 = g2 . Thus we can write j · (g1 ⊗ g2 ) · j −1 →

(g1 , g2 ).
60 see theorem 6.7 point 2a
7.7. ON THE GALOIS GROUP OF THE COMPOSITE 111
I mean is G1 embedded to the product by identifying it with G1 × e

where e is the neutral element of G2 .
The intersection LG1 ∩ LG2 is xed by G so LG1 ∩ LG2 = K . 61 Linear
disjoint follows from LG1 ∩ LG2 = K since we are in the Galois case.
62
Let me give you a small example:
Example 7.17. We have a Composite extension Q (ζn ) Q (ζm ) = Q (ζn , ζm )

63
Q (ζn , ζm ) = Q ζLCM (n,m) 64 therefore if (n, m) = 1 then
2πi
where ζn = e n .
Q (ζn ) and Q (ζm ) are linearly disjoint. 65 It can be seen as follows: we can
apply proposition 7.16 to our Galois groups then Q (ζn , ζm ) = Q (ζnm ) but by
61
Lets dene the action of g = (g1 , g2 ) ∈ G1 × G2 on L1 L2 . By denition Composite
P (1) (2)
extension, any element of L1 L2 can be represented in the form x = ij li lj , where
(1) (2)
li ∈ L1 , lj ∈ L2 (see also Theorem 7.12 (statement 4)). Thus

(1) (2)
X
g(x) = (g1 , g2 )(x) = g1 li g2 lj .
ij
We also have
(1) (2)
X
g1 (x) = g1 li lj
ij
and
(1) (2)
X
g2 (x) = li g2 lj .
ij
Let x ∈ LG1 ∩ LG2 then ∀g1 ∈ G1 , g2 ∈ G2 : g1 (x) = x, g2 (x) = x. Therefore

(1) (1) (2) (2)
g1 li = li , g2 lj = lj .
(1,2)
I.e. li,j ∈ K . Or in other words ∀g ∈ G : g = (g1 , g2 ) we have
X
(1) (2) (1) (2)
X
g(x) = (g1 , g2 )(x) = g1 li g2 lj = li lj = x.
ij ij
Therefore G xes LG1 ∩ LG2 , but G = Gal (L/K) thus LG1 ∩ LG2 = K .
62 We can use theorem 7.15 as soon as LG1 and LG2 are both Galois (see theorem 6.7
point 2a).
63 We can consider L = Q (ζ ), L = Q (ζ ) and L L = Q (ζ , ζ ).
1 n 2 m 1 2 n m
64 LCM - least common multiple. For instance multiples for 4 are 4, 8, 12, . . . . Multiples
for 6 are 6, 12, 18, . . . . Thus LCM (4, 6) = 12.
65 See example 7.14 where we got if ([L : K] , [L : K]) = (n, m) = 1 then L and L
1 2 1 2
are linearly disjoint.
66
the Chinese remainder (Theorem 4.20)
(Z/nmZ)× ∼ × ×
= (Z/nZ) × (Z/mZ) .
Thus Gal (Q (ζnm )) = Gal (Q (ζn )) × Gal (Q (ζm )). So the linear disjoint is
just got from the proposition 7.16.
66 and also by Fundamental theorem of cyclic groups (Theorem A.24)

Chapter 8
Solvability by radicals, Abel's
theorem. A few words on relation
to representations and topology
We nally arrive to the source of Galois theory, the question which motivated
Galois himself: which equation are solvable by radicals and which are not?
We explain Galois' result: an equation is solvable by radicals if and only if
its Galois group is solvable in the sense of group theory. In particular we see
that the "general" equation of degree at least 5 is not solvable by radicals.
We briey discuss the relations to representation theory and to topological
coverings.
8.1 Extensions solvable by radicals. Solvable

groups. Example
8.1.1 Extensions solvable by radicals
Let K is a eld of Characteristic 0: char(K) = 0. It is embedded into its
Algebraic closure.
Denition 8.1 (Extension solvable by radicals). A nite extension E of K is

solvable by radicals if ∃α1 , . . . , αr generating E such that αini ∈ K (α1 , . . . , αi−1 )
for some ni ∈ N.
√ p √
Example 8.2. Let K = Q, E = Q 3 2 + 3 7, 5 4 + 5 11 . We have
p
√ √ p3
√ p
5
√
α1 = 7, α2 = 11, α3 = 2 + 3 7, α4 = 4 + 5 11.
113
114CHAPTER 8. SOLVABILITY BY RADICALS, ABEL'S THEOREM. A FEW WORDS O
Denition 8.3 (Polynomial solvable by radicals). P ∈ K [X] is called solv-

able by radicals if exists a E - Extension solvable by radicals and containing
all roots of P .
So more precisely, it would say that the equation, P = 0 is solvable by

radicals.
Property 8.4. 1. Composite extension of solvable by radicals is itself

solvable by radicals
2. If L K is solvable by radicals (by denition L should be

extension of
nite extension of K ) then exists a nite Galois extension E containing
L and solvable by radicals.
Proof. For the rst property (the proof is missing in the lectures): let L =
L1 L2 where L1 and L2 are solvable i.e. L1 = K (α1 , . . . , αn ) , L2 = K (β1 , . . . , βm )
with αi , βj which satises properties from denition 8.1. In the case we
can assume L = L1 L2 = K (L1 ∪ L2 ) = K (α1 , . . . , αn , β1 , . . . , βm ) and all
properties from denition 8.1 will also be satised. Therefore the composite
extension L = L1 L2 is also solvable.
For the second property: Indeed take a composite
1 of all images of L in
K̄ or, in other words, images of L by Gal K̄/K .
1 Lets prove by induction. For the rst step we have that K (α) is solvable therefore
∃d = min{j | αj ∈ K}. Let ζd is a root of unity i.e. root of X d − 1 then by proposition
7.3, the K (ζd , α) ⊃ K (α) is a Galois extension and it has a nite degree (≤ d · d).
Using induction hypothesis we have that if K (α1 , . . . , αi ) is solvable then there exists a
Galois extension of nite degree F = K (β1 , . . . , βk ) that is solvable and K (α1 , . . . , αi ) ⊂
F . By Galois correspondence (Theorem 6.7) (item 2b) we have that ∀g ∈ G =
Gal K̄/K :

g (F ) = F.
Lets consider F (αi+1 ). As we know (by induction hypothesis) K (α1 , . . . , αi , αi+1 ) is

solvable i.e. ∃d = min{k | αi+1
k
∈ K (α1 , . . . , αi )} therefore αi+1 is a root of the following
irreducible polynomial X − a where a = αi+1
d d
. The Galois group G = Gal K̄/K
permutes the roots of the polynomial and as result ∀g ∈ G
Lg = g (F (αi+1 )) = F (βg ) ,
where βg = g (αi+1 ) is a root of X d − a. I.e. Lg is solvable. If we take a composite

extension L ∪g∈G Lg then by rst property of 8.4 it will be solvable but as soon as L is the
image of all g ∈ G then ∀g ∈ G : g (L) = L and by Galois correspondence (Theorem 6.7)
(item 2b) the extension L is Galois. L is a splitting eld of X d − a (because it contains
all its roots) but by theorem 2.7 it has nite degree: [L : F ] ≤ d!.
8.2. PROPERTIES OF SOLVABLE GROUPS. SYMMETRIC GROUP115
8.1.2 Solvable groups

This shall be a brief reminder since this is not a course on group theory, you
are supposed to know some group theory already. So I somehow I presume
that you are familiar with this denition but I will recall the denition of
basic properties.
Denition 8.5 (Solvable group). G is called solvable if it has a ltration

i. e. G = G0 ⊃ G1 ⊃ · · · ⊃ Gr−1 ⊃ Gr = {e}, such that Gi is a Normal
subgroup of Gi−1 and the Quotient group Gi−1 /Gi is an Abelian group.
Example 8.6 (Group of permutations S3 ). Consider S3 - the group of per-

mutations (see also example A.56). It's solvable because S3 ⊃ A3 ⊃ {e}.
We have |S3 /A3 | = 2 (see example A.57) i.e. S3 /A3 is cyclic of order 2.
|A3 | = 3 i.e. A3 - cyclic of order 3. 2
Example 8.7 (Group of permutations S4 ). Consider S4 - the group of per-

mutations (see also example A.56). It's solvable because S4 ⊃ A4 ⊃ K ⊃ {e},
3
where K - is a subgroup of double transpositions (see example A.49 for per-
mutation cycles notation):
K = {e, (12)(34), (13)(24), (14)(23)} .
A double transposition is a product of two Transpositions with distinct sup-

port, right, which permute the distinct elements. A4 / S4 , |S4 /A4 | = 2, i.e.
S4 /A4 is cyclic of order 2. 4
5
K / A4 , |A4 /K| = 3, i.e. A4 /K is cyclic of order 3.
K is Abelian group and K ∼ = Z/2Z × Z/2Z.
So this shows that S4 is solvable.
8.2 Properties of solvable groups. Symmetric

group
Property 8.8. If G is solvable and H ⊂G is a subgroup of G then H is
solvable.
2 As it was mentioned at [64] there is only one group of order 3. We also have (with
theorem A.41) that A3 is Abelian group
3 There is a well known Klein four group V - the only non-cyclic group of order 4. It
4
also denoted as K4 . We have already seen it at example 7.1.
4 i.e. theorem A.41 gives us that S /A is Abelian group
4 4
5 i.e. theorem A.41 gives us that A /K is Abelian group
4
6
Proof. Indeed Gi ∩ H gives a ltration with required property.
Property 8.9. If G is solvable and H /G is a normal subgroup of G then

G/H is solvable.
Proof. Indeed consider a projection map
π : G → G/H (8.1)
7
then π (Gi ) gives a ltration (G/H)i on G/H with required properties.
6 We have Hi = Gi ∩ H and as soon as Gi / Gi−1 we also get Hi / Hi−1 . Really
∀h ∈ Hi−1 we have h ∈ Gi−1 ∩ H ⊂ Gi−1 . Thus, using normality (see denition A.13)
Gi / Gi−1 ,
hGi = Gi h.
The last equation also holds for any subset of Gi−1 and especially for Hi = Gi ∩ H ⊂ Gi .
I.e. ∀h ∈ Hi−1 : hHi = Hi h or in other words, Hi / Hi−1 .
We have that Gi−1 /Gi is an Abelian group and now we have to prove that the Quotient
group Hi−1 /Hi is abelian. ∀h0 ∈ Hi−1 /Hi , ∃hi−1 ∈ Hi−1 such that h0 = {hi−1 , Hi }.
From other hand hi−1 ∈ Gi−1 and Hi ⊂ Gi therefore we can associate with h0 the
g 0 = {hi−1 , Gi } ∈ Gi−1 /Gi . I.e. we just got an injection f : Hi−1 /Hi −−0−−→
0
Gi−1 /Gi .
h →g
Obviously we have ∀h0 , h00 ∈ Hi−1 /Hi the following relation
f (h0 h00 ) = g 0 g 00 = f (h0 ) f (h00 ) ,
where g 00 = {h00 , Gi } and g 0 g 00 = {h0 h00 , Gi }. Therefore f is Homomorphism. Thus

∀a, b ∈ Hi−1 /Hi we have f (ab) = f (a)f (b), but as soon as Gi−1 /Gi is abelian,
f (a)f (b) = f (b)f (a) = f (ba), or, using property of homomorphism, one can get that
ab = ba i.e. Hi−1 /Hi is an Abelian group and as result H is solvable.
7 We can associate each element g ∈ G with ḡ ∈ G/H via the following map: π :
G −−−→ G/H . I.e. π(g) = ḡ = gH . Thus we also have π (Gi ) = Ḡi = Gi H . Lets prove
g→ḡ
that Ḡi forms the required ltration i.e. that Ḡi / Ḡi−1 and Ḡi−1 /Ḡi is Abelian group.
For normality (see denition A.13) prove we have that Gi / Gi−1 i.e. ∀g ∈ Gi−1 : gGi =
Gi g but ∀ḡ ∈ Ḡi−1 we have
ḡ Ḡi = gHGi H
but H / G i.e. ∀g 0 ∈ Gi ⊂ G : g 0 H = Hg 0 therefore
ḡ Ḡi = gGi H = Gi gH = Gi gHH = Gi HgH = Ḡi ḡ
that nished the normality proof.

For the abelian quotient proof we have that Gi−1 /Gi is Abelian group therefore ∀a, b ∈
Gi−1 : abGi = baGi . Using the fact that H is a Normal subgroup one can get
āb̄Ḡi = aHbHGi H = abGi H = baGi H =

= baGi HH = baHGi H = baHHGi H = bHaHGi H =
= b̄āḠi ,
i.e. Ḡi−1 /Ḡi is Abelian group.

Property 8.10. If H / G, H and G/H are solvable then G is solvable.

Proof. Put togeter the ltration Hi and π −1 (G/H)j (see (8.1) for π de-
8
nition).
The following property is not a part of the lectures but it's required for
the property 8.12 proof
Property 8.11. If G is nite Abelian group than G is solvable and there

exists a nite ltration with cyclic quotients for G.
Proof. Accordingly theorem A.47) each nite Abelian group G can be repre-
sented as
G = K1 ⊕ K2 ⊕ · · · ⊕ Kn−1 ⊕ Kn ,
where Ki is a Cyclic group. If we denote
Gi = K1 ⊕ K2 ⊕ · · · ⊕ Ki
8 ∀ḡ ∈ Ḡ = G/H we can consider it as a set of elements: ḡ = {∀h ∈ H : hg} where
g ∈ G. If we combine then we will get the following: ∪g ḡ = G. We also have
GH = ∪g ḡH = ∪g gHH = ∪g gH = ∪g ḡ = G. (8.2)
Thus from Ḡi / Ḡi−1 one can get ∀ḡ ∈ Ḡi−1 : ḡ Ḡi = Ḡi ḡ or in other words (with Gi−1 =
∪ḡ∈Ḡi−1 ḡ ) ∀g ∈ Gi−1 we have the following
gGi−1 = gHGi−1 H = ḡ Ḡi−1 = Ḡi−1 ḡ = Gi−1 HgH = Gi−1 HHg = Gi−1 g,
i.e. Gi / Gi−1 . As result the ltration
{e} = H̄ / · · · / Ḡi / Ḡi−1 / · · · / Ḡ0 = Ḡ
produces the following one
H / · · · / Gi / Gi−1 / · · · / G0 = G. (8.3)
If Ḡi−1 /Ḡi is abelian, i.e. ∀ā, b̄ ∈ Ḡi−1 : āb̄Ḡi = b̄āḠi , we can get the following ∀a, b ∈ Gi−1
(as soon as (8.2) gives us Gi = Gi H )
abGi = abGi H = abGi HH = abHGi H = abHHGi H = aHbHGi H =

= āb̄Ḡi = b̄āḠi = bHaHGi H = baHGi H = baGi H = baGi
i.e. the Quotient group Gi−1 /Gi is an Abelian group.

Finally combining (8.3) with H solvability one can get
{e} / · · · / Hj / Hj−1 / · · · / H / · · · / Gi / Gi−1 / · · · / G0 = G
i.e. G is solvable.
when, by property A.35, we can get
{e} = G0 / G1 / · · · / Gn = G.
The gotten ltration has cyclic quotients because (see property A.36) Gi /Gi−1 =
Ki , where Ki is a cyclic group.
Property 8.12. If G is nite than G is solvable (i.e. has a nite ltration

with Abelian quotients) if and only if there exists a nite ltration with cyclic
quotients.
Proof. This is just because a nite Abelian group is just a product of cyclic
groups (see see The fundamental theorem of nitely generated abelian groups
(Theorem A.47) ). 9
Lets also look at another denition of solvable group
Denition 8.13 (Solvable group). G is called solvable if the following se-

quence is nite:
G ⊇ [G, G] = G(1) ⊇ G(1) , G(1) = G(2) ⊇ · · · ⊇ G(n−1) , G(n−1) = G(n) = {e}

where G(i) = G(i−1) , G(i−1) is the Commutator subgroup.

Remark 8.14. Denitions of solvable group 8.13 and 8.5 are equivalent.
Proof. Our ltration with Commutator subgroups G ⊇ G(1) ⊇ · · · ⊇ G(n) =
{e} is a ltration with abelian quotient because G/ [G, G] as well as G(i) / G(i) , G(i) =

G(i) /G(i+1) are Abelian groups. 10

9 If exists a nite ltration of G with cyclic quotient then the G is solvable because
Cyclic group is abelian (see theorem A.41).
The other direction is not so simple. Lets G is solvable then Gi /Gi−1 and Gi−1 /Gi = K
- abelian. The property 8.11 (that follows from The fundamental theorem of nitely
generated abelian groups (Theorem A.47) ) says us that there exists a nite ltration
with cyclic quotients for K i.e.
{e} = K (n) / · · · / K (j) / K (j−1) / · · · / K0 = K
where K (j−1) /K (j) is cyclic. The Correspondence theorem (Theorem A.18) says us that
(j) (j) (j−1) (j−1) (j)
∀K (j) , ∃Gi ⊂ Gi−1 with the same properties i.e. Gi / Gi and Gi /Gi is cyclic.
As result we can complete each Gi / Gi−1 with
(n) (j) (j−1) (0)
Gi = Gi / · · · / Gi / Gi / · · · / Gi = Gi−1
and the result ltration will have cyclic quotients.

10 See denition A.22 and theorem A.21
From the other hand if G/H is an Abelian group then H ⊃ [G, G]. 11 So
if a nite ltration with abelian quotient exists then the ltration given by
G(i) is also nite. It must terminate after a nite steps. So, this proves the
equivalence.
Theorem 8.15 (Sn solvability). Sn - the permutation of n elements (see

example A.53) is not solvable for n ≥ 5.
Proof. It's easy to use denition 8.13. Main steps are the following
1. we know that [Sn , Sn ] = An - subgroup of even permutations (see def-

inition A.52). It can be seen from the fact that any 3-cycle is a Com-
mutator subgroup 12 and 3-cycles generate An 13
2. If n ≥ 5 then [An , An ] = An thus the ltration generated by com-

mutators will never terminate i.e. will never reach the unity ({e})
and will stabilize on An . How we can see it? We can remember that
[A4 , A4 ] = K (see example 8.7) - the subgroup of double transpositions.
11 See theorem A.21
12 Let n ≥ 3 and (a, b, c) ∈ Sn is a 3-cycle then (see example A.62)
2
(a, b, c) = (a, b, c)2 = ((a, c)(a, b)) =
= (a, c)(a, b)(a, c)(a, b) = (c, a)−1 (b, a)−1 (a, c)(a, b) =
= (a, c)−1 (a, b)−1 (a, c)(a, b) =
= [(a, c), (a, b)] ⊂ [Sn , Sn ] .
13 Any even permutation can be represented in a form of product of transpositions

(see theorem A.63). The number of transpositions should be even. We have several cases
there:
(a) The transpositions with dierent elements produces 2 3-cycles
(a, b)(c, d) = (a, b, c)(b, c, d).
Really
c→d
d→b→c
(a, b, c)(b, c, d) = = (a, b)(c, d)
b→c→a
a→b
(b) The transpositions that have a same element produces a 3-cycle (see example A.62)
(a, c)(a, b) = (a, b, c)
As soon as we have even number of transposition then we will always have a translation
to a product of 3-cycles for any product of transposition pairs.
A4 ,→ An in many dierent ways. Because you can pick any 4 elements

among our n elements and just consider the permutations of those 4
elements as a subgroup of permutations of n elements and then taking
the commutators of those A4 , we see that all double transpositions are
in the [An , An ] (Commutator subgroup of An ). 14 But if n ≥ 5, they
generate An . 15
8.3 Galois theorem on solvability by radicals

The following claim missing in the lectures but it's important for the theorem
8.17 proof.
Claim 8.16. Let K⊂L⊂M and both M and L are Galois extensions over
K then
Gal (M/K) ∼
= Gal (L/K) .
Gal (M/L)
Proof. Consider the following group Homomorphism
φ : Gal (M/K) −−−→ Gal (L/K) .

g→g|L
14 The double transposition consists of 4 elements. We can consider all permutations

of the 4 elements only and ignore all others. For instance if n = 6 and we are interested
in all double transpositions of 1, 2, 4, 5. In the case we have to look at the following
permutations:
1 2 3 4 5 6
π1,2,4,5 = .
i1 i2 3 i3 i4 6
Note that we consider the even permutations only i.e. π1,2,4,5 ∈ A6 . From example 8.7
we know that the commutator of such permutations contains all double transpositions
for 1, 2, 4, 5: {e, (12)(45), (14)(25), (15)(24)} (the elements 3, 6 are not changed and dis-
appeared in the commutator). Continue such way we can conclude that for any double
transposition:
(i1 , i2 )(i3 , i4 ) ∈ [An , An ]
15 Really if n ≥ 5 then for any 3 elements a, b, c (which form a 3-cycle (a, b, c)) exist
2 elements e, d such that all elements a, b, c, d, e are dierent. Using the (e, d)(e, d) = id
one can get that 2 double transpositions (a, b)(d, e) and (d, e)(b, c) generate the 3-cycle
(a, b, c):
(a, b)(d, e)(d, e)(b, c) = (a, b)(b, c) = (a, b, c).
I.e. any 3-cycle can be generated by 2 double transpositions. Therefore (see note 13)
the whole An is generated by the double transpositions.
Thanks Arnur Nigmetov for the hint.
8.3. GALOIS THEOREM ON SOLVABILITY BY RADICALS 121
The Galois correspondence (Theorem 6.7) says that φ is a Surjection with

kernel Gal (L/F ). Therefore by First isomorphism (Theorem A.134) one
can get the required statement.
Theorem 8.17. Let P ∈ K [X]. P is a Polynomial solvable by radicals if

and only if Gal (P ) is solvable. There Gal (P ) is (by denition) Gal (F/K)
where F is a Splitting eld of P over K.
Proof. First of all lets proof that if Gal (P ) is solvable then P is solvable. Let
n = [F : K] and consider L = K (ζn ) where ζn - n-th root of 1. Let M = F L
is a Composite extension. So this is the splitting and eld of P of which we
have adjoined all the n-th roots of unity. Then M is a Galois extension 16
and Gal (M/L) ,→ Gal (F/K). 17 ∀g ∈ Gal (M/L) leaves F invariant. 18 If
g |F = id then g = id. Then the image in fact of this map is Gal (F/F ∩ L).
19
So G = Gal (M/L) is solvable 20 i.e.
G = G0 ⊃ G1 ⊃ · · · ⊃ Gr = {e}
and Gi /Gi+1 - cyclic 21 of order ni | n. 22 And as soon as ni | n (very

important), all n-th roots of 1 are in M (this is why we adjoin the L).
Let Mi = M Gi . We know Mi ,→ Mi+1 is a cyclic Galois extension of
order ni | n and roots of 1 are in it, thus there is Kummer extension. 23 and
16 Accordingly property 7.11 the composition of two Galois extensions will also be Galois.
17 See theorem 7.15 where L1 = F, L2 = L, L1 L2 = M = F L.
18 g(F ) = F see Galois correspondence (Theorem 6.7) (point 2b)
19 See theorem 7.15 where L = F, L = L, L L = M .
1 2 1 2
20 Using property 8.8 G is solvable as soon as G ⊂ Gal (F/K) and Gal (F/K) is solvable
21 see property 8.12
22 As soon as G , G
i i+1 ⊂ G then using Lagrange (Theorem A.8) one can get that
|Gi | | |G| as well as |Gi+1 | | |G| and therefore |Gi /Gi+1 | | |G|. But G ⊂ Gal (F/K), thus
|G| | |Gal (F/K)| = n and, as result, |Gi /Gi+1 | divides n.
23 We have M = M Gi where
i
M0 = M G0 = M G = M Gal(M/L) = L
and
Mr = M Gr = M {e} = M.
Thus by Galois correspondence (Theorem 6.7) , as soon as Gr / Gr−1 / · · · / G1 / G0 = G,
we have the following tower:
L = M0 ⊂ M1 ⊂ · · · ⊂ Mr−1 ⊂ Mr = M.
If we consider Mi ⊂ Mi+1 then we can get the following: Mi contains all n-th roots of
unity as soon as K (ζn ) = L ⊂ Mi ; Mi+1 is Galois extension over Mi with Galois group
Gi /Gi+1 (we have Gi = Gal (M/Mi ), Gi+1 = Gal (M/Mi+1 ) and therefore, by claim 8.16,
√
therefore Mi+1 = Mi ni ai (see proposition 7.4). So M = K (ζn , α1 , . . . , αr )
√
where αi = ni ai . Therefore M is solvable by radicals.
For another direction: if P is solvable then G is solvable. Let E is solvable
extension containing F . We may suppose (using property 8.4) that this is
Galois. Then write E = K (α1 , . . . , αr ) where αini ∈ K (α1 , . . . , αi−1 ). Then
let L = K (ζn ) where n = LCM ({ni }) so ∀ni : ni | n. And take M = LE .
We have K (α1 , . . . , αi−1 ) ,→ K (α1 , . . . , αi ) - cyclic extension of order ni .
We have Gal (M/L) is solvable by this cyclic subgroups. Gal (M/K) is also
solvable since
Gal (M/L) ⊂ Gal (M/K)
and the quotient (see claim 8.16)
Gal (M/K) ∼
= Gal (L/K)
Gal (M/L)
24 25
which is abelian. Gal (F/K) is a quotient of Gal (M/K) thus it is
solvable too. 26
8.4 Examples of equations not solvable by rad-

icals."General equation"
As we can see there exist equations which are not solvable in radicals.
Example 8.18 (Not solvable polynomial of degree 5). Let P ∈ Q [X] is an

irreducible polynomial with rational coecients of degree 5. It has 3 real roots
(and 2 complex conjugate roots) as it shown on the picture.
y
We claim that Gal (P ) = S5 . This is because
Gi /Gi+1 = Gal (Mi+1 /Mi )) which is cyclic of order ni | n and as result Mi+1 will be also
be cyclic of the same order ni | n. Therefore Mi+1 is a Kummer extension (see section
7.2) over Mi ,
24 as result the property 8.10 gives us the solvability of Gal (M/K)
25 ??? subset
26 by the property 8.8
8.4. EXAMPLES OF EQUATIONS NOT SOLVABLE BY RADICALS."GENERAL EQUATION"123
1. Gal (P ) contains the complex conjugation (we have 2 complex conju-

gated roots but Galois group is the group of automorphisms which ex-
change roots and the complex conjugation will exchange the 2 complex
27
roots). The complex conjugation is the transposition of roots
2. As soon as P is irreducible then Gal (P ) should act transitively (see

denition A.31) on roots (see remark 5.21 item 2). We have an irre-
ducible polynomial. We can always send one of its roots to another of
its roots. We have this isomorphism of stem elds which extends to an
automorphism of the splitting eld. But, what is the subgroup of S5 ,
which adds transitively?
Gal (P ) ⊂ S5 acts transitively. This means that 5 | |Gal (P )|. That

is because (see Orbit-stabilizer theorem (Theorem A.30) ) |G| =
|Orb (x)| |Gx | (Orb (x) - is the Orbit, Gx - Stabilizer subgroup) but the
28
orbit has 5 elements and therefore 5 divides the cardinality of G.
This means, by Sylow (Theorem A.37) theorems, that our group con-
tains something of order 5. But only 5-cycle has order 5 (see theorem
A.9). But a 5-cycle and transposition generate S5 (see corollary A.64).
So Gal (P ) = S5 .
In fact, the same argument is valid for Sp with every p - prime. I.e. applies
to an arbitrary prime number p instead of 5.
So Gal (P ) = S5 - not solvable and therefore P is not solvable by radicals.
Example 8.19 (General equation of degree n). What's the general equation.
It is the following
X n − T1 X n−1 + T2 X n−2 + · · · + (−1)n Tn ,
where Ti is a variable. Where does it come from? Let X1 , . . . , X n are roots

of a polynomial of degree n when the polynomial itself is
!
X
(X − X1 ) · · · · · (X − Xn ) = X n − Xi X n−1 +
i
!
X Y
+ Xi X j X n−2 + · · · + (−1)n Xi ,
i,j i
27(1, 2) = 1 → 2 → 1
28|Orb (x)| = 5 because there are 5 dierent roots that can be permuted by the Galois
group.
P P Q
i.e. T1 = Xi , T2 = i,j Xi Xj , . . . , Tn = i Xi .
i
One has K [T1 , . . . , Tn ] ⊂ K [X1 , . . . , Xn ] (multi-variable polynomial rings).
We have the same also for eld extensions: K (T1 , . . . , Tn ) ⊂ K (X1 , . . . , Xn ).
The K (X1 , . . . , Xn ) is algebraic and a splitting eld for our general polyno-
mial. So it has degree at most n!, i.e. [K (X1 , . . . , Xn ) : K] ≤ n! (see theorem
Sn 29
2.7). On the other hand K (T1 , . . . , Tn ) ⊂ K (X1 , . . . , Xn ) so degree of
30
the extension is n! and
K (T1 , . . . , Tn ) = K (X1 , . . . , Xn )Sn .
In particular the Galois group is Sn and our general polynomial is not solvable
by radicals if n ≥ 5. This is known as Abel theorem
8.5 Galois action as a representation. Normal

base theorem
Connection with group representations.
Denition 8.20 (Group representation). Let G is a nite group. V is a

Vector space over K . Representation of G is a Homomorphism ρ : G →
GL (V ) (where GL (V ) is the General linear group of a vector space i.e. the
group of Automorphisms of the vector space V ).
If L is a nite extension of K we can talk about it as about K -vector

space. So we have a representation of G as Galois group Gal (L/K): ρ :
G → GLK (L) - this is something that we have as the denition because we
dene the Galois group as the group of automorphisms of L over K .
We can ask the question: what's kind of representation is the ρ. We claim
that ρ is something that is called as Regular representation.
Denition 8.21 (Regular representation). Let a vector space V has a ba-

sis indexed by elements of group G: eg where g ∈ G. ρreg (h) 31 acts by
29 This is because Sn permutes the roots, for instance Xk → Xl → Xk , but the equations
for Ti are not changed during the transformation.
30 We have that S ⊂ Gal (K (X , . . . , X ) /K (T , . . . , T )) and therefore
n 1 n 1 n
[K (X1 , . . . , Xn ) : K (T1 , . . . , Tn )] ≥ |Sn | = n!.
Thus with [K (X1 , . . . , Xn ) : K] ≤ n! one can get that [K (X1 , . . . , Xn ) : K (T1 , . . . , Tn )] =

n! and [K (X1 , . . . , Xn ) : K] = n!.
31 where h ∈ G.
8.5. GALOIS ACTION AS A REPRESENTATION. NORMAL BASE THEOREM125
32
permutations:
ρreg (h) eg = ehg .
We claim that the representation of Galois group is the regular represen-

tation. We have seen that (see proof of the theorem 5.11)
L ⊗K K̄ ∼
= K̄ n .
The K̄ n is a Direct sum of n (n = |G = Gal (L/K)|) copies of K̄ . The

sum is indexed by the embeddings of L into K̄ . 33 Pick one j : L ,→ K̄ and
all others can be obtained by group Action j ◦ g, g ∈ G. 34 So K̄ n has a
basis indexed by G and the Action of G permutes the basis vectors. 35 So
32 ρreg (h) is an Automorphism of V i.e.
ρreg (h) : V −−−−−→ V.

eg →egh
33 The embeddings are also noticed asthe set of homomorphisms HomK L, K̄ . And

n
K̄ n = i=1 K̄ where n = HomK L, K̄ (accordingly (5.2)).
Q
34 This is because the Galois group acts transitively on the set of homomorphisms from
L to K̄ (HomK L, K̄ ), see theorem 5.17.

35 We have L is a Galois extension i.e., via Primitive element (Theorem 5.11) , ∃α ∈ L
such that L = K (α), where α is a root of a polynomial Pmin (α, K) of degree [L : K] = n.
The consider the set of the following vectors in K̄ n :
ei = φ1 (αi−1 ), φ2 (αi−1 ), . . . , φn (αi−1 ) ,

where φj -distinct homomorphisms such that φj (α) = αj - is another root of Pmin (α, K).
The set is linearly independent:
n n n n
!
X X X X
i−1 i−1 i−1
ci ei = ci α1 , ci α2 , . . . , ci αn =
i=1 i=1 i=1 i=1
= (P (α1 ), P (α2 ), . . . , P (αn )) ,
Pn
where P (X) = i=1 ci X i−1 . It can have α1 , α2 , . . . , αn as roots only if all ci = 0 because
the minimal polynomial ∀αi has degree n > n − 1. As result {ei } are linearly independent.
If we put the vectors into a matrix as rows we can get
 
φ1 (1) φ2 (1) φ3 (1) ... φn (1)
 φ1 (α) φ2 (α) φ3 (α) ... φn (α) 
 φ1 (α2 ) 2 2
φn (α2 ) 
 
A= φ2 (α ) φ3 (α ) ... .
.. .. .. .. ..
. . . . .
 
 
φ1 (αn−1 ) φ2 (αn−1 ) φ2 (αn−1 ) . . . φn (αn−1 )
The columns are also linearly independent (as rows) and therefore
vi = φi (1), φi (α), . . . , φi (αn−1 ) (8.4)

L ⊗K K̄ ∼= K̄ n ∼
= R, where R is a Regular representation of G over K̄ . In
particular ∃x ∈ L ⊗K K̄ such that gx |g∈G forms a basis of L ⊗K K̄ over K̄ .
36
And, as result, the elements of G are linearly independent in the space of
Endomorphisms EndK̄ L ⊗K K̄ 37
Theorem 8.22 (Normal base). ∃x ∈ L such that {gx | g ∈ G} is a K basis

of L.
G are linearly independent in the space of Endomorphisms EndK (L)
Proof. First of all consider a case when K is innite. Let pick some basis
e1 , . . . , en - K -basis in L. g1 , . . . , gn ∈ G. Let x ∈ L then g1 (x) , . . . , gn (x)
is a basis if and only if matrix formed by gi (x) in the basis ej has non zero
determinant. 38 But this determinant is a polynomial in the coecient,
is a basis of K̄ n . If g ∈ G such that g(αi ) = αj then
g(vi ) = g 1, αi , . . . , αin−1 =

= 1, αj , . . . , αjn−1 = φj (1), φj (α), . . . , φj (αn−1 ) = vj

36 you can take one basis vector from (8.4): x = vi

37 See also theorem A.98 about linearly independents of elements ofPa group G
38 We have x = X e + . . . X e where X ∈ K . ∀g ∈ G : g (x) =
1 1 n n i l l i Xi gl (ei ) because
gl (Xi ) = Xi (remember that LG = K ). For each gl (ei ) we can write
X (l)
gl (ei ) = kij ej (8.5)
j
(l)
We have kij are predened (because {gl } and {ei } are predened) elements of L and
therefore Xi can only be changed.
We want {gj (x)} to be linearly independent set of elements from L. Therefore the
equation
c1 g1 (x) + . . . cn gn (x) = 0,
where ci ∈ K , should holds only for c1 = c2 = · · · = cn = 0. We also can write
! !
(1) (2)
X X X X
c1 Xi kij ej + c2 Xi kij ej + · · · +
j i j i
!
(n)
X X X
+cn Xi kij ej = 0ej = 0
j i j
The equation can also be rewritten in the matrix form

P (1) P (2) P (n)  

···
 
i Xi ki1 i Xi ki1 i Xi ki1 c1 0
(1) (2) (n)   
c 0
P P P
 i Xi ki2 X k · · · X k

i i i2 i i i2   
 2
.. .. .. .. ..  =  .. 
  
. . .

. . .
  
 
P (1) P (2) P (n) c 0
i Xi kin i Xi kin ··· i Xi kin
n
8.5. GALOIS ACTION AS A REPRESENTATION. NORMAL BASE THEOREM127
which is not identically zero. 39 Well, why? Because if it was identically

zero, it would remain identically zero also after the base changed to K̄ . 40
Since it has a K̄ point where it does not vanish.
There are many x ∈ L ⊗K K̄ such that gi (x) form a basis. And over
an innite eld, a polynomial which is not identically zero cannot vanish
identically (P 6= 0). And over an innite eld, only a polynomial which is
identically 0 can vanish at every point. 41
Let me to clarify the point. P ∈ K [X] has at most deg P roots. 42 So if
K innite and P has every element of K as a root then P = 0 (P is zero as
an element of K [X]).
By induction we can get the same statement for a polynomial in several
variables. 43 So, our polynomial which is the determinant of the matrix, is
non zero, as a polynomial of several variable because it does not have roots
over algebraic closure K̄ . And so, it also does not have roots over K . So,
The equation has non-trivial solutions only if

P (1) P (2) P (n)

i Xi ki1 Xi ki1 ··· i Xi ki1
P (1) Pi (2) P (n) 
 i Xi ki2 i Xi ki2 ··· i Xi ki2 

(8.6)

det  .. .. ..  6= 0.
..

. . . .
 
 
P (1) P (2) P (n)
i i kin
X i i kin
X ··· X
i i ink
39 The determinant(8.6) can always be written in the form of multi-variable polynomial

P ∈ L [X1 , . . . , Xn ] i.e. the polynomial coecients are from L.
40 The element g ∈ G acts on l ⊗ k ∈ L ⊗ K̄ as follows g (l ⊗ k) = g(l) ⊗ k (see proof
K
for the theorem 4.13). Using proposition 4.11 we have x = X1 e1 ⊗ 1K̄ + . . . Xn en ⊗ 1K̄
and Xi ∈ K̄ . With equation (8.5) we have
(l)
X
gl (ei ⊗ 1K̄ ) = gl (ei ) ⊗ 1K̄ = kij ej ⊗ 1K̄
j
(l)
i.e. all kij are the same as in the (8.6) and as result the polynomial coecients will be
the same (i.e. the polynomial will be identically zero).
41 I.e. polynomial of the following polynomial ring P ∈ L [X , . . . , X ] can vanish at
1 n
every point only if it is identically zero P = 0.
42 Therefore it is impossible to have each k ∈ K as a root of the polynomial because in
the case the number of roots |K| = ∞ > deg P .
43 Let P
n ∈ Kn = K [X1 , X2 , . . . , Xn−1 , Xn ] and we proved the statement for
any Pn−1 ∈ Kn−1 = K [X1 , X2 , . . . , Xn−1 ] i.e. if ∀k1 , k2 , . . . , kn−1 ∈ K we have
Pn−1 (k1 , k2 , . . . , kn−1 ) = 0 then Pn−1 = 0K .
We want to proof that if Pn vanishes for any choice of k1 , . . . , kn ∈ K then it is identically
0. Consider Pn as a polynomial of single variable Xn with coecients from Kn−1 . If
the polynomial is 0 for every kn ∈ K then its coecients are 0. But by the induction
hypothesis they are identically 0 (as soon as they stays zero for any ki choice). Thus the
result polynomial is also identically zero: Pn = 0K
there exists a point x ∈ L (not anymore in L ⊗K K̄ ) such that det(. . . ) 6= 0

at x so gi (x) form a basis.
If K is nite then the argument with roots of a polynomial does not
apply any more. But in the case the Galois group is cyclic i.e. G = hσi (see
corollary 3.17). We have id, σ, . . . , σ n−1 are linearly independent because
they are linearly independent over K̄ . 44 Then the minimal polynomial of
σ as an Endomorphism of L over K is X n − 1. This is because a lower
degree polynomial cannot vanish at σ since, the lower power of σ are linearly
independent. 45 Therefore, as a K [X]-module with X acting by σ , 46 L is
isomorphic to
L∼
= K [X] / (X n − 1) .
47
This is a Cyclic module and any generator x shall do i.e. x, σx, . . . , σ n−1 x
form a basis. 48
8.6 Relation with coverings

Remark 8.23. If L is a nite Galois extension of K then L ⊗K L is a
direct sum of elds 49 which are isomorphic to L. Sums are permuted by
G = Gal (L/K).
Proof. So if L = K (α) is a splitting eld of the polynomial P = (X − α1 ) ·

· · · · (X − αn ) (where α ∈ {α1 , . . . , αn }) that is isomorphic to K [X] /(P ). If
44 See theorem A.98.
45 Pm
σ n = 1 as soon as G = hσi (cyclic group of order n). If ∃m < n such that i=0 ci σ i =
0 and ci 6= 0 then the set 1, σ, . . . , σ m is not linearly independent that is in contradiction
with Dedekind (Theorem A.98) . Thus the minimal polynomial for σ is X n − 1.
46 We consider L as a module over K [X] where X acts on L as σ i.e. if f (X) ∈ K [X]
then f (X) : L −−−−−−−→ L [4], for instance f (X) = X 2 + X + 1 acts on α ∈ L as follows:
α→f (σ)(α)
σ 2 (α) + σ(α) + α.
47 By claim A.108 there is a cyclic K [X]-module. I.e. if x ∈ L is the generator then
L = K [X] x.
48 As soon as X acts as σ on L and σ n − 1 = 0 then any element from K [X] can
be represented as follows k0 + k1 X + · · · + kn−1 X n−1 , where ki ∈ K . Let x ∈ L is the
generator of L as K [X]-module thus one can get
L = k0 + k1 X + · · · + kn−1 X n−1 X=σ x =

= k0 x + k1 σ(x) + · · · + kn−1 σ n−1 (x)
i.e. x, σx, . . . , σ n−1 x is a K basis of L.

49 see also denition A.104
8.6. RELATION WITH COVERINGS 129
we tensor it to L we will get (see examples 4.16 and 5.1)
L ⊗K L ∼
= L ⊗K K [X] / (X − α1 ) · · · · · (X − αn ) ∼ =
∼ ∼
= L [X] / (X − α1 ) · · · · · (X − αn ) =
∼
= L [X] / (X − α1 ) × · · · × L [X] / (X − αn ) ∼ =
∼
= L × ··· × L
that is a product of copies of L permuted by Galois action
In topology one has Galois covering Y → X . G acts on Y , X is a

Quotient of the group action. The covering is characterized by the property
that Y ×X Y = tg∈G Yg (is a Disjoint union), Yg = {(y, gy)}. 50
50 ??? add an explanation

Chapter 9
Ring extensions, norms and
traces, reduction modulo p
We build a tool for nding elements in Galois groups, learning to use the
reduction modulo p. For this, we have to talk a little bit about integral ring
extensions and also about norms and traces.
9.1 Integral elements over a ring

Let P ∈ Z [X]. We want to know what is Gal (P ). Just a reminder that
Gal (P ) = Gal (K/Q) where K is a Splitting eld of P . We have already
done the work for several types of polynomials: n-th cyclotomic polynomials,
Kummer extensions (section 7.2) and so on.
Sometimes, if our polynomial is a kind of combination of them, then the
explicit information about the roots helps to calculate the Galois
√ group.
√ For
instance if we have polynomial X − 2 we know it's roots: 2, j 2, where
5 5 k 5
2πi
j = e 5 , 1 ≤ k ≤ 4. Now we have a lot about the Galois group. If K is the
splitting eld of the polynomial then we have the following towers:
Q (j)
4 5
√
5

Q Q 2, j = K
5 4
√
5

Q 2
131
132CHAPTER 9. RING EXTENSIONS, NORMS AND TRACES, REDUCTION MODULO
From that we know we can conclude that it follows that our Galois group,
contains a normal cyclic subgroup of a order 5: Z/5Z. 1 And then the
quotient is the Galois group of cyclotomic extension 2 , so this is (Z/5Z)× .
So this is a group of order 20. 3 You can show that this is noncommutative,
and from this exact sequence, you have some information about it. But what
will we do if we don't know the roots. One of the tool that we will use is the
reduction of modulo prime and this will be the subject of the lecture.
Example 9.1 (Galois group). 4 Let investigate the Galois group Gal (K/Q)
via GAP:
gap> n:=5;;
gap> i:=GaloisType(x^n-2);;
gap> g:=TransitiveGroup(n,i);;
gap> g;
F(5) = 5:4
gap> Order(g);
20
9.1.1 Ring extensions

Denition 9.2 (Integral element). Let A be an Integral domain, i.e. a ring
without zero divisors and let B is an extension of A. The element α ∈ B is
called integral over A if α is a root of a Monic polynomial P ∈ A [X].
So one can write the following relation
αn + an−1 αn−1 + · · · + a1 α + a0 = 0, ai ∈ A.
1
√ √
We have the following towers: Q ⊂ Q (j) ⊂ Q j, 5 2 = K , where Q (j) and Q j, 5 2
are Galois extensions. By Galois correspondence (Theorem 6.7) one can get that
Gal (K/Q (j)) / Gal (K/Q) = Gal (P ) .
As soon as (see theorem 5.19),
|Gal (K/Q (j))| = [K : Q (j)] = 5,
the Gal(P ) has a normal subgroup of order 5.

2 The quotient is
Gal (K/Q) /Gal (K/Q (j)) .
×
It has order 20/5 = 4 and therefore it is (Z/5Z) . It is also (see claim 8.16) Gal (Q (j) /Q)
3 There is a General ane group: Gal(P ) ∼ GA(1, 5) / S [11, 10]
= 5
9.1. INTEGRAL ELEMENTS OVER A RING 133
√
Example 9.3. 1
2
is not integral element over Z but 2 is an Integral element
over Z.
This is because the polynomial in the denition 9.2 is monic i.e. the
leading coecient is 1.
Lemma 9.4. The following conditions are equvivalent
1. α is integral over A.
2. A [α] is a nitely generated A-module (see denition A.106).
3. A [α] ⊂ C ⊂ B where C is a a nitely generated A-module (see deni-

tion A.106). I.e. A [α] is contained in a nitely generated A-module.
Proof. 1 → 2 → 3 is easy 5 and we will concentrate on 3 → 1.

Let x1 , . . . , xr generate C as A-module then we can write 6
X
αxi = λij xj ,
where λij ∈ A. Consider the matrix Λ = {λij } and let M = α · id − Λ. Then
 
x1
 .. 
M ·  .  = 0.
xr
Thus (see (4.3) at theorem 4.23 proof)
 
x1
 .. 
det M ·  .  = 0.
xr
Therefore det M · C = 0 but 1 ∈ C thus det M = 0. 7 The equation
det M = 0 can be considered as a polynomial with α as a root. 8
5 1 → 2 is really easy because the nite set {1, α, . . . , αn−1 } generates A [α] i.e. ∀x ∈
Pn−1
A [α] , ∃{xi } ⊂ A such that x = i=0 xi αi .
2 → 3 is even more easy because C = A [α] will work.
6 As soon as αx ∈ C is an element of C that can be written as a linear combination
i
of x1 , . . . , xr with coecients from A.
7 There is a question why do we need det M · C = 0, may be we can directly write
det M = 0? Sta provided a good example why it's necessary: Sending generators to zero
is not sucient for having zero determinant. For example, consider a ring A, its ideal
I ⊂ A and A/I as a A-Cyclic module i.e. module with one generator. If x ∈ A is the
generator then we can just say that det M x ∈ I and as result det M not necessary to be
zero.
8 The polynomial will be a Monic polynomial because the matrix element can have α
with coecient equal 1.
9.2 Integral extensions, integral closure, ring of

integers of a number eld
9.2.1 Integral extensions and integral closure
Denition 9.5 (Integral extension). Let A ⊂ B . B is integral over A if
∀α ∈ B , α is an Integral element over A.
The following proposition is not a part of the lectures but it is required
for propositions 9.7 and 9.8 proof.
Proposition 9.6. Let A is a sub-ring of C and α1 , α2 , . . . , αn ∈ C , α1 is an
integral over A, α2 is an integral over A [α1 ], and so on, αn is an integral over
A [α1 , α2 , . . . , αn−1 ] then A [α1 , α2 , . . . , αn ] is a nitely generated A module.
Proof. Lets proof by induction. The case n = 1 follows directly from lemma
9.4.
Induction hypothesis gives us that B = A [α1 , α2 , . . . , αn−1 ] is a nitely
generated A module and we have to prove that S = A [α1 , α2 , . . . , αn ] is a
nitely generated A module.
For B as a nitely generated A module we have
X
B= Abi ,
i
but S = B [αn ] is a nitely generated B module (thanks lemma 9.4) and

therefore X
S= Bsj
j
and as result X X
S= Abi sj = Akij
ij ij
where bi sj = kij ∈ S form a nite set of generators for S .
Proposition 9.7. Let A ⊂ B ⊂ C. B integral over A, C integral over B

then C is an Integral extension over A.
9
Proof. Proof is left as an exercise.
9 Let x ∈ C then
xn + bn−1 xn−1 + · · · + b1 x + b0 = 0,
where bi ∈ B . Thus x is an integral element over A [b0 , b1 , . . . , bn−1 ]. We also have that B
is integral over A therefore bi is an integral over A. As result we have b0 is integral over
A, b1 is integral over A [b0 ], and so on x is integral over A [b0 , b1 , . . . , bn−1 ]. As result with
proposition 9.6 we have that A [b0 , b1 , . . . , bn−1 , x] is a nitely generated A module and as
A [x] ⊂ A [b0 , b1 , . . . , bn−1 , x], the lemma 9.4 gives us that x is an integral element over A.
9.2. INTEGRAL EXTENSIONS, INTEGRAL CLOSURE, RING OF INTEGERS OF A NUMBER F
Proposition 9.8. B is a nitely generated over A as a module (see denition

A.106) if and only if B = A [α1 , . . . , αr ] where each αi is an Integral element
over A.
10
Proof. Proof is left as an exercise
Proposition 9.9. Let A ⊂ B. I.e. B is an arbitrary extension of A. The

elements of B which are integral over A form a subring of B (one calls it as
the integral closure of A in B ).
Proof. Let α, β are integral over A then A [α, β] - nitely generated A-module
(see denition A.106). This follows directly from lemma 9.4. It contains α+β
and αβ and by lemma 9.4 the α + β and αβ are integral over A. But this is
exactly we need to proof. 11
Denition 9.10 (Integrally closed). Let A ⊂ B . A is integrally closed in B

if the integral closure of A in B equals to A.
A is integrally closed (without mention of any B ) if it is integrally closed
in Fraction eld Frac(A).
Example 9.11. Z is Integrally closed. 12
Remark 9.12. More generally any Unique factorization domain is Integrally
closed.
Proof. Let A be a Unique factorization domain and x ∈ Frac(A) such that
x 6= 0. So x = pq such that p, q ∈ A, (p, q) = 1 (this means no common prime
divisor). If x integral over A then
n n−1
p p p
+ an−1 + · · · + a1 + a0 = 0.
q q q
10 If αi is an integral over A then it also an integral over A [α1 , α2 , . . . , αi−1 ] and therefore
by proposition 9.6, A [α1 , α2 , . . . , αr ] = B is nitely
Pgenerated A-module.
r
Let B is nitely generated A module i.e. B = i=1 αi xi , where αi ∈ B - the module
generators and xi ∈ A. If we look at any αi we can notice that A [αi ] ⊂ B - nitely
generated A module and therefore by lemma 9.4, αi is an integral element. Pr We also have
(from denition A.106) that ∀b ∈ B, ∃{xi } ∈ A such that b = i=1 αi xi . Therefore
B = A [α1 , α2 , . . . , αr ].
11 A [α + β] ⊂ A [α, β] and as soon as A [α, β] - netely generated A module then α + β
is an integral element by lemma 9.4.
Same result can be got for αβ from the following inclusion: A [αβ] ⊂ A [α, β]
12 Frac(Z) = Q. An arbitrary element q ∈ Q is an integral element over Z if and
only if q ∈ Z. For example 5 ∈ Z ⊂ Q is integral over Z as a root of a monic polynomial
P = X −5. But 52 ∈ Q is not integral over Z because it is a root of a non-monic polynomial
P = 2X − 5.
Thus
pn + qan pn−1 + q 2 an−1 pn−2 + · · · + q n−1 a1 p + q n a0
=0
qn
therefore q | pn 13 which is in contradiction with (p, q) = 1. As result we have
that q is invertible and therefore x ∈ A. 14
9.2.2 Ring of integers in a number eld

Denition 9.13 (Number eld). Let K is a nite extension of Q i.e. [K : Q] <
∞. In the case K is a number eld.
Let K is a Number eld and [K : Q] = N .
Denition 9.14 (Ring of integers). Let K is a Number eld. The ring of
integers OK ⊂ K is the integral closure of Z in K .
Note: We know that integral closure of Z in Q is Z but now we consider
the closure in K but not in Q.
Property 9.15. 1. ∀α ∈ K, ∃d ∈ Z \ {0} such that dα ∈ OK .
2. If α ∈ OK then Pmin (α, Q) ∈ Z [X].

Proof. For the rst part lets Pmin (α, Q) = X m +am−1 X m−1 +· · ·+a1 X +a0 ∈
Q [X].
∃d ∈ Z (the common denominator) such that ∀i : dai ∈ Z. So bi =
d ai ∈ Z for any i. Therefore
m−i
(dα)m + bm−1 (dα)m−1 + · · · + b0 = 0.
Thus dα ∈ OK .
The second part is also easy. If we have such α ∈ OK , it is a root of some
Monic polynomial Q ∈ Z [X]. Then the Pmin | Q. So Q = Pmin R. If we pick
Pmin to be monic, then by an argument very similar to that of the Gauss
(Equation 1.34) , we conclude that both Pmin , R ∈ Z [X]. 15
13 This is because we have
pn = q −an pn−1 − qan−1 pn−2 − · · · − q n−2 a1 p − q n−1 a0

and
−an pn−1 − qan−1 pn−2 − · · · − q n−2 a1 p − q n−1 a0 ∈ A

14 q −1 ∈ A and x = pq = pq −1 ∈ A.
15 I.e. we can always write
Q = mnP1 R1
9.3. NORM AND TRACE 137
9.3 Norm and trace

9.3.1 Norms and traces
(The material was given inside the proof of theorem 9.21 and can be consid-
ered as a recall. The remarks 9.16 and 9.19 are not parts of the lectures and
were given for better understanding the material)
Remark 9.16 (K embedding of E into the algebraic closure K̄ ). Let K ⊂
E ⊂ K̄ . When we say about K embedding of E into the algebraic closure
K̄ we assume σ ∈ HomK E, K̄ i.e. σ is a Homomorphism of K-algebras
i.e. the map that preserves the structure and especially σ(K) = K i.e.
∀k ∈ K : σ(k) = k .
If E is a normal extension then all such homomorphisms have the same
image accordingly theorem 5.17. The image is E and therefore the homomor-
phisms can be considered as automorphisms i.e. Gal (E/K) = HomK E, K̄ .

Denition 9.17 (Norm). Let K ,→ E - nite separable eld extension. Let

α ∈ E . Dene the norm of alpha with respect to this extension as
Y
NE/K (α) = σi (α)
σi :E,→K̄
i.e. we took a product by all K embeddings of E into the algebraic closure

of K (see remark 9.16). And we also assume that E is nite and as result
i = 1, . . . , r.
where P1 , R1 ∈ Z [X]. Choose p is a prime divisor of mn we can write
P¯1 R̄1 = Q̄ = 0
where Q̄ = Q mod p, P̄1 = P1 mod p, R̄1 = R1 mod p thus P̄1 or R̄1 are equal to 0.
Let R̄1 = 0 thus all coecients of R1 are divided by p i.e. R1 = Rp2 where R2 ∈ Z [X].
Therefore
mn
Q= P1 R2
p
Continue this way we can conclude that
Q = Ps R t ,
where Ps , Rt ∈ Z [X]. As soon as Q is monic then both Ps and Rt are monic. Using the
fact that Ps = zPmin where z ∈ Z we have Ps (α) = 0 and therefore we can conclude that
Ps is the minimal polynomial.
Denition 9.18 (Trace). Let K ,→ E - nite separable eld extension. Let

α ∈ E . Dene the norm of alpha with respect to this extension as
X
TrE/K (α) = σi (α)
σi :E,→K̄
i.e. we took a sum by all K embeddings of E into the algebraic closure of

K (see remark 9.16). And we also assume that E is nite and as result
i = 1, . . . , r.
In the denitions 9.17 and 9.18 we assume that the extension E is Sepa-
rable extension. If the extension is not separable then you have to take it to
the power equal to the pure inseparable degree of E/K , 16 but for simplicity,
we shall suppose that everything is separate.
Remark 9.19. We have NE/K (α) ∈ K and TrE/K (α) ∈ K
Let g ∈ Gal (E/K). We have to prove that g NE/K (α) = NE/K (α).

Proof.
In the case NE/K (α) ∈ K because E Gal(E/K) = K .
The just permutes the homomorphisms i.e.

g Hom K E, K̄ gHomK E, K̄ =
HomK E, K̄ . If this is not the truth then ∃σi , σj , σk ∈ HomK E, K̄ such

that gσi = σk , gσj = σk and σi 6= σj . Therefore
σj = g −1 σk = g −1 gσi = σi
that is contradiction. By the eld denition (A.91) the product in Norm

does not depend on the order and the result become the same after the
permutation.
The same result is for Trace.
Property 9.20. 1. NE/K : E × → K × 17 is multiplicative i.e. homomor-

phism of groups. TrE/K : E → K is additive, K -linear i.e. homomor-
18
phism of K -vector spaces.
16 [14] p. 284 gives the following denitions for norm and trace in not separable case:
 [E:K]i
Y
NE/K (α) =  σi (α) ,
σi :E,→K̄
X
TrE/K (α) = [E : K]i σi (α) ,
σi :E,→K̄
where [E : K]i is Inseparable degree.

17 E × = E \ {0} and K × = K \ {0}
18 The property statement is the truth i.e. there really should be K (not K̄)(see remark
9.19).
2. If E = K (α), n = [E : K] and Pmin (α, K) = X n + a1 X n−1 + · · · +

an−1 X + an then NE/K (α) = (−1)n an and TrE/K (α) = −a1 .
3. If we have the tower of extensions K⊂F ⊂E then
NE/K = NF/K ◦ NE/F

and the same for trace
TrE/K = TrF/K ◦ TrE/F

19
4. Consider T : E × E −−−−−−−−−−→ K . This is a non-degenerate K-

(x,y)→TrE/K (xy)
bilinear form (see denition A.120)
20
5. α integral over Z. Then NE/Q (α) , TrE/Q (α) are integers.
21
Proof. The rst property is obvious from the denition.
19 Or in other words
NE/K (α) = NF/K NE/F (α)
and
TrE/K (α) = TrF/K TrE/F (α) .
Lets also note [7] that the following expression does not make sense:

TrE/F TrF/K (α)
because TrE/F : E → F and TrF/K : F → K (see remark 9.19). The same is valid for
Norm.
20 We have K = Q in the property.
21 Let α, β ∈ E then
Y
NE/K (αβ) = σi (αβ) =
σi :E,→K̄
Y
= σi (α) σi (β) = NE/K (α) NE/K (β) .
σi :E,→K̄
With remark 9.19 we can get that
NE/K : E × → K ×
If we take a, b ∈ K, σ ∈ HomK E, K̄ then one can get that σ (aα) = aσ (α). This is

because a is xed under every embedding of E over K ([14] page 286), see also remark
9.16.
Therefore one can get
TrE/K (aα + bβ) = aTrE/K (α) + bTrE/K (β)
I.e. there is a K -linear map of E to K (see remark 9.19).

The second one uses the following fact: σi (α) are roots of Pmin (α, K).
The Norm is a product and it's assigned to its constant term (an ) and the
sum is the rst coecient term (a1 ) (see also example 8.19).
The third property is somewhat less trivial, so this follows from, the
fact that if τ1 , . . . , τk are K embeddings of F into K̄ and, µ1 , . . . , µs are
F embeddings of E into K̄ then the embedings of E into K̄ are just the
compositions {τj µi }. 22
For the 4th property. Indeed if x ∈ ker T , 23 that means TrE/K (xy) =
0, ∀y ∈ E (see denition A.120), but this can't be a case when xy ∈ K \ {0}
by denition 9.18 24 TrE/K (xy) = [E : K] xy . 25
For the 5th property we know that

TrE/Q (α) = TrQ(α)/Q TrE/Q(α) (α) =
= TrQ(α)/Q ([E : Q(α)] α) = [E : Q(α)] TrQ(α)/Q (α)
but TrQ(α)/Q (α) ∈ Z because TrQ(α)/Q (α) is a coecient of Pmin (α, Q) ∈
Z [X]. 26
Why such names are used? Consider the following map (multiplication
by a)
fa : E −−−→ E
x→ax
then the TrE/K (a) is exactly the trace of the linear map (i.e. sum of diag-
onal elements of the linear map matrix in a basis) and the NE/K (a) is the
determinant 27 . Now this fa is a a K -linear map. It's an Endomorphism of
a vector space E/K , and the TrE/K (a) is the trace of this endomorphism,
and the NE/K (a) is the determinant of this endomorphism.
22 We have K ⊂ F ⊂ K̄ , and (see remark 9.16):
(
τj : F → K̄ such that ∀k ∈ K : τj (k) = k,
µi : E → K̄ such that ∀f ∈ F : µi (f ) = f.
Lets extend homomorphism τj : F → K̄ to the automorphism K̄ → K̄ and denote the

result with the same τj ([14] p. 285). As result we have τj µi : E → K̄ . We also have
∀k ∈ K ⊂ F µi (k) = k (as soon as k ∈ F ) and therefore τj (µi (k)) = τj (k) = k . Thus
τj µi ∈ HomK E, K̄ .
23 T : E × E → K
24 and by taking into consideration the following fact (see remark 9.16): if xy ∈ K then
σi (xy) = xy
25 we proved that T has a trivial kernel i.e. only x = 0 is in the ker T . There is one of
denitions of the non-degenerate K -bilinear form (see denition A.120)
26 Property 9.15 says that P
min (α, Q) ∈ Z [X]. But 2d item of the property 9.20 says
that the Trace is a coecient of the polynomial i.e. TrQ(α)/Q (α) ∈ Z
27 Consider an easy case when E = K (a). In the case a is a root of P
min (a, K) =
an + an−1 X + · · · + a1 X n−1 + X n . We have the following basis 1, a, . . . , an−1 . The basis
9.3.2 Theorem about rings of integers

Theorem 9.21. Ok is a nitely generated (see denition A.106) Z-module
that is a Free module of rank (see denition A.103) n, where n = [K : Q].
Proof. If e1 , . . . , en is a Q-basis of K then ∀i∃di ∈ Z\{0} such that di ei ∈ OK

(see property 9.15). Therefore OK contains a free Z-submodule of rank n 28 .
What is the Z-module this is a nitely generated Finitely generated
abelian group and we know a lot of things about such groups. The Finitely
generated abelian group is the same as nitely generated Z-module. Any
is transformed by multiplication via the following rules

1 → a,


a → a 2 ,

..


 .

 n−1
a → an = −an − an−1 a − · · · − a1 an−1 .
Therefore the Endomorphism matrix can be written as follows:

 
0 1 0 ··· 0 0
 0 0 1 ··· 0 0 
 .. . . . . .. 
 
M = . .. .. .. .. . 
 
 0 0 0 ··· 0 1 
−an −an−1 −an−2 · · · −a2 −a1
It can be easy seen (with 2d item of the property 9.20) that
Tr (M ) = −a1 = TrE/K (a)
and
det (M ) = (−1)n an = NE/K (a) .
28 This is because d1 e1 , . . . , dn en are linearly independent and form a basis of a free

Z-module. The number n is the cardinality of the basis.
It can be proved by contradiction i.e. let there exists a set {ci } such that ∃j : cj 6= 0
and
Xn
ci di ei = 0,
i=1
as soon as dj 6= 0 then kj = cj dj 6= 0 and

n
X
ki ei = 0,
i=1
i.e. {ei } are not linearly independent. This is in contradiction with the initial conditions.
such group is isomorphic to (see theorem A.47)
Zn ⊕ A,
where A is a nite group (torsion part). A subgroup B ⊂ Zn is itself a free

module (B ∼= Zm ) of rank m ≤ n.
We have to show that OK ⊂ A where A is a free Z-submodule of rank
n = [K : Q]. Let e1 , . . . , en is a Q-basis of K (as above) contained in OK .
Consider the following map (T : K × K → Q):
(x, y) → TrK/Q (xy)
this is Non-degenerate bilinear form (see 4th property 9.20) therefore ∃v1 , . . . , vn
- Dual basis (Q-basis of K ) and we have the property that TrK/Q (ei vj ) = δij .
29
We claim that Z submoduleP generated by v1 , . . . , vn contains OK . Indeed

let α ∈ OK and write α = αi vi , αi ∈ Q. We can do it because {vi } is a
Q basis of K . But one can see that αi ∈ Z because αi = TrK/Q (αei ) (by
denition of vj ). 30 Since α and ei are elements of OK then αei ∈ OK too.
Therefore TrK/Q (αei ) ∈ Z. So αi ∈ Z and this one is what we want to proof.
We have expressed any element of OK as a combination of vi with integral
coecients. So OK is contained in a Z submodule, generated by {vi }.
29
Lets consider the following map T : K × K → Q. We can use it to construct a
linear map by the following rule: ∀x ∈ K we have fx (y) = T (x, y) : K → Q. We have
fx+y = fx + fy and fax = afx i.e. the set {fx } = K ∗ is the linear space. The map
K −−−−→ K ∗ is Surjection by the fx construction. The map is also Injection as soon as
x→fx
the map T : K × K → Q is Non-degenerate bilinear form and as result the fx (y) = T (x, y)
has a trivial kernel (if x 6= 0): ker f = {0}. Therefore we can conclude that K ∗ is Dual
space. I.e. exists a set of elements {fvj } ⊂ K ∗ which form the Dual basis to {ei } i.e.
fvj (ei ) = δij . Each element fvj of dual space corresponds to vj ∈ K . For such vj we have:
δij = fvj (ei ) = T (vj , ei ) = TrK/Q (vj ei ) .
30
 
Xn
TrK/Q (αei ) = TrK/Q  αj vj ei  =
j=1
n
X n
X
= αj TrK/Q (vj ei ) = αj δij = αi
j=1 j=1
9.4. REDUCTION MODULO A PRIME 143
9.4 Reduction modulo a prime

Let P ∈ Z [X] is an irreducible polynomial with integer coecients. K is a
def
Splitting eld of P over Q and n = [K : Q]. Let G = Gal (P ) == Gal (K/Q).
We denote roots of P as α1 , . . . , αn and they are elements of OK . G acts on
the set of roots, and on OK . We will denote OK as A. Let p is a prime
number and we will consider A/pA. As we have seen 31
A/pA ∼
= A ⊗Z Z/pZ = A ⊗Z Fp
there A ⊗Z Fp is a n-dimension vector space over Fp 32 . Maximal ideals of

A/pA are in one-to-one correspondence with maximal ideals of A containing
p. As we know (see theorem 4.23) there are only nitely many maximal ideal
in a nite algebra over a eld. Therefore A also has nitly many maximal
ideals J1 , . . . , Jr containing p. Our group G acting on A must permute these
maximal ideals in some way. 33
Lets consider a subgroup Di ⊂ G which stabilizes Ji (see denition A.29)
i.e.
Di = {g ∈ G | gJi = Ji }.
Let also ki = A/Ji - this is a eld 34 and there is a nite extension of Fp .
35
Then there exists a natural homomorphism Di → Gal (ki /Fp ). Since Di
stabilizes Ji and it acts on the residual classes of modulo Ji so there is a
homomorphism of Di into the Galois group. 36
Theorem 9.22. 1. G acts transitively (see denition A.31) on {J1 , . . . , Jr }

and the map Di → Gal (ki /Fp ) is a Surjection i. e. Di Gal (ki /Fp )
31 see proposition 4.14 where M = A, A = Z and I = pZ.
32 As soon as A = OK is a free Z-module of rank n (see theorem 9.21) then proposition
4.11 gives us that A ⊗Z Fp is a free Fp -module with rank equal to n.
33 Let J is an Ideal of A, then AJ = J (i.e. ∀a ∈ A, j ∈ J : aj ∈ J ). Let also g ∈ G
then g (A) = A (see [14] p. 341) and
Ag (J) = g (A) g (J) = g (AJ) = g (J)
i.e. g (J) is also an ideal of A.

34 see theorem A.94
35 We have that A is a F -algebra. J is a Maximal ideal of A then by remark 4.22 we
p i
have that ki = A/Ji is also Fp -algebra i.e. ki is a eld extension of Fp (see denition 1.4).
The extension is nite as soon as A is nite.
36 We can notice that both k and D are related to J because any g (k) ∈ Gal (k /F ) is
i i i i p
an automorphism of ki i.e. it should preserve Oki = Ji (as soon as it is an injection). Or in
other words any element of Gal (ki /Fp ) acts in the same way as elements of Di (preserves
Ji ). See also theorem 9.22 about the map structure (it's surjection and in several situations
is also bijection).
2. If the reduction P̄ = P mod p has no multiple roots then the map

Di → Gal (ki /Fp ) is bijection and ki is a splitting eld of the reduction
P̄ .
Proof. For the rst part. Suppose that for some i and ∀g ∈ G, g (J1 ) 6= Ji i.e.
suppose that there is not a Transitive group action. By Chinese remainder
(Theorem 4.20) ∃x ∈ A such that x ≡ 0( mod Ji ), x ≡ 1( mod g (J1 ))∀g ∈
G. 37 Consider a product of all such things:
Y
a= gx
g
it's an integer a ∈ Z. 38 But since x ∈ Ji (remember that ≡ 0 mod Ji ) then

a is also in Ji : 39 a ∈ Z ∩ Ji = (p) 40 - the ideal generated by the prime
number p. So one has a ∈ J1 since all Ji , and especially J1 , contains p. But
this is impossible as soon as J1 is a Prime ideal (see note 40). That is because
if we have k xk ∈ J1 (xk = gk (x)) then ∃i such that xi ∈ J1 but there is not
Q
a case in our construction. 41
We still need to proof that Di Gal (ki /Fp ). 42 We may assume that
i = 1. By the Primitive element (Theorem 5.11) ∃z ∈ Fp such that
k1 = Fp (z) i.e. z generates ki /Fp . By Chinese remainder (Theorem 4.20)
∃y ∈ Q A such that y ∈ Ji , i 6= 1, y ≡ z( mod J1 ). 43 Consider polynomial
Q = g∈G (X − g (y)). There is a polynomial with integral coecients i.e.
Q ∈ Z [X]. This is because we know that coecients are G invariant i.e.
Q ∈ Q [X] 44 moreover the roots are Integral elements over Z 45 and therefore
37 We have that all Ji are relatively prime (see note 40 below) and as result the map
π (a) from (4.2) is Surjection i.e. for any residuals and therefore for the chosen ones (≡ 0
mod Ji and ≡ 1 mod J1 ) exists x that produces the residuals.
38 This is because Q gx = N
g A/Q (x), but by 5th item of property 9.20 one can get that
such Norm (in the denition of norm we have homomorphism but not automorphism but
accordingly remark 9.16 it is not important as soon as A is a normal extension of Q) is an
integer. Q
39 a =
g6=id gx and by ideal denition A.69 we can conclude that a ∈ Ji .
Q
g gx = x
40 We have J is a Maximal ideal and A is a Principal ideal domain. Therefore by lemma
i
A.80, Ji is a Prime ideal that contains p ∈ Z, but the prime ideal in Z that contains p is
(p) (see lemma A.79).
41 We have x ∈ J and x ∈ J but x is obtained from x by applying an element from
i i 1 i
G i.e. ∃gi ∈ G such that xi = gi (x). But if we consider a reverse element: g = gi−1 then
g(xi ) = x ∈ Ji that is in contradiction with g(J1 ) 6= Ji .
42 i.e. that there is a Surjection or in other words that ∀ḡ ∈ Gal (k /F ) , ∃d ∈ D such
i i p i i
that di → ḡi .
43 I.e. using the same argument as in note 37 one can conclude that ∃y ∈ A such that
its residual are the following ∀i 6= 1: ≡ 0 mod Ji and ≡ z mod J1 .
44 Q = P
min (y, Q).
45 as soon as y ∈ A = O .
K
9.5. FINDING ELEMENTS IN GALOIS GROUPS 145
the polynomial arguments are in Z. 46

47
Lets study Q̄ = Q mod J1 . If g ∈ / D1 then ∃i such that g (Ji ) = J1
and particularly g(y) ∈ J1 . Therefore for such g we have
X − g (y) = X − g (y) mod J1 = X.
So we have for Q̄ ∈ Fp [X]
Y Y
Q̄ = X X − g (y) ,
g∈G\D1 g∈D1

48
but g∈D1 X − g (y) has z as a root and D1 acts transitively on its
Q
roots.
Now recall that z generates k1 (i.e. k1 = Fp (z)). Thus an element of
Gal (k1 /Fp ) is determined by the image of z . 49 And we have an element of
D1 which sends z to any possible image of it. But this means that D1
Gal (k1 /Fp )
For the second part of the theorem we assume that P̄ has no multiple
roots. So α1 , . . . , αn -roots of P and ᾱ1 , . . . , ᾱn -roots of P̄ where ᾱi = αi
mod J1 .
Lets g ∈ D1 acts as id on k1 . Then, of course, g (ᾱi ) = ᾱi . But g (αi ) ∈
{α1 , . . . , αn } and it can not be dierent from αi since they will have dieent
reduction mod J1 . So ∀i, g (αi ) = αi and therefore g = id. 50 Thus
conclusion that D1 ∼ = Gal (k1 /Fp ).
By the same argument 51
Gal (k1 /Fp [ᾱ1 , . . . , ᾱn ]) = id
therefore k1 = Fp [ᾱ1 , . . . , ᾱn ] i.e. k1 is a splitting eld.
9.5 Finding elements in Galois groups

How can we apply the above material to study Galois groups? One uses this
theorem to construct elements of a certain type in the Galois group to show
that the Galois group is large.
46 Ekaterina also said that ??? as soon as Z is Integrally closed.
47 If g ∈
/ D1 then g −1 ∈
/ D1 i.e. g −1 (J1 ) 6= J1 i.e. as soon as g ∈ G permutes the ideals
(see note 33) then ∃i such that g −1 (J1 ) = Ji i.e. g (Ji ) = J1 .
48 Let g = id ∈ D then g (y) = y mod J = z
1 1
49 Because we have bijection g → g(z).
50 This proves injection of the map D → Gal (k /F ) (see lecture 6 note 42).
1 1 p
51 i.e. ∀g ∈ Gal (k /F [ᾱ , . . . , ᾱ ]) and ∀x ∈ F [ᾱ , . . . , ᾱ ] we have g(x) = x i.e.
1 p 1 n p 1 n
g = id.
So let P ∈ Z [X] be an irreducible polynomial and suppose that there is

a prime p ∈ Z such that P̄ = P mod p is also irreducible. Then Gal (P )
contains a subgroup that is isomorphic to Gal P̄ . 52 But we know Galois

group of nite elds and we conclude that this Galois group contains an n
cycle. This is because Gal P̄ is cyclic generated by n cycle (see corollary

3.17).
Sometimes, there is no such prime, but of course, a variant of this argu-
ment exists also in other cases. Suppose, for instance that P is irreducible
of degree 5 and that P̄ = R2 R3 where Ri is irreducible of degree i. Then the
same argument, gives that Gal (P ) contains the permutation (1, 2)(3, 4, 5).
53
And in this way one can construct elements of particular type in the
Galois group and use this to show that those groups are very large.
52
See theorem 9.22 there Di ∼ = Gal (ki /Fp )
53
The Gal (P ) contains a subgroup of 2-cycle (associated with the rst 2 roots) and a
subgroup of 3-cycle (associated with the last 3 roots).
Appendices
147
Appendix A
Course prerequisites
There are several prerequisites for the course there. They consists of deni-
tions, theorems and examples mostly taken from Wikipedia.
A.1 Sets
Denition A.1 (Class). A class is a collection of sets (or sometimes other
mathematical objects) that can be unambiguously dened by a property that
all its members share.
Denition A.2 (Disjoint union). Let [62] {Ai : i ∈ I} be a family of sets

indexed by I . The disjoint union of this family is the set
ti∈I Ai = ∪i∈I {(x, i) : x ∈ Ai } .
The elements of the disjoint union are ordered pairs (x, i). Here i serves as
an auxiliary index that indicates which Ai the element x came from.
Example A.3 (Disjoint union). Let [62] we have 2 sets A0 = {1, 2, 3} and
A1 = {1, 2}. We can construct the following sets of pairs
A∗0 = {(1, 0) , (2, 0) , (3, 0)} ,

A∗1 = {(1, 1) , (2, 1)}
so
A0 t A1 = A∗0 ∪ A∗1 = {(1, 0) , (2, 0) , (3, 0) , (1, 1) , (2, 1)}
149
150 APPENDIX A. COURSE PREREQUISITES
Table A.1: Cayley table for Z/2Z

◦ 0 1
0 0 1
1 1 0
A.2 Groups
Denition A.4 (Monoid). The set of elements M with dened binary op-
eration ◦ we will call as a monoid if the following conditions are satised.
1. Closure: ∀a, b ∈ M : a ◦ b ∈ M
2. Associativity: ∀a, b, c ∈ M : a ◦ (b ◦ c) = (a ◦ b) ◦ c
3. Identity element: ∃e ∈ M such that ∀a ∈ M : e ◦ a = a ◦ e = a
Denition A.5 (Group). Let we have a set of elements G with a dened

binary operation ◦ that satised the following properties.
1. Closure: ∀a, b ∈ G: a ◦ b ∈ G
2. Associativity: ∀a, b, c ∈ G: a ◦ (b ◦ c) = (a ◦ b) ◦ c
3. Identity element: ∃e ∈ G such that ∀a ∈ G: e ◦ a = a ◦ e = a
4. Inverse element: ∀a ∈ G ∃a−1 ∈ G such that a ◦ a−1 = e
In this case (G, ◦) is called as group.
Therefore the group is a Monoid with inverse element property.
Example A.6 (Group Z/2Z). Consider a set of 2 elements: G = {0, 1} with

the operation ◦ dened by the table A.1.
0 i.e. e = 0.
The identity element is Inverse element is the element itself
because ∀a ∈ G: a ◦ a = 0 = e.
See also example A.17
Denition A.7 (Order of element in group). Order, sometimes period, of

an element a of a group is the smallest positive integer m such that am = e
(where e denotes the identity element of the group, and am denotes the
product of m copies of a). If no such m exists, a is said to have innite order.
A.2. GROUPS 151
Theorem A.8 (Lagrange). For any nite group G, the order (number of
elements) of every subgroup H of G divides the order of G.
Theorem A.9 (Finite group of prime order). Let G is a nite group of prime
order i.e. |G| = p, where p is prime number. Then G is a Cyclic group.
Proof. Let g ∈ G such that g 6= e then, by theorem A.8, we have |hgi| | |G|.
Thus |hgi| = p because |G| = p and p is a prime number. As result we have
G = hgi and therefore G is a cyclic group.
Denition A.10 (Subgroup). Let we have a Group (G, ◦). The subset
S ⊂ G is called as subgroup if (S, ◦) is a Group.
Denition A.11 (Proper subgroup). A proper subgroup of a group G is a

Subgroup H which is a proper subset of G (i.e. H 6= G) [57]
Denition A.12 (Coset). If G is a group, and H is a subgroup of G, and g

is an element of G, then
gH = {gh|h ∈ H}
is the left coset of H in G with respect to g , and
Hg = {hg|h ∈ H}
is the right coset of H in G with respect to g .
Denition A.13 (Normal subgroup). A subgroup, N , of a group G, is called

a normal subgroup if it is invariant under conjugation i.e.
N / G ⇔ ∀n ∈ N, ∀g ∈ G, gng −1 ∈ N
The denition taken from [51]

We also can write normality in the Coset notation as follows. N / G if
∀g ∈ G : gN = N g .
Denition A.14 (Simple group). A simple group is a nontrivial group whose

only Normal subgroups are the trivial group and the group itself.
The denition taken from [56]
Denition A.15 (Quotient group). A quotient group or factor group is

a mathematical group obtained by aggregating similar elements of a larger
group using an equivalence relation that preserves the group structure. For
example, the Cyclic group of addition modulo n can be obtained from the
integers by identifying elements that dier by a multiple of n and dening
a group structure that operates on each such class (known as a congruence
class) as a single entity. It is part of the mathematical eld known as group

theory [53].
In a quotient of a group, the equivalence class of the identity element is
always a normal subgroup of the original group, and the other equivalence
classes are precisely the cosets of that normal subgroup. The resulting quo-
tient is written G/N , where G is the original group and N is the normal
subgroup.
In other words the quotient group can be dened as a set of all left Cosets
(they also equal to the right cosets as soon as N is a Normal subgroup):
G/N = {aN : a ∈ G}
See also example S3 /A3 quotient group (Example A.57) .
Theorem A.16 (Quotient group). If G is a group and H / G then the

operation aH · bH = a · H makes G/H a group with identity H and inverse
−1
element (aH) = a−1 H .
Proof. See [68] p. 33.
Example A.17 (Quotient group). Consider [53] a group of integers Z (under

addition) and the subgroup 2Z of all even integers. This is a normal subgroup,
because Z is Abelian group. There are only two Cosets: the set of even
integers and the set of odd integers; therefore, the quotient group Z/2Z is the
Cyclic group with two elements. This quotient group is isomorphic with the
set {0, 1} with addition modulo 2; informally, it is sometimes said that Z/2Z
equals the set {0, 1} with addition modulo 2.
See also example A.6.
Theorem A.18 (Correspondence theorem). The correspondence theorem,

sometimes referred to as the fourth isomorphism theorem or the lattice the-
orem, states that if N G, then there exists
is a Normal subgroup of a group
a bijection from the set of all subgroupsA of G containing N , onto the set
of all subgroups of the quotient group G/N . The structure of the subgroups
of G/N is exactly the same as the structure of the subgroups of G containing
N , with N collapsed to the identity element [26].
Denition A.19 (Commutator). The commutator of two elements, g and

h, of a group G, is the element [24]
[g, h] = g −1 h−1 gh
A.2. GROUPS 153
Denition A.20 (Commutator subgroup). The commutator subgroup or

derived subgroup of a group is the subgroup generated by all the Commuta-
tors of the group [25].
Theorem A.21 (About quotient group and commutator subgroup). Given a
group G a Quotient group G/N is an Abelian group if and only if N ⊇ [G, G]
[25]
Denition A.22 (Abelianization). The quotient G/[G, G] is an abelian

group (as it follows from theorem A.21) called the abelianization of G or
G made abelian. [25]
A.2.1 Cyclic group

Denition A.23 (Cyclic group). A cyclic group or monogenous group is a
group that is generated by a single element. Note that Group Z/2Z (Ta-
ble A.6) is a cyclic group. See example A.68.
Theorem A.24 (Fundamental theorem of cyclic groups). In abstract alge-
bra, every subgroup of a Cyclic group is cyclic. Moreover, for a nite cyclic
group of order n, every Subgroup's order is a divisor of n, and there is exactly
one subgroup for each divisor.This result has been called the fundamental the-
orem of cyclic groups [58]
Theorem A.25 (About subgroups of a cyclic group). Let [19] G = hai be a

cyclic group.
1. Every subgroup S of G is cyclic
2. If |G| = n, then G has a unique subgroup of order d for each divisor d

of n
See example A.68
A.2.2 Group action

Denition A.26 (Action). An action of a group is a way of interpreting the
elements of the group as "acting" on some space in a way that preserves the
structure of that space. See also [43].
Denition A.27 (Orbit). Consider [43] a group G acting on a set X . The
orbit of an element x ∈ X is the set of elements in X to which x can be
moved by the elements of G:
Orb (x) = {y ∈ X : ∃g ∈ G : y = g · x}
The orbit of element x is also denoted as G (x).
Denition A.28 (Fixed point). The set of points of X xed by a group

action are called the group's set of xed points, dened by
{x : gx = x, ∀g ∈ G} .
see also [18].
Denition A.29 (Stabilizer subgroup). For every x in X , we dene [43] the
stabilizer subgroup of G with respect to x (also called the isotropy group) as
the set of all elements in G that x x:
Gx = {g ∈ G | g · x = x}
Theorem A.30 (Orbit-stabilizer theorem). If group G and the set the group
acting X are nite then
|G| = |Orb (x)| |Gx |

where x ∈ X , G (x) - is the Orbit, Gx - Stabilizer subgroup.
Note: the result was got from [43] as orbit-stabilizer theorem + Lagrange
(Theorem A.8)
Denition A.31 (Transitive group action). The action of G on X is called

[43] transitive if X is non-empty and if for each pair x, y ∈ X there exists a
g ∈ G such that gx = y .
Denition A.32 (Free group action). The action of G on X is called [43]
free if, given g, h ∈ G, the existence of an x ∈ X with g(x) = h(x) implies
g = h.
Denition A.33 (Quotient of the group action). Consider [43] a group G
acting on a set X . The set of all Orbits of X under the action of G is written
as X/G, and is called the quotient of the action.
A.2.3 Direct product

Denition A.34 (Direct product). Given groups G and H , the direct prod-
uct G × H is dened as follows:
The underlying set is the Cartesian product, G × H . That is, the ordered
pairs (g, h), where g ∈ G and h ∈ H . The binary operation on G × H is
dened component-wise:
(g1 , h1 ) (g2 , h2 ) = (g1 · g2 , h1 ◦ h2 )
The resulting algebraic object satises the axioms for a group. See [30].
The direct product of 2 Abelian groups is also called Direct sum
A.2. GROUPS 155
Property A.35 (Direct product of groups). Let G1 , G2 - Groups and G=

G1 × G2 - Direct product of the groups. Then
1. G1 ∼
= (G1 , 1G2 ) and G2 ∼
= (1G1 , G2 )
2. G1 and G2 are Normal subgroups in G
See [30].
Property A.36 (Quotient of direct product). Let G1 , G2 - Groups and

G = G1 × G2 - Direct product of the groups. Then
G/G1 ∼
= G2
and
G/G2 ∼
= G1
Proof.Lets prove the rst claim: G/G1 ∼

= G2 (the second one is analogous).
Consider projection π : G −−−−−−−−−−→ G2 . The π is Homomorphism.
(g1 ,g2 )→(1G1 ,g2 )
Really let a = (a1 , a2 ), b = (b1 , b2 ) ∈ G where a1 , b1 ∈ G1 , a2 , b2 ∈ G2 :
π (a · b) = π ((a1 , a2 ) · (b1 , b2 )) =
= π ((a1 b1 , a2 b2 )) = a2 b2 = π (a) π (b)
As result π is a Homomorphism.
The π is also a Surjection because ∀a2 ∈ G2 ∃a ∈ G such that π(a) = a2 .
Really we can use a = (1G1 , a2 ).
Therefore by First isomorphism (Theorem A.134) one can get
G/ ker π ∼
= G2 ,
but ∀g1 ∈ G1 we have (g1 , 1G2 ) ∈ ker π . And conversely ∀k ∈ ker π we have
k = (k1 , k2 ) where k1 ∈ G1 and k2 = 1G2 . As result (see property A.35)
ker π = (G1 , 1G2 ) ∼

= G1 .
Therefore
G/G1 ∼
= G2 .
A.2.4 Sylow theorems

Corollary A.37 (Sylow). Given a nite group G and a prime number p
dividing the order of G, then there exists an element (and hence a subgroup)
of order p in G [59]
A.2.5 Abelian group

Denition A.38 (Abelian group). Let we have a Group (G, ◦). The group
is called an Abelian or commutative if ∀a, b ∈ G it holds a ◦ b = b ◦ a.
Theorem A.39 (About order of element of an Abelian group). If G is a

nite Abelian group and m is the maximal order of the elements of G then
the order of every element of G divides m
Theorem A.40. Let G is an Abelian group and n = |G| the group order
(number of elements) then ∀g ∈ G the following statement holds
g n = e,
there e is the group identity.
Proof. Let m is the maximal order of group G. In this case by Lagrange

(Theorem A.8) m | n i. e. n = k1 m where k1 ∈ Z. Let l is the order of g
i.e. g l = e. By the theorem A.39 l | m i.e. m = k2 l. Thus
k2 k1
g n = (g m )k1 = g l = e.
Theorem A.41 (Cyclic Group is Abelian). Let G be a Cyclic group then G

is Abelian group
Proof. As soon as G = hgi then ∀x, y ∈ G, ∃n, m ∈ N such that x = g n , y =

g . In the case
m
xy = g n g m = g n+m = g m g n = yx,
i.e. G is abelian.
Remark A.42 (Not every Abelian group is cyclic). The theorem statement
cannot be reversed i.e. there exist abelian group that are not cyclic. For
example well known Klein four group V4 that is an Abelian group but not
Cyclic group
A.2. GROUPS 157
Table A.2: Cayley table for V4

◦ e a b c
e e a b c
a a e c b
b b c e a
c c b a e
Denition A.43 (Klein four group). The Klein four group V4 is a group of
1
order 4 The Cayley table for the group is on the table A.2 [65].
Example A.44 (Klein four group). There are several examples of V4 :
• V4 ∼
= Z/2Z ⊕ Z/2Z
• V4 ∼
= (Z/8Z) ∼
×
= (Z/12Z)
×
Denition A.45 (Direct sum). The direct sum of two abelian groups A and
B is another abelian group A ⊕ B consisting of the ordered pairs (a, b) where
a ∈ A and b ∈ B . [31] See also Direct product.
Denition A.46 (Finitely generated abelian group). An Abelian group

(G, +) is called nitely generated [38] if there exist nitely many elements
x1 , . . . , xs in G such that ∀x ∈ G:
x = n1 x1 + · · · + ns xs (A.1)
with ni ∈ Z. In this case we say that {x1 , . . . , xs } is a generating set of G.

In the (A.1) we have the following:
n i x i = x i + . . . xi
| {z }
ni times
Theorem A.47 (The fundamental theorem of nitely generated abelian

groups). Every Finitely generated abelian group G is isomorphic to a Direct
sum of primary cyclic groups and innite cyclic groups. A primary cyclic
group is one whose order is a power of a prime. That is, every nitely
generated abelian group is isomorphic to a group of the form
Zn ⊕ Zq1 ⊕ · · · ⊕ Zqt
1 there are only 2 groups of order 4: V4 and Z/4Z (the only cyclic group of order 4)
where the rank n ≥ 0, and the numbers q1 , . . . , qt are powers of (not nec-
essarily distinct) prime numbers. In particular, G is nite if and only if
n = 0. The values of n, q1 , . . . , qt are (up to rearranging the indices) uniquely
determined by G. The statement was took from [38].
Theorem A.48 (Simple subgroup of an abelian group). Every non-simple

abelian group has a simple normal subgroup. ??? add link ???
A.3 Permutations
Example A.49 (Permutation). The following permutation
1→2
2→5
π= 3→4
4→3
5→1
can be also written in dierent forms. The most common one is the following:

1 2 3 4 5
π= .
2 5 4 3 1
In the permutation we can see 2 cycles: 1 → 2 → 5 → 1 and 3 → 4 → 3. The
rst cycle can be written as (1, 2, 5) (or (5, 1, 2) or (2, 5, 1)) and the second
one as (3, 4) (or (4, 3)). The cycles gives us the shortest form of writing the
permutation:
π = (1, 2, 5)(3, 4) = (3, 4)(5, 1, 2).
If we have 2 permutations

1 2 3 4 5
π1 = .
3 1 5 2 4
and
1 2 3 4 5
π2 = .
2 5 4 3 1
then we can combine them into the new one (via a multiplication)
1→2→1
2→5→4
1 2 3 4 5
π = π1 π2 = 3 → 4 → 2 = = (1)(2, 4, 5, 3) = (2, 4, 5, 3)
1 4 2 5 3
4→3→5
5→1→3
A.3. PERMUTATIONS 159
We have identity element

1 2 3 4 5
e=
1 2 3 4 5
such that ∀π : πe = eπ = π .
−1
For every π we can dene π such that ππ −1 = π −1 π = e.
For our example

1 2 3 4 5
π= = (1, 2, 5)(3, 4)
2 5 4 3 1
we have

−1 2 5 4 3 1 1 2 3 4 5
π = = = (1, 5, 2)(3, 4)
1 2 3 4 5 5 1 4 3 2
As result we have a Group of permutations.
Denition A.50 (Parity of a permutation). When X is a nite set of at

least two elements, the permutations of X (i.e. the bijective functions from
X to X ) fall into two classes of equal size: the even permutations and the
odd permutations. If any total ordering of X is xed, the parity (oddness or
evenness) of a permutation σ of X can be dened as the parity of the number
of inversions for σ , i.e., of pairs of elements x, y of X such that x < y and
σ(x) > σ(y) [52].
Example A.51 (Parity of a permutation). For the following permutation

(2, 5, 4, 1, 3) we have the following inversions
(2, 5, 4, 1, 3) →(1,2) (1, 5, 4, 2, 3) →(5,2)

(1, 2, 4, 5, 3) →(3,4) (1, 2, 3, 5, 4) →(5,4) (1, 2, 3, 4, 5)
We have made 4 inversions and as result the permutation is even.

The same result can be got if we use the following equation l−1 where l
is the circle length (5 in our case)
Denition A.52 (Alternating group). Alternating group [23] is the group

of even permutations (see denition A.50) of a nite set. The alternating
group on a set of n elements is called the alternating group of degree n, or
the alternating group on n letters and denoted by An .
Example A.53 (Sn group). If we a have a permutation of n elements then

it's possible to do by means of n! ways.
Table A.3: Cayley table for S2

◦ e τ
e e τ
τ τ e
Example A.54 (S1 group). S1 permutation of 1 element consists of only

one element e - the simplest possible group
Example A.55 (S2 group). S2 permutation consists of 2 elements:

1 2
1. identity: e =
1 2

1 2
2. transposition: τ =
2 1
It's easy to see that the Cayley table has the form A.3
Example A.56 (S3 group). S3 permutation consists of 6 elements: e, τ, τ1 , τ2 , σ, σ1 .

The most important are e, τ and σ and all others can be obtained from this
ones (see table A.4).

1 2 3
1. identity e =
1 2 3

1 2 3
2. transposition: τ =
2 1 3

1 2 3
3. circle: σ =
2 3 1

1 2 3 1 2 3
Another elements of S3 : τ1 = , τ2 = and σ1 =
1 3 2 2 1 3
1 2 3
.
3 1 2
As we can see from the table A.4 the elements e, σ, σ1 forms a subgroup
of S3 moreover all the permutation (see denition A.50). I.e. there we will
have Alternating group A3 .
Example A.57 (S3 /A3 quotient group). Lets consider the following Quotient
group S3 /A3 . S3 can be divided into 2 classes
As we can see all elements of
each of them with size 3 = |A3 |: E = A3 = {e, σ, σ1 } and G = {τ, τ1 , τ2 }. If
A.3. PERMUTATIONS 161
Table A.4: Cayley table for S3 [20]

◦ e σ σ1 τ τ1 τ2
e e σ σ1 τ τ1 τ2
σ σ σ1 e τ2 τ τ1
σ1 σ1 e σ τ1 τ2 τ
τ τ τ1 τ2 e σ1 σ
τ1 τ1 τ2 τ σ e σ1
τ2 τ2 τ τ1 σ1 σ e
we take an element x1 ∈ E and multiply it on another element of x2 ∈ E we

will get x1 x2 ∈ E (see table A.4) i.e. E · E = E . For G we can get G · G = E
and E · G = G · E = G. Therefore S3 /A3 = {E, G} forms a group of order
2. Thus
S3 /A3 ∼
= Z/2Z
Denition A.58 (Cycle). A cyclic permutation (or cycle) is a permutation

of the elements of some set X which maps the elements of some subset S of X
to each other in a cyclic fashion, while xing (that is, mapping to themselves)
all other elements of X . If S has k elements, the cycle is called a k -cycle [28].
Example A.59 (Cycle). The following permutation is a 3-cycle:
1→2
(1, 2, 3) = 2 → 3
3→1
Denition A.60 (Transposition). A cycle with only two elements is called

a transposition. [28]
Example A.61 (Transposition). The following permutation is a transposi-
tion:
1→2
(1, 2) =
2→1
See also example A.62
Example A.62 (Transposition product). The example shows a product of 2

Transpositions with a same element:
a→b
(a, c)(a, b) = b → a → c = (a, b, c)
c→a
Theorem A.63. Every permutation can be represented as a product of Trans-

positions
Proof. This is because any cycle (a1 , a2 , a3 , . . . , an−1 , an ) can be represented

as a product of transpositions as follows
(a1 , a2 , a3 , . . . , an−1 , an ) = (a1 , an )(a1 , an−1 ) . . . (a1 , a3 )(a1 , a2 )
Corollary A.64. Sn is generated by any combination of a Transposition and

n-cycle if and only if n is prime
Proof. See [12]
A.4 Rings and Fields

A.4.1 Rings
Denition A.65 (Ring). Consider a set R with 2 binary operations dened.
The rst one ⊕ (addition) and elements of R forms an Abelian group under
this operation. The second one is (multiplication) and the elements of
R forms a Monoid under the operation. The two binary operations are
connected each other via the following distributive law
• Left distributivity: ∀a, b, c ∈ R: a (b ⊕ c) = a b ⊕ a c
• Right distributivity: ∀a, b, c ∈ R: (a ⊕ b) c = a c ⊕ b c

The identity element for (R, ⊕) is denoted as 0 (additive identity). The
identity element for (R, ) is denoted as 1 (multiplicative identity).
The inverse element to a in (R, ⊕) is denoted as −a
In this case (R, ⊕, ) is called as ring.
The Ring is a generalization of integer numbers conception.
Example A.66 (Ring of integers Z). The set of integer numbers Z forms a
Ring under + and · operations i.e. addition ⊕ is + and multiplication is
·. Thus for integer numbers we have the following Ring: (Z, +, ·)
Denition A.67 (Multiplicative group). If R is a ring then the multiplica-
×
tive group (R) is a group of invertible elements of R with the dened mul-
tiplication operation.
A.4. RINGS AND FIELDS 163
Example A.68 (Multiplicative group of integers modulo n). Lets consider

×
the following group (Z/9Z) = {1, 2, 4, 5, 7, 8} [49]
1 2 3
The group has order 6. The group generator is 2: 2 = 2, 2 = 4, 2 =
8, 24 = 7, 25 = 5, 26 = 1.
Accordingly theorem A.25, there are only 2 Proper subgroup not equal to
id.
One of them is a cyclic group of order 2: {1, 8} ∼
= Z/2Z.
Another one is also cyclic and has order 3: {1, 4, 7} ∼
= Z/3Z.
A.4.2 Ideals
Denition A.69 (Ideal). Lets we have the Ring (R, ⊕, ). Subset I ⊂ R
will be an ideal if it satised the following conditions
1. (I, ⊕) is Subgroup of (R, ⊕)
2. ∀i ∈ I and ∀r ∈ R: i r ∈ I and r i ∈ I
Example A.70 (Ideal 2Z). Consider even numbers. They forms an Ideal
in Z. Because multiplication of any even number to any integer is an even.
The ideal's symbolic name is 2Z.
Example A.71 (Ring of integers modulo n: Z/nZ). Let n∈Z and n > 1.
Then nZ is an Ideal.
Two integer a, b ∈ Z are said to be congruent modulo n, written
a ≡ b( mod n)
if their dierence a−b is an integer multiple of n.
Thus we have a separation of set Z into subsets of numbers that are con-
gruent. Each subset has the following form
{r}n = r + nZ = {r + nk | k ∈ Z}
, thus
Z = {0}n ∪ {1}n ∪ · · · ∪ {n − 1}n .
Very often use the following notation
r̄ = {r}n .
We can dene the following operations
k̄ ⊕ ¯l = k + l
k̄ ¯l = k · l
The Ring where the objects are dened is called as Z/nZ.
Denition A.72 (Ideal generated by a set). Let R be a Ring and S is a sub

set of R. Consider the following set
I = {r1 s1 + · · · + rn sn |n ∈ N, ri ∈ R, si ∈ S}
I is called by an ideal generated by set S if ∀r ∈ R, i ∈ I : r · i ∈ I .

The sum in the denition of the ideal should be nite. The ring is assumed
commutative in the denition.
Denition A.73 (Principal ideal). The ideal that is generated by one el-
ement a is called as principal ideal and is denoted as (a) i.e. left principal
ideal: (a) = {ra | ∀r ∈ R} and right principal ideal: (a) = {ar | ∀r ∈ R}
Denition A.74 (Integral domain). In mathematics, and specically in ab-

stract algebra, an integral domain is a nonzero commutative Ring in which
the product of any two nonzero elements is nonzero.
Denition A.75 (Principal ideal domain). In abstract algebra, a principal

ideal domain, or PID, is an Integral domain in which every ideal is principal,
i.e., can be generated by a single element.
Denition A.76 (Maximal ideal). A maximal ideal is an ideal that is max-

imal (with respect to set inclusion) amongst all Proper ideals i.e. I is a
maximal ideal of a ring R if there are no other ideals contained between I
and R [48].
Example A.77 (Maximal ideal). If F is a Field then the only maximal ideal
is {0} [48].
Denition A.78 (Prime ideal). An ideal I of a commutative ring R is prime

2
if it has the following 2 properties [66]
1. If a, b ∈ R such that ab ∈ I then a ∈ I or b ∈ I
2. I is not equal the whole ring R
Lemma A.79 (nZ prime ideal). A positive integer n is a prime if and only
if nZ is a Prime ideal in Z [66]
Lemma A.80. All nonzero Prime ideals are maximal in a principal ideal
domain [48]
Denition A.81 (Proper ideal). I is a proper ideal of a ring R if I ( R.

2 There is a generalization of prime numbers in arithmetic
Theorem A.82 (About proper ideal). An ideal I of ring R is proper if and

only if 1R ∈
/ I.
Denition A.83 (Quotient ring). Quotient ring is a construction where one
starts with a ring R and a two-sided ideal I in R, and constructs a new ring,
the quotient ring R/I , whose elements are the Cosets of I in R subject to
special + and · operations.
Given a ring R and a two-sided ideal I ⊂ R, we may dene an equivalence
relation ∼ on R as follows: a ∼ b if and only if a − b ∈ I . The equivalence
class of the element a in R is given by
ā = {a} = a + I := {a + r : r ∈ I} .
This equivalence class is also sometimes written as a mod I and called the
"residue class of a modulo I" (see also example A.71).
The special + and · operations are dened as follows
∀x̄, ȳ ∈ R/I : x̄ + ȳ = (x + I) + (y + I) = (x + y) + I = x + y.
∀x̄, ȳ ∈ R/I : x̄ · ȳ = (x + I) · (y + I) = (x · y) + I = x · y.
As result we will get the following ring (R/I, +, ·) is called the quotient
ring of R by I .
See also Quotient group
A.4.3 Polynomial ring K [X]

Let we have a commutative Ring K . Lets create a new Ring B with the
following innite sets as elements:
f = (f0 , f1 , . . . ) , fi ∈ K, (A.2)
such that only nite number of elements of the sets are non zero.
We can dene addition and multiplication on B as follows
f + g = (f0 + g0 , f1 + g1 , . . . ) ,
f · g = h = (h0 , h1 , . . . ) , (A.3)
where X
hk = fi gj .
i+j=k
The sequences (A.2) forms a Ring with the following identities:

• Additive identity: (0, 0, . . . )
• Multiplicative identity: (1, 0, . . . )

The sequences k = (k, 0, . . . ) added and multiplied as elements of K this
allows say that such elements are elements of original Ring K . Thus K is
sub-ring of the new ring B .
Let
X = (0, 1, 0, . . . ) ,
X 2 = (0, 0, 1, . . . )
thus if we have
f = (f0 , f1 , f2 , . . . , fn , 0, . . . ) ,
where fn is the last non-zero element of (A.2), when one can get
f = f0 + f1 X + f2 X 2 + · · · + fn X n .
Denition A.84 (Polynomial ring). The Ring of sequences (A.2) with op-
erations dened by (A.3) is called as polynomial ring K [X].
Lemma A.85 (Bezout). Let a and b be nonzero integers and let d be their
greatest common divisor. Then there exist integers x and y such that
ax + by = d.
Denition A.86 (Monic polynomial). Monic polynomial is a univariate

polynomial in which the leading coecient (the nonzero coecient of highest
degree) is equal to 1. Therefore, a monic polynomial has the form
xn + an−1 xn−1 + · · · + a1 x + a0
Denition A.87 (Irreducible polynomial). An irreducible polynomial is,

roughly speaking, a non-constant polynomial that cannot be factored into
the product of two non-constant polynomials.
Example A.88 (Irreducible polynomial). The following polynomial is irre-
2
ducible in R [X]: X + 1. The following one is also irreducible despite it has
a root: X + 1.
Theorem A.89 (About irreducible polynomials). Let π(X) is an Irreducible
polynomial in K [X] and let α be a root of π(X) in a some larger eld.
∀h(x) ∈ K (X) if have the following statement: h (α) = 0 if and only if
π(X) | h(X) in K [X].
Proof. If h(X) = π(X)g(X) then h(α) = 0

From other side let π - h in K [X] this means that they are relatively
prime in K [X] and by Bezout (Lemma A.85) we can get Q, R ∈ K [X]
such that
π(X)R(X) + h(X)Q(X) = 1,
and especially for X = α we will get that 0 = 1 that is impossible.
Theorem A.90 (About ideal generated by irreducible polynomial). Let P ∈

K [X] is a polynomial and I = (P ) is an Ideal generated by the polynomial.
The I is Maximal ideal if and only if P is irreducible in K [X]
Proof. Let P is reducible i.e. P = GF . In the case (P ) ⊂ (G) and (P ) ⊂ (F )

i.e. by denition it is not a maximal ideal.
If P is irreducible then K [X] /(P ) is a eld (see claim 1.11) and by
theorem A.94 (P ) is a maximal ideal.
A.4.4 Fields
Denition A.91 (Field). The ring (R, ⊕, ) is called as a eld if (R \ {0}, )
is an Abelian group.
The inverse element to a in (R \ {0}, ) is denoted as a−1
Example A.92 (Field Q). Note that Z is not a eld because not for every
integer number an inverse exists. But if we consider a set of fractions Q=
{a/b | a ∈ Z, b ∈ Z \ {0}} when it will be a eld.
The inverse element to a/b in (Q \ {0}, ·) will be b/a.
Denition A.93 (Unique factorization domain). Unique factorization do-

main (UFD) is a commutative ring, which is an Integral domain, and in
which every non-zero non-unit element can be written as a product of prime
elements (or irreducible elements), uniquely up to order and units, analogous
to the fundamental theorem of arithmetic for the integers.
Theorem A.94 (About Quotient Ring and Maximal Ideal). Let (R, +, ·) is
a commutative Ring with additive identity 0R and multiplicative identity 1R .
Let I be an Ideal of R then I is Maximal ideal if and only if Quotient ring
R/I is a Field
Proof. See the end of section 2.3.3.

Denition A.95 (Fraction eld). The eld of fractions of an integral domain

is the smallest eld in which it can be embedded. The elements of the
eld of fractions of the integral domain R are equivalence classes (see the
construction below) written as ab with a, b ∈ R and b 6= 0. The eld of
fractions of R is sometimes denoted by Quot(R) or Frac(R) [37].
A.4.5 Characters
Denition A.96 (Character). For an abelian group G, nite or innite, a
character of G is a group homomorphism φ : G → F × where F is a eld.
The denition was taken from [4]
Example A.97 (Character). A eld homomorphism K→F is a character

×
by restricting it to the nonzero elements of K (that is, using G = K ) and
ignoring the additive aspect of a eld homomorphism.
The example was taken from [4]
Theorem A.98 (Dedekind). If we have n distinct Characters φ1 , . . . , φ n :

×
G→F then they are linearly independent i.e. if c1 , . . . , c n ∈ F satisfy
c1 φ1 (g) + · · · + cn φn (g) = 0
for all g∈G then c1 = · · · = cn = 0 (see also [4]).
A.5 Modules and Vector spaces

A.5.1 Modules
A module over a ring is a generalization of the notion of vector space over
a eld, wherein the corresponding scalars are the elements of an arbitrary
given ring (with identity) and a multiplication (on the left and/or on the
right) is dened between elements of the ring and elements of the module.
Denition A.99 (Module). Let R is a Ring and 1R is it's multiplicative

identity. A left R-module M consists of an Abelian group (M, +) and an
operation · : R × M → M such that ∀r, s ∈ R and ∀x, y ∈ M the following
relations are hold:
1. r · (x + y) = r · x + r · y
2. (r + s) · x = r · x + s · x
3. (rs) · x = r · (s · x)
A.5. MODULES AND VECTOR SPACES 169
4. 1R · x = x
Example A.100 (Module). If K is a Field then concepts of K-Vector space
and K -module are the same
Denition A.101 (Generating set of a module). A generating set G of a

module M over a ring R is a subset of M such that the smallest submodule
of M containing G is M itself [42]
Denition A.102 (Free module). The Module that has a basis (i.e. linearly
independent generating set) is called as free module [40].
For a R-module M the set E ⊆ M is a basic for M if
1. E is a generating set (see
P denition A.101) for M i.e. ∀m ∈ M ∃n < ∞:
∃ei ∈ E, ri ∈ R: m = ni=1 ri ei
2. E is linearly independent, i.e. if r1 e1 + · · · + rn en = 0M for distinct
elements e1 , . . . , en ∈ E then r1 = · · · = rn = 0R .
Denition A.103 (Rank of free module). The cardinality of any (and there-
fore every) basis is called the rank of the free module M [40].
Denition A.104 (Direct sum of modules). In abstract algebra, the direct
sum is a construction which combines several modules into a new, larger
module. The direct sum of modules is the smallest module which contains
the given modules as submodules with no "unnecessary" constraints, making
it an example of a coproduct. Contrast with the direct product, which is the
dual notion [32].
Example A.105 (Direct sum of modules). If we have 2 Free modules M
and N with bases m1 , m2 , . . . , mm and n1 , n2 , . . . , nn . Then the Direct sum
of modules A = M ⊕ N will also be a free module with composite basis:
m1 , m2 , . . . , mm , n1 , n2 , . . . , nn
Denition A.106 (Finitely generated module). Finitely generated module
is a module that has a nite generating set (see also denition A.101) [39].
Denition A.107 (Cyclic module). A (left) R-Module M is called cyclic if
M can be generated by a single element i.e.
M = hxi = Rx = {rx | r ∈ R}
for some x ∈ M [27].
Claim A.108. M is a cyclic R module if and only if exists left ideal I⊂R
such that M∼
= R/I
Proof. See [15]. Note that [14] p. 149 has the claim as a denition of the
Cyclic module.
A.5.2 Linear algebra

Denition A.109 (Vector space). Let F is a Field. The set V is called as
vector space under F if the following conditions are satised
1. We have a binary operation V × V → V (addition): (x, y) → x + y

with the following properties:
(a) x + y = y + x
(b) (x + y) + z = x + (y + z)
(c) ∃0 ∈ V such that ∀x ∈ V : x + 0 = x
(d) ∀x ∈ V ∃ − x ∈ V such that x + (−x) = x − x = 0
2. We have a binary operation F × V → V (scalar multiplication) with

the following properties
(a) 1F · x = x
(b) ∀a, b ∈ F, x ∈ V : a · (b · x) = (ab) · x.
(c) ∀a, b ∈ F, x ∈ V : (a + b) · x = a · x + b · x
(d) ∀a ∈ F, x, y ∈ V : a · (x + y) = a · x + a · y
Lemma A.110 (About vector space isomorphism). 2 vector spaces L and

M with same dimension dimL = dimM then there exists an Isomorphism
between them
Denition A.111 (Image). The image or range of a linear map f : V → W

is the following set [46]:
Im f = {w ∈ W : w = f (v), v ∈ V }
Denition A.112 (Kernel). The kernel of a linear map f : V → W is the

following set [45]:
ker f = {v ∈ V : f (v) = 0}
Denition A.113 (Rank). The rank of a linear map f : V → W is dimen-

sion of Image: rgf = dim Im f [54]:
Denition A.114 (Nullity). The nullity of a linear map f : V → W is

dimension of Kernel: nulf = dim ker f [45]:
A.5. MODULES AND VECTOR SPACES 171
Theorem A.115 (Ranknullity theorem). Let V and W be vector spaces

over some eld and let T : V → W be a linear map. Then the Rank of T is
the dimension of the image of T and the Nullity of T is the dimension of the
kernel of T , so we have
dim (Im T ) + dim (ker T ) = dim V

or, equivalently
rg(T ) + nul(T ) = dim(V )
Proof. See [55]

Denition A.116 (General linear group of a vector space). If V is a Vector
space over eld K the general linear group of V , written GL (V ) or Aut (V ),
is the group of all automorphisms of V , i.e. the set of all bijective linear trans-
formations V → V , together with functional composition as group operation
[41].
Denition A.117 (Dual space). Given any vector space V over a eld F ,
the dual space V is dened as the set of all linear maps φ : V → F (linear
∗
functionals). The dual space V ∗ itself becomes a vector space over F when
equipped with an addition and scalar multiplication satisfying:
(ϕ + ψ)(x) = ϕ(x) + ψ(x)
(aϕ)(x) = a (ϕ(x))
for all φ, ψ ∈ V ∗ , x ∈ V , and a ∈ F .
This is also named as algebraic dual space at [34].
Denition A.118 (Dual basis). If {ei } is a basis of V then exists a basis
{vj } of Dual space such that vj (ei ) = δij and is called as dual basis. [33]
Denition A.119 (Degenerate bilinear form). A degenerate bilinear form
f (x, y) on a vector space V is a bilinear form such that the map from V to
V ∗ (the Dual space of V ) given by v → (x → f (x, v)) is not an isomorphism
[29].
An equivalent denition when V is nite-dimensional is that it has a non-
trivial kernel: there exist some non-zero x ∈ V such that ∀y ∈ V f (x, y) = 0
Denition A.120 (Non-degenerate bilinear form). A nondegenerate or non-

singular form is one that is not degenerate, meaning that the map from V to
V ∗ (the Dual space of V ) given by v → (x → f (x, v)) is an isomorphism [29]
or equivalently when V is nite-dimensional if and only if ∀y ∈ V f (x, y) = 0
implies x = 0, i.e. the map f has a trivial kernel.
Denition A.121 (Eigenspace). Consider a linear map T : V → V . The

set of all eigenvectors of T corresponding to the same eigenvalue, together
with the zero vector, is called an eigenspace or characteristic space of T [63]
Denition A.122 (Diagonalizable map). If V is a nite-dimensional vector

space, then a linear map T : V → V is called diagonalizable if there exists
an ordered basis of V with respect to which T is represented by a diagonal
matrix.
A linear map T : V → V is diagonalizable if and only if the sum of the
dimensions of its eigenspaces is equal to n = dim(V ), which is the case if and
only if there exists a basis of V consisting of eigenvectors of T . With respect
to such a basis, T will be represented by a diagonal matrix. The diagonal
entries of this matrix are the eigenvalues of T [61].
Theorem A.123 (About eigenvalues of a diagonalizable linear map). A

linear map T : V → V with n = dim(V ) is diagonalizable if it has n distinct
eigenvalues [61].
Theorem A.124 (About invertible matrix). Let A be a matrix over F. If

A is diagonalizable, then so is any power of it. Conversely, if A is invertible,
n
F is algebraically closed, and A is diagonalizable for some n that is not an
integer multiple of the characteristic of F, then A is diagonalizable.
Proof. See [61]
A.6 Functions aka maps

A.6.1 Functions
Denition A.125 (Surjection). The function f : X → Y is surjective (or
onto) if ∀y ∈ Y , ∃x ∈ X such that f (x) = y .
Denition A.126 (Injection). The function f : X → Y is injective (or

one-to-one function) if ∀x1 , x2 ∈ X , such that x1 6= x2 then f (x1 ) 6= f (x2 ).
Denition A.127 (Bijection). The function f : X → Y is bijective (or

one-to-one correspondence) if it is an Injection and a Surjection.
Denition A.128 (Homomorphism). The homomorphism is a function (map)

between two sets that preserves its algebraic structure. For the case of
groups (X, ◦) and (Y, ) the function φ : X → Y is called homomorphism if
∀x1 , x2 ∈ X it holds φ (x1 ◦ x2 ) = φ (x1 ) φ (x2 ).
A.6. FUNCTIONS AKA MAPS 173
Denition A.129 (Field homomorphism). Let [16] (F, +, ×) and (K, ⊕, ⊗)

are Fields. Let φ : F → K is a mapping such that both (F, +) → (K, ⊕)
and (F, ×) → (K, ⊗) are Homomorphisms under φ.
That is ∀a, b ∈ F :
1. φ (a + b) = φ (a) ⊕ φ (b)
2. φ (a × b) = φ (a) ⊗ φ (b)
Then φ : (F, +, ×) → (K, ⊕, ⊗) is a eld homomorphism.
Note that in the lectures we often refer the Field homomorphism as just
Homomorphism.
Denition A.130 (Isomorphism). If a map is Bijection as well as Homo-
morphism when it is called as isomorphism.
We use the following symbolic notation for isomorphism between X and
Y: X ∼
= Y.
Denition A.131 (Endomorphism). An endomorphism is a morphism (or
homomorphism) from a mathematical object to itself [35]
Denition A.132 (Automorphism). Automorphism is an isomorphism from
a mathematical object to itself.
Denition A.133 (Embedding). When some object X is said to be em-
bedded in another object Y , the embedding is given by some injective and
structure-preserving map f : X → Y . The precise meaning of "structure-
preserving" depends on the kind of mathematical structure of which X and
Y are instances.
The fact that a map f : X → Y is an embedding is often indicated by
the use of a "hooked arrow", thus: f : X ,→ Y . On the other hand, this
notation is sometimes reserved for inclusion maps.
Theorem A.134 (First isomorphism). Let G is a group and φ:G→H is
a surjective Homomorphism. Then if N = ker φ we have
H∼
= G/N
Theorem A.135 (Isomorphism extension theorem). Let F is a Field and

0 0
E is an Algebraic extension of F . F is another Field and E the Algebraic
extension of F 0.
If there exists an Isomorphism φ : F → F0 then it can be extended to an
0
isomorphism τ : E → E .
Proof.The proof of the isomorphism extension theorem depends on Zorn

(Lemma 2.15) .
??? The theorem seems to be very close to the theorem 2.17.
Theorem A.136 (About group homomorphism). A group homomorphism

is Injection i the kernel is trivial [44]
A.6.2 Category theory

Denition A.137 (Commutative diagram). A commutative diagram is a
diagram of objects (also known as vertices) and morphisms (also known as
arrows or edges) such that all directed paths in the diagram with the same
start and endpoints lead to the same result by composition
The following diagram commutes if fAB = fCB fAC or fAB (x) = fCB (fAC (x)).
A fAB B
fAC fCB
A.7 Number theory

Denition A.138 (Euler's totient function). In number theory, Euler's to-
tient function counts the positive integers up to a given integer n that are
relatively prime to n. It is written using the Greek letter phi as φ (n), and
may also be called Euler's phi function. It can be dened more formally
as the number of integers k in the range 1 ≤ k ≤ n for which the greatest
common divisor gcd (n, k) = 1. The integers k of this form are sometimes
referred to as totatives of n.
The denition was taken from [36]
Example A.139 (Euler's totient function). For example [36], the totatives
of n = 9 are the six numbers 1, 2, 4, 5, 7 and 8. They are all relatively
prime to 9, but the other three numbers in this range, 3, 6, and 9 are not,
because gcd (9, 3) =gcd (9, 6) = 3 and gcd (9, 9) = 9. Therefore, φ (9) = 6.
As another example, φ (1) = 1 since for n = 1 the only integer in the range
from 1 to n is 1 itself, and gcd (1, 1) = 1.
Index
K = F2 / (X 2 + X + 1) example, 20
example, 13 n-th cyclotomic polynomial, 93,
K embedding of E into the 94, 131
algebraic closure K̄ denition, 92
remark, 137 example, 92
K -algebra is not a eld n=5
example, 11 example, 99
K (α1 , . . . , αn ) n=8
denition, 19 example, 97
S1 group nZ prime ideal lemma
example, 160 declaration, 164
S2 group
Abelian group, 40, 115119,
example, 160
152154, 156, 157, 162,
S3 group
167, 168
example, 160
denition, 156
S3 /A3 quotient group Abelianization
example, 160 denition, 153
S3 /A3 quotient group example, About Algebraic closure theorem
152 declaration, 30
Sn group About eigenvalues of a
example, 159 diagonalizable linear map
Sn solvability theorem theorem
declaration, 119 declaration, 172
X − 2 over Q
3
About extension of
example, 29 homomorphism theorem,
C 77, 78
example, 16, 30 declaration, 32
Fpn About group homomorphism
remark, 36 theorem
Q extension declaration, 174
175
176 INDEX
About homomorphism of elds declaration, 51

lemma, 13, 104, 106 About towers theorem
About ideal generated by About uniqueness of object
irreducible polynomial dened by universal
theorem property lemma
About invertible matrix theorem About vector space isomorphism
declaration, 172 lemma, 63
About irreducible polynomials declaration, 170
theorem, 15 Action, 79, 125
declaration, 166 denition, 153
About minimal polynomial Algebraic closure, 30, 32, 35, 37,
existence lemma, 18 45, 90, 104, 113
About order of element of an Algebraic element, 15, 20, 21, 42,
Abelian group theorem 44
About proper ideal theorem Algebraic extension, 20, 21, 30,
declaration, 165 32, 35, 44, 47, 173
About quotient eld denition, 17
remark, 14 Algebraically closed eld, 30
About quotient group and denition, 30
commutator subgroup Alternating group, 89, 160
theorem denition, 159
declaration, 153 Artin theorem, 85
About Quotient Ring and declaration, 80
Maximal Ideal theorem, 13 Artin-Schreier extension, 103, 104
About separable extensions Artin-Schreier theorem
theorem, 49 declaration, 102
declaration, 45 Automorphism, 26, 29, 42, 43, 78,
About splitting elds theorem 80, 124, 125
About stem eld
remark, 26 Bezout lemma, 12, 13, 39, 167
About subgroups of a cyclic group declaration, 166
theorem Base-change theorem, 58, 69, 105
About tensor product existence Bijection, 13, 8587, 91, 173
lemma denition, 172
INDEX 177
Chain, 3133 example, 93

denition, 31
Character, 168 Dedekind theorem, 128
denition, 168 declaration, 168
example, 168 Degenerate bilinear form
Characteristic, 36, 92, 99, 103, 113 denition, 171
denition, 12 Degree 2
Chinese remainder theorem, 60, example, 87
64, 65, 67, 70, 112, 144 Degree 3
declaration, 62 example, 88
Class, 35, 36 Degree of inseparability
denition, 149 denition, 44
Commutative diagram, 50, 60 Degree of separability
Commutativity proof Diagonalizable map, 101
Commutator, 153 Direct product, 110, 155, 157
Commutator subgroup, 118120 Direct sum, 125, 154, 157
Composite extension, 111, 114, Direct sum of modules, 169
121 denition, 169
Correspondence theorem theorem, discriminant, 88
118 denition, 88
declaration, 152 Discriminant of polynomial degree
Coset, 151, 152, 165 3
Cycle Disjoint union, 129
Cyclic extension, 102, 103 Dual basis, 142
Cyclic group, 40, 41, 99, 100, 117, Dual space, 142, 171
118, 151153, 156 denition, 171
denition, 153
Cyclic Group is Abelian theorem Eigenspace, 101
Cyclic module, 128, 133, 169 Eisenstein criterion
Cyclotomic polynomial Eisenstein criterion lemma, 23
178 INDEX

Embedding, 35, 36, 96 First isomorphism theorem, 62,
denition, 173 121, 155
Endomorphism, 126, 128, 140, 141 declaration, 173
denition, 173 Fixed eld, 98, 109
Euler's totient function, 41, 92 denition, 84
denition, 174 Fixed point
Extension solvable by radicals, Fraction eld, 135
104, 114 denition, 168
denition, 113 Free group action, 79
Extension which cannot be denition, 154
generated by a single Free module, 51, 54, 141, 169
element denition, 169
example, 76 Frobenius homomorphism
remark, 36
Field, 9, 13, 17, 68, 164, 167, 169,
Frobenius homomorphism remark,
170, 173
37, 47
denition, 167
Fundamental theorem of cyclic
Field Q
groups theorem, 91, 100,
example, 167
112
Field extension, 25, 27, 37, 49
declaration, 153
denition, 10
Field extensions
example, 11 Galois correspondence theorem,
Field homomorphism, 10, 13, 104, 90, 91, 98, 101, 109, 114,
106 121, 132
Field of complex numbers C Galois extension, 7981, 83, 85,
example, 9 97, 100, 108, 114, 120,
Finite extension, 18, 21 121, 125, 128, 132
Finite eld, 75 Galois group, 81, 85, 89, 90,
denition, 36 97102, 121, 123125, 128,
Finite group of prime order 131, 132, 143
theorem denition, 80
Finitely generated abelian group, GAP example
141, 157 Cyclotomic polynomial, 93
denition, 157 Galois group, 132
Finitely generated module Minimal polynomial, 16
INDEX 179
Polynomial factorization over 72, 96, 100, 106108, 142,

a nite eld in GAP, 40 172, 174
Gauss lemma, 23, 94, 95, 136 denition, 172
declaration, 22 Inseparable degree, 138
General equation of degree n denition, 45
example, 123 Integral domain, 62, 63, 132, 164,
General linear group of a vector 167
space, 124 denition, 164
denition, 171 Integral element, 133135, 144
Generating set of a module denition, 132
denition, 169 Integral extension, 134
Group, 78, 151, 155, 156, 159 denition, 134
denition, 150 Integrally closed, 135, 145
Group Z/2Z denition, 135
example, 150 Irreducible polynomial, 12, 13, 25,
Group Z/2Z example, 153 38, 41, 166
Group of permutations S3 denition, 166
Group of permutations S4 Isomorphism, 10, 12, 13, 25, 26,
example, 115 28, 29, 33, 35, 38, 42, 50,
Group representation 51, 63, 80, 170, 173
Isomorphism extension theorem
Homomorphism, 10, 13, 17, 3537, theorem
44, 45, 49, 50, 52, 57, 60, declaration, 173
69, 77, 100, 106, 116, 120,
124, 137, 155, 173 K-algebra, 10, 11, 13, 16, 59, 137
K-algebra and homomorphism
Ideal, 12, 15, 58, 61, 70, 143, 163, lemma, 63
167 declaration, 10
denition, 163 Kernel, 103, 170
Ideal 2Z denition, 170
example, 163 Klein four group, 40, 98, 99, 115,
Ideal generated by a set, 31 156
Idempotent, 74 example, 157
denition, 74 Kummer extensions section, 103,
Image, 103, 104, 170 104, 131
denition, 170
Injection, 10, 13, 17, 33, 47, 63, Lagrange theorem, 121, 154, 156
180 INDEX
declaration, 151 Not every Abelian group is cyclic

Linearly disjoint extensions remark, 156
denition, 106 Not solvable polynomial of degree
local, 72 5
Nullity, 171
Maximal ideal, 13, 31, 32, 60, denition, 170
6265, 68, 143, 144, 167 Number eld, 136
example, 164
Minimal polynomial, 16, 33, 44, 49 Of extension of homomorphism
example, 16 on normal extensions
Module, 9, 50, 55, 56, 169 remark, 79
denition, 168 Orbit, 79, 81, 123, 154
Monic polynomial, 25, 38, 94, 132, Orbit-stabilizer theorem theorem,
133, 136 123
Monoid, 150, 162 Order of element in group, 41, 102
Multiplicative group, 95
denition, 162 Parity of a permutation
Multiplicative group of integers denition, 159
modulo n example, 159
example, 163 Perfect eld, 47, 69
denition, 47
Nilpotent element, 66, 68 Permutation
Non-degenerate bilinear form, 142 Polynomial factorization over a
denition, 171 nite eld
Norm, 138140, 144 example, 39
denition, 137 Polynomial factorization over a
Normal base theorem nite eld in GAP
Normal extension, 77, 79, 83, 86 Polynomial ring
remark, 77 Polynomial solvable by radicals,
Normal subgroup, 85, 87, 115, 121
116, 151, 152, 155 denition, 114
denition, 151 Prime eld, 93
INDEX 181
denition, 12 reduced, 72
Prime ideal, 62, 64, 65, 144, 164 denition, 68
denition, 164 Regular representation, 124, 126
Primitive element denition, 124
example, 75 Relatively prime ideals
Primitive element example, 98 denition, 61
Primitive element theorem, 81, 82, Ring, 9, 12, 55, 56, 61, 68,
125, 144 162168
Primitive roots of unity, 92, 94, Ring of integers
97, 101 denition, 136
denition, 92 Ring of integers Z
Principal ideal example, 162
denition, 164 Ring of integers modulo n: Z/nZ
Principal ideal domain, 15, 144 example, 163
denition, 164
Proper ideal, 31, 58, 65, 164 Separable closure
Proper subgroup, 99, 163 Separable degree, 44
Pure inseparable polynomial Separable element, 45, 46, 49
Separable extension, 45, 74, 79,
Quotient group, 115117, 153, 83, 138
160, 165 denition, 44
denition, 151 Separable polynomial, 44
Quotient group theorem Set of invariants
Quotient of the group action, 129 Simple group
Quotient ring, 32, 65, 167 Simple subgroup of an abelian
denition, 165 group theorem
declaration, 158
Rank, 103, 171 Solvable group
denition, 170 denition, 115, 118
Rank of free module Splitting eld, 28, 31, 37, 38, 42,
denition, 169 77, 83, 84, 121, 131, 143
Ranknullity theorem theorem, denition, 27
103, 105 Stabilizer subgroup, 123, 154
182 INDEX
Stem eld, 27, 28, 38, 84 declaration, 18

denition, 25 Trace, 138, 140
Structure of nite K -algebra denition, 137
theorem Transitive group action, 7880,
declaration, 63 102, 123, 144
Subgroup, 40, 100, 151, 153, 163 denition, 154
denition, 151 Transposition, 115, 161, 162
Surjection, 10, 13, 17, 33, 47, 62, denition, 161
63, 85, 87, 121, 142144, example, 161
155, 172 Transposition product
Sylow corollary, 123
Unique factorization domain, 23,
Tensor product 39, 135
The complexication of a real Universal property, 5054, 104,
vector space 106
example, 55 denition, 50, 59
The fundamental theorem of Vector space, 9, 11, 16, 17, 19, 54,
nitely generated abelian 63, 64, 73, 124, 169, 171
groups theorem, 40, 118 denition, 170
declaration, 157
The multiplicativity formula for Zorn lemma, 32, 174
degrees theorem, 27, 28 declaration, 31
Bibliography
[1] Baker, A. An introduction to Galois theory / Andrew Baker. 2013.
http://www.maths.gla.ac.uk/~ajb.
[2] Brown, K. The primitive element theorem / Ken Brown. http://www.

math.cornell.edu/~kbrown/6310/primitive.pdf.
[3] Conrad, K. Finite elds / Keith Conrad. http://www.math.uconn.

edu/~kconrad/blurbs/galoistheory/finitefields.pdf.
[4] Conrad, K. Linear independence of characters / Keith Conrad.

http://www.math.uconn.edu/~kconrad/blurbs/galoistheory/
linearchar.pdf.
[5] Conrad, K. Separability ii / Keith Conrad. http://www.math.uconn.

edu/~kconrad/blurbs/galoistheory/separable2.pdf.
[6] Conrad, K. Tensor poroducts / Keith Conrad. http://www.math.

uconn.edu/~kconrad/blurbs/linmultialg/tensorprod.pdf.
[7] Conrad, K. Trace and norm ii / Keith Conrad. http://www.math.

uconn.edu/~kconrad/blurbs/galoistheory/tracenorm2.pdf.
[8] Gap - groups, algorithms, programming - a system for computational

discrete algebra. https://www.gap-system.org/.
[9] Gap - reference manual. https://www.gap-system.org/Manuals/

doc/ref/chap0.html.
[10] Groupwiki. "general ane group: ga(1, 5)". 2017. [Online; ac-
cessed 15-March-2017]. https://groupprops.subwiki.org/w/index.
php?title=General_affine_group:GA(1,5)&oldid=45747.
[11] Groupwiki. "subgroup structure of symmetric group:s5 ". 2017. [On-

line; accessed 15-March-2017]. https://groupprops.subwiki.org/
183
184 BIBLIOGRAPHY
w/index.php?title=Subgroup_structure_of_symmetric_group:
S5&oldid=49180.
[12] Gruber, A. Why is s5 generated by any combination of a

transposition and a 5-cycle? Mathematics Stack Exchange.
URL:http://math.stackexchange.com/q/357673 (version: 2013-05-11).
http://math.stackexchange.com/q/357673.
[13] (http://math.stackexchange.com/users/326053/starfall), S. Tensor

product of galois extension and algebraic closure. Mathematics Stack
Exchange. URL:http://math.stackexchange.com/q/2217402 (version:
2017-04-04). http://math.stackexchange.com/q/2217402.
[14] Lang, S. Algebra / Serge Lang. Graduate Texts in Mathematics. 3 edi-

tion. Springer Science and Business Media, 2002. https://books.
google.ru/books?id=eOUlBQAAQBAJ.
[15] Miller, K. Isomorphism and cyclic modules. Mathematics Stack

Exchange. URL:http://math.stackexchange.com/q/1003598 (version:
2014-11-03). http://math.stackexchange.com/q/1003598.
[16] ProofWiki. Denition: Field homomorphism. 2017. [Online; ac-

cessed 18-June-2017]. https://proofwiki.org/w/index.php?title=
Definition:Field_Homomorphism&oldid=283682.
[17] Ribenboim, P. Classical Theory of Algebraic Numbers / P. Ribenboim.

Universitext. Springer New York, 2001. https://books.google.
ru/books?id=u5443xdaNZcC.
[18] Rowland, T. Group xed point. 2016. From MathWorldA Wol-

fram Web Resource, created by Eric W. Weisstein. http://mathworld.
wolfram.com/GroupFixedPoint.html.
[19] user58289. Subgroups of a cyclic group and their order. Mathemat-

ics Stack Exchange. URL:https://math.stackexchange.com/q/410389
(version: 2013-07-27). https://math.stackexchange.com/q/410389.
[20] Wikibooks. Abstract algebra/group theory/permutation groups

wikibooks, the free textbook project. 2016. [Online; accessed 27-
September-2016]. https://en.wikibooks.org/w/index.php?title=
Abstract_Algebra/Group_Theory/Permutation_groups&oldid=
3070727.
BIBLIOGRAPHY 185
[21] Wikipedia. Simple extension wikipedia, the free encyclopedia.

2014. [Online; accessed 20-September-2016]. https://en.wikipedia.
org/w/index.php?title=Simple_extension&oldid=595508851.
[22] Wikipedia. "Íîðìàëüíîå ðàñøèðåíèå Âèêèïåäèÿ, ñâîáîäíàÿ

ýíöèêëîïåäèÿ". 2014. http://ru.wikipedia.org/?oldid=
65922374.
[23] Wikipedia. Alternating group wikipedia, the free encyclopedia.

org/w/index.php?title=Alternating_group&oldid=736689793.
[24] Wikipedia. Commutator wikipedia, the free encyclopedia. 2016.

[Online; accessed 30-September-2016]. https://en.wikipedia.org/w/
index.php?title=Commutator&oldid=740207690.
[25] Wikipedia. Commutator subgroup wikipedia, the free encyclopedia.

org/w/index.php?title=Commutator_subgroup&oldid=737371946.
[26] Wikipedia. Correspondence theorem (group theory) wikipedia,

the free encyclopedia. 2016. [Online; accessed 2-March-2017].
https://en.wikipedia.org/w/index.php?title=Correspondence_
theorem_(group_theory)&oldid=744531320.
[27] Wikipedia. Cyclic module wikipedia, the free encyclopedia.

2016. [Online; accessed 13-March-2017]. https://en.wikipedia.
org/w/index.php?title=Cyclic_module&oldid=752351296.
[28] Wikipedia. Cyclic permutation wikipedia, the free encyclopedia.

2016. [Online; accessed 6-March-2017]. https://en.wikipedia.org/
w/index.php?title=Cyclic_permutation&oldid=755419947.
[29] Wikipedia. Degenerate bilinear form wikipedia, the free en-

cyclopedia. 2016. [Online; accessed 10-September-2016].
https://en.wikipedia.org/w/index.php?title=Degenerate_
bilinear_form&oldid=738751020.
[30] Wikipedia. Direct product of groups wikipedia, the free

encyclopedia. 2016. [Online; accessed 19-February-2017].
https://en.wikipedia.org/w/index.php?title=Direct_product_
of_groups&oldid=723991376.
186 BIBLIOGRAPHY
[31] Wikipedia. Direct sum wikipedia, the free encyclopedia. 2016.

index.php?title=Direct_sum&oldid=738351383.
[32] Wikipedia. Direct sum of modules wikipedia, the free en-

https://en.wikipedia.org/w/index.php?title=Direct_sum_
of_modules&oldid=730018916.
[33] Wikipedia. Dual basis wikipedia, the free encyclopedia. 2016.

[Online; accessed 1-April-2017]. https://en.wikipedia.org/w/index.
php?title=Dual_basis&oldid=748729591.
[34] Wikipedia. Dual space wikipedia, the free encyclopedia. 2016.

[Online; accessed 6-October-2016]. https://en.wikipedia.org/w/
index.php?title=Dual_space&oldid=742901318.
[35] Wikipedia. Endomorphism wikipedia, the free encyclopedia.

2016. [Online; accessed 2-October-2016]. https://en.wikipedia.
org/w/index.php?title=Endomorphism&oldid=726230579.
[36] Wikipedia. Euler's totient function wikipedia, the free en-

https://en.wikipedia.org/w/index.php?title=Euler%27s_
totient_function&oldid=736552571.
[37] Wikipedia. Field of fractions wikipedia, the free encyclopedia.

org/w/index.php?title=Field_of_fractions&oldid=720271734.
[38] Wikipedia. Finitely generated abelian group wikipedia, the

free encyclopedia. 2016. [Online; accessed 3-June-2016].
https://en.wikipedia.org/w/index.php?title=Finitely_
generated_abelian_group&oldid=723506843.
[39] Wikipedia. Finitely generated module wikipedia, the free

encyclopedia. 2016. [Online; accessed 5-October-2016].
https://en.wikipedia.org/w/index.php?title=Finitely_
generated_module&oldid=735554374.
[40] Wikipedia. Free module wikipedia, the free encyclopedia. 2016.

[Online; accessed 9-January-2016]. https://en.wikipedia.org/w/
index.php?title=Free_module&oldid=699002213.
BIBLIOGRAPHY 187
[41] Wikipedia. General linear group wikipedia, the free encyclopedia.

org/w/index.php?title=General_linear_group&oldid=738480571.
[42] Wikipedia. Generating set of a module wikipedia, the

free encyclopedia. 2016. [Online; accessed 5-October-2016].
https://en.wikipedia.org/w/index.php?title=Generating_set_
of_a_module&oldid=732648521.
[43] Wikipedia. Group action wikipedia, the free encyclopedia. 2016.

[Online; accessed 28-August-2016]. https://en.wikipedia.org/w/
index.php?title=Group_action&oldid=735249701.
[44] Wikipedia. Group homomorphism wikipedia, the free encyclopedia.

w/index.php?title=Group_homomorphism&oldid=737886552.
[45] Wikipedia. Kernel (linear algebra) wikipedia, the free en-

https://en.wikipedia.org/w/index.php?title=Kernel_(linear_
algebra)&oldid=735290769.
[46] Wikipedia. Linear map wikipedia, the free encyclopedia. 2016.

index.php?title=Linear_map&oldid=740065546.
[47] Wikipedia. Local ring wikipedia, the free encyclopedia. 2016. [On-
line; accessed 23-April-2016]. https://en.wikipedia.org/w/index.
php?title=Local_ring&oldid=716779040.
[48] Wikipedia. Maximal ideal wikipedia, the free encyclopedia.

2016. [Online; accessed 29-July-2016]. https://en.wikipedia.org/
w/index.php?title=Maximal_ideal&oldid=704326783.
[49] Wikipedia. Multiplicative group of integers modulo n wikipedia,

the free encyclopedia. 2016. [Online; accessed 20-September-2016].
https://en.wikipedia.org/w/index.php?title=Multiplicative_
group_of_integers_modulo_n&oldid=739054442.
[50] Wikipedia. Nilpotent wikipedia, the free encyclopedia. 2016. [On-

line; accessed 29-July-2016]. https://en.wikipedia.org/w/index.
php?title=Nilpotent&oldid=727125150.
188 BIBLIOGRAPHY
[51] Wikipedia. Normal subgroup wikipedia, the free encyclopedia.

org/w/index.php?title=Normal_subgroup&oldid=737429252.
[52] Wikipedia. Parity of a permutation wikipedia, the free en-

https://en.wikipedia.org/w/index.php?title=Parity_of_a_
permutation&oldid=736707840.
[53] Wikipedia. Quotient group wikipedia, the free encyclopedia.

org/w/index.php?title=Quotient_group&oldid=726415849.
[54] Wikipedia. Rank (linear algebra) wikipedia, the free en-

https://en.wikipedia.org/w/index.php?title=Rank_(linear_
algebra)&oldid=739885289.
[55] Wikipedia. Ranknullity theorem wikipedia, the free en-

cyclopedia. 2016. [Online; accessed 15-February-2017].
https://en.wikipedia.org/w/index.php?title=Rank%E2%80%
93nullity_theorem&oldid=730054227.
[56] Wikipedia. Simple group wikipedia, the free encyclopedia.

w/index.php?title=Simple_group&oldid=709515196.
[57] Wikipedia. Subgroup wikipedia, the free encyclopedia. 2016.

index.php?title=Subgroup&oldid=737399477.
[58] Wikipedia. Subgroups of cyclic groups wikipedia, the free

https://en.wikipedia.org/w/index.php?title=Subgroups_of_
cyclic_groups&oldid=744131880.
[59] Wikipedia. Sylow theorems wikipedia, the free encyclopedia.

org/w/index.php?title=Sylow_theorems&oldid=735518140.
[60] Wikipedia. Adjugate matrix wikipedia, the free encyclopedia.

2017. [Online; accessed 3-April-2017]. https://en.wikipedia.org/
w/index.php?title=Adjugate_matrix&oldid=765833302.
BIBLIOGRAPHY 189
[61] Wikipedia. Diagonalizable matrix wikipedia, the free en-

cyclopedia. 2017. [Online; accessed 10-February-2017].
https://en.wikipedia.org/w/index.php?title=Diagonalizable_
matrix&oldid=764159566.
[62] Wikipedia. Disjoint union wikipedia, the free encyclopedia.
2017. [Online; accessed 13-April-2017]. https://en.wikipedia.org/
w/index.php?title=Disjoint_union&oldid=774047863.
[63] Wikipedia. Eigenvalues and eigenvectors wikipedia, the free
https://en.wikipedia.org/w/index.php?title=Eigenvalues_
and_eigenvectors&oldid=764288456.
[64] Wikipedia. Finite group wikipedia, the free encyclopedia. 2017.
[Online; accessed 5-February-2017]. https://en.wikipedia.org/w/
index.php?title=Finite_group&oldid=759713845.
[65] Wikipedia. Klein four-group wikipedia, the free encyclopedia.
2017. [Online; accessed 9-February-2017]. https://en.wikipedia.
org/w/index.php?title=Klein_four-group&oldid=759017274.
[66] Wikipedia. Prime ideal wikipedia, the free encyclopedia. 2017.
[Online; accessed 29-March-2017]. https://en.wikipedia.org/w/
index.php?title=Prime_ideal&oldid=763442154.
[67] Yuan, Q. Using vieta's theorem for cubic equations to de-
rive the cubic discriminant. Mathematics Stack Exchange.
URL:http://math.stackexchange.com/q/103504 (version: 2012-01-29).
http://math.stackexchange.com/q/103504.
[68] Êîñòðèêèí À. È. Ââåäåíèå â àëãåáðó. ÷. 3. Îñíîâíûå ñòðóêòóðû
àëãåáðû / Êîñòðèêèí À. È. ÌÖÍÌÎ, Ìîñêâà, 2012.

Apuntes Introducion Teoria de Galois

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Apuntes Introducion Teoria de Galois

Uploaded by

Copyright:

Available Formats

Lectures notes in Introduction to Galois

Theory by Ekaterina Amerik

1 Generalities on algebraic extensions 9

2 Stem eld, splitting eld, algebraic closure 25

3 Finite elds. Separability, perfect elds 35

3.6 Perfect elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 Tensor product. Structure of nite K-algebras 49

5 Structure of nite K-algebras continued 67

6 Galois correspondence and rst examples 83

7 Galois correspondence and rst examples. Examples contin-

7.5 Linearly disjoint extensions. Examples . . . . . . . . . . . . . 106

8 Solvability by radicals, Abel's theorem. A few words on re-

9 Ring extensions, norms and traces, reduction modulo p 131

A.4.1 Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

1.1 Field extensions: examples

• Compatibility with scalars: (ax) · (by) = (ab)(x · y)

1.1.2 Field extension

Proof. Really if I have a K-algebra I can dene the Homomorphism f (k) =

Lemma 1.6 (About homomorphism of elds). Any Field homomorphism is

Proof. Lets proof by contradiction. Really if f (x) = f (y) and x 6= y then

Example 1.7 (Field extensions). C is a eld extension for R. R is a eld

where x1 , x2 ∈ R. I will dene the multiplication for L (our K algebra) as

Denition 1.9 (Characteristic). If L is a eld there are 2 possibilities

1. 1 + 1 + · · · 6= 0. In this case Z ⊂ L but Z is not a eld therefore L is

2. 1 + 1 + · · · + 1 = m i=1 1 = 0 for some m ∈ Z. The rst time when

Denition 1.10 (Prime eld). If L is a eld Q ⊂ L, i.e. charL = 0 then Q

Claim 1.11. Let K [X] Ring of polynomials. The P ∈ K [X] is an Irre-

Proof.How we can see it? If Q ∈ K [X] is a polynomial that Q ∈

Proof. The K [X] / (X − a) is a eld just because X − a is Irreducible poly-

φ : K [X] / (X − a) −−−−−−−→ K (1.2)

φ (P1 + P2 ) = P1 (a) + P2 (a) = φ (P1 ) + φ (P1 ) .

The same holds with the multiplication:

φ (P1 · P2 ) = P1 (a) · P2 (a) = φ (P1 ) · φ (P1 ) .

That completes the proof of the Field homomorphism.

1.2 Algebraic elements. Minimal polynomial

Example 1.13 (K = F2 / (X 2 + X + 1)). Lets consider the following eld

X̄ 2 + X̄ + 1 = P (X) + (P (X)) = (P (X)) ≡ 0 mod P.

So this is the structure of a eld of four elements. The cardinality of

L [X] is an irreducible polynomial over L. Let α is a root of the polynomial

and as result ∀x ∈ L/ (P ) exists xi ∈ L such that

1.2.2 Algebraic elements

1.2.3 Minimal polynomial

Proof. We know that K [X] is a Principal ideal domain and a polynomial

Denition 1.17 (Minimal polynomial). If α is Algebraic element then the

Example 1.18 (Minimal polynomial)

1.3 Algebraic elements. Algebraic extensions

As soon as K [α] is a K-algebra it is a Vector space over K generated by

Example 1.20 (C).

Proposition 1.21. The following assignment are equivalent

2. K [α] is a nite dimensional Vector space over K

Proof.Lets proof that 1 implies 2. If α is algebraic over K then using lemma

Pmin (α, K) = αd + ad−1 αd−1 + a1 α + a0 = 0,

this means that any αn can be represented as a linear combination of -

Denition 1.22 (Algebraic extension). L an extension of K is called alge-

Proposition 1.23. If L is algebraic over K then any K-subalgebra of L is

Proof. Let L0 ⊂ L is a subalgebra and let α ∈ L0 . We want to show that α

Proposition 1.24. LetK ⊂ L ⊂ M . α ∈ M - algebraic over K then α

We can consider the following example as an illustration for proposition

1.4 Finite extensions. Algebraicity and nite-

Theorem 1.27 (The multiplicativity formula for degrees). Let K ⊂ L ⊂ M .

also L is a vector sub space of M thus if [M : K] < ∞ then [L : K] < ∞.

• L-basis over M : (e1 , e2 , . . . , en )

2 Stem eld, splitting eld, algebraic closure 25

3 Finite elds. Separability, perfect elds 35

3.6 Perfect elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 Tensor product. Structure of nite K-algebras 49

5 Structure of nite K-algebras continued 67

6 Galois correspondence and rst examples 83

7 Galois correspondence and rst examples. Examples contin-

Proof. Really if I have a K-algebra I can dene the Homomorphism f (k) =

Lemma 1.6 (About homomorphism of elds). Any Field homomorphism is

Example 1.7 (Field extensions). C is a eld extension for R. R is a eld

where x1 , x2 ∈ R. I will dene the multiplication for L (our K algebra) as

Denition 1.9 (Characteristic). If L is a eld there are 2 possibilities

1. 1 + 1 + · · · 6= 0. In this case Z ⊂ L but Z is not a eld therefore L is

2. 1 + 1 + · · · + 1 = m i=1 1 = 0 for some m ∈ Z. The rst time when

Denition 1.10 (Prime eld). If L is a eld Q ⊂ L, i.e. charL = 0 then Q

Proof. The K [X] / (X − a) is a eld just because X − a is Irreducible poly-

Example 1.13 (K = F2 / (X 2 + X + 1)). Lets consider the following eld

So this is the structure of a eld of four elements. The cardinality of

Denition 1.17 (Minimal polynomial). If α is Algebraic element then the

2. K [α] is a nite dimensional Vector space over K

this means that any αn can be represented as a linear combination of -

Denition 1.22 (Algebraic extension). L an extension of K is called alge-

1.4 Finite extensions. Algebraicity and nite-

Denition 1.28 (K (α1 , . . . , αn )). K (α1 , . . . , αn ) ⊂ L generated by α1 , . . . , αn

• K is a eld, α is an Algebraic element over K if it is a root of a

• If an extension is nite then it is algebraic

• An extension is nite if and only if it is algebraic and generated by a

2.1 Stem eld. Some irreducibility criteria

Such things exist, for instance we can take K [X] / (P ). It is a eld

Proof.Existence: K [X] / (P ) can be took as the stem eld.

Remark 2.3 (About stem eld). 1. In particular: If a stem eld contains

2. If E stem eld then [E : K] = deg P

3. If [E : K] = degP and E contains a root of P then E is a stem eld

Lets K (α) is a stem eld of P over K then [K (α) : K] = deg P = n.

2.2 Splitting eld

Theorem 2.7 (About splitting elds). 1. Splitting eld L exists and [L : K] ≤

2. If L and M are 2 splitting elds then ∃φ : L ∼

Proof. Lets prove by induction on d. The rst case (d = 1) is trivial the K

Example 2.11 (C). C is an Algebraically closed eld. This will be proved

Denition 2.12 (Algebraic closure). An algebraic closure of K is a eld L

Theorem 2.13 (About Algebraic closure). Any eld K has an Algebraic

Denition 2.14 (Chain). Let P is a partially ordered set (≤ is the order

If we take a Quotient ring by maximal ideal it's always a eld 18 otherwise

Ki is a eld where each polynomial P ∈ Ki−1 has a root. The eld Ki is

3.1 An example (of extension)s. Finite elds

Corollary 3.2. Any Algebraic extension of K embeds (see denition A.133)

3.1.1 Finite elds

3.2 Properties of nite elds