Computation of Galois groups and corresponding

polynomials
Evie Roebroek

Bachelor thesis

Bachelor of Mathematics and Physics

Supervisor: Prof. Frits Beukers

January 2018
Contents
1 Introduction 1

2 Galois theory 2
2.1 Groups, rings and fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Polynomial rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Galois groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Theorems that help find the Galois group . . . . . . . . . . . . . . . . . . . 8
2.5 The fundamental theorem of Galois Theory . . . . . . . . . . . . . . . . . 11

3 An algorithm to find the Galois group 13

3.1 Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Resolvents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Tschirnhausen transformations . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 General algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4 Inverse Galois Theory 21

4.1 Galois groups for degree 3 polynomials . . . . . . . . . . . . . . . . . . . . 21
4.2 Galois groups for degree 4 polynomials . . . . . . . . . . . . . . . . . . . . 22
4.3 Symmetric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4 Subgroups and subfields of order 2 . . . . . . . . . . . . . . . . . . . . . . 29
4.5 Overview of developments in Inverse Galois Theory . . . . . . . . . . . . . 31

5 Conclusion 33

6 Appendices 34
6.1 Appendix 1: Proof of Theorem 4.1 for V4 . . . . . . . . . . . . . . . . . . . 34
6.2 Appendix 2: Proof of Theorem 4.1 for C4 . . . . . . . . . . . . . . . . . . . 36
1 Introduction
In the 18th century, mathematicians had found solutions for the general quadratic, cubic
and quartic equation, but did not manage to find a solution for the general quintic equa-
tion. At the beginning of the 19th century, Abel proved that there was no general solution
for the quintic equation, but Évariste Galois was the first to prove why there was no solu-
tion. The theory he used to prove this was difficult to understand at that time, and since
he died before his paper could be published, it took several years before the importance of
his theory was acknowledged.

Nowadays, Galois theory is usually explained using field theory, and looking at the
Galois group as a group of field automorphisms. Galois himself however started his theory
by looking at permutation groups, and investigated which permutations acted on the roots
of a certain polynomial such that any equation that is satisfied by the roots will still be
satisfied after permuting them. That is how we will approach Galois theory in this thesis,
by looking at permutations of roots rather than automorphisms of fields, and minimizing
the amount of field theory needed to understand the concepts.

We will start with some elementary definitions, and after that we will look into poly-
nomial rings and irreducibility of polynomials. We then have enough knowledge to define
what a Galois group is the way Galois himself did, and investigate some properties of
Galois groups which we will need later on. We will then study Dedekinds and Frobe-
nius theorem, together with some tricks that help us find the Galois group and look at
the fundamental theorem, which explicitly describes the relation between groups and fields.

Knowing the basics of Galois theory, we can investigate an algorithm given by Cohen
in his book ’A Course in Computational Algebraic Number Theory’ [1]. This algorithm is
a very practical one to determine the Galois group of polynomials for degrees up to 7. In
this thesis we will look into the theory behind the algorithm. We first examine the theory
of invariants, which we will need to make resolvents. After looking into resolvents and
Tschirnhausen transformations we have all the instruments to understand a more general
version of Cohen’s algorithm, with which we can determine the Galois group of any poly-
nomial.

In the last chapter, we will do the opposite of what Cohen did. We will now look at a
transitive group and try to find polynomials with rational coefficients that have this group
as Galois group. Whether or not this is always possible is an unresolved problem, called
the Inverse Galois Problem (IGP). We will not attempt to study this problem, since it is
far above the range of this thesis, but we will look at subgroups of S3 and S4 and construct
infinitely many polynomials for these groups. We will then use elementary symmetric
functions to find out more about the group structure and the correspondence to invariant
fields and subfields.

1
2 Galois theory
In this first chapter we will describe the basics of Galois theory. For the sake of com-
pleteness, we will start with some elementary definitions of a group, a ring and a field
and quickly move on to polynomial rings and their properties. We can then state what a
Galois group is, and look at this special correspondence between polynomials and groups.
Next, we will look at ways to reduce the number of options for the Galois group, including
Dedekind’s and Frobenius’ theorems. In the last section we will state the fundamental
theorem of Galois theory, which describes the relation between groups and fields, and its
special form when considering normal groups.

2.1 Groups, rings and fields

Galois theory makes a connection between polynomials, groups and fields. Before we will
explain what Galois theory is and what it does, we first make sure that it is clear what we
mean when talking about a group, a ring and a field. We will therefore start with some
basic definitions. Since a group is the most elementary one, let us look at this definition
first.

Definition 2.1 (Group). A group is a set G with a multiplication denoted by ∗ which

satisfies the following three conditions:

1. The multiplication must be associative, that means if a, b, c ∈ G, then (a ∗ b) ∗ c =

a ∗ (b ∗ c).

2. The set has an identity e, such that for all a ∈ G, e ∗ a = a ∗ e = a.

3. Each element of the set has an inverse, which means that for all a ∈ G, there exists
an element a−1 such that a ∗ a−1 = e.

We call a group abelian or commutative if in addition:

4. For all a, b ∈ G we have a ∗ b = b ∗ a.

Groups come in all shapes and sizes, but the group we will be using most in this thesis
is the permutation group, Sn . Sn is the group containing all possible permutations of the
elements 1 up to n. We usually write elements of Sn as cycles, e.g. (12) is the action in
which 1 and 2 are reversed and all the potential other elements are left the same. The
identity is thus denoted by (1)(2)...(n) or simply by id. Note that there is not a unique way
to write an element of a group, for example (13)(23)=(132). There is however a unique
way to write the elements of a permutation group as disjoint cycles. For that reason, we
will write the permutations as disjoint cycles in this thesis.

When looking at sets that form a group under a certain multiplication, we are sometimes
interested in whether or not a subset also forms a group under that multiplication. For

2
example, knowing Z forms a group under addition, we might wonder whether the set of
even integers, denoted by 2Z, forms a group under addition. We therefore need to check
the following conditions:
Definition 2.2 (Subgroup). Let G be a group under a certain multiplication, and H a
subset of G. We call H a subgroup of G if it satisfies the following conditions:
1. The identity e is an element of H.
2. Each element of H has an inverse.
3. The set H is closed under multiplication, which means that ∀ a, b ∈ H we have that
a ∗ b ∈ H.
Note that we do not have to check if H satisfies the first condition of a group, namely
associativity. We know that H is a subset of G so we automatically know that it is
associative under the multiplication.
For all n, Sn has subgroups. One of those is An , the group of even permutations. We
will call a permutation even if we can write it as an even amount of 2-cycles. Note that
(132)=(13)(23) is even, and so are all other 3-cycles. Other useful groups are Cn , which is
the group that represents the permutations we get by rotating a n-gon, a polygon with n
sides, and Dn which adds the action of reflection to Cn .
Aside from the multiplication within a group, we can also let the elements of a group
act on a set. We define this operation as follows:
Definition 2.3 (Group action). Let G be a group and X be a set. Then a group action φ
of G on X is a function from G ×X → X defined by φ(g, x) = g · x satisfying the following
two conditions:
1. e · x = x for all x ∈ X.
2. (gh) · x = g(h · x) for all g, h ∈ G.
We call a group action transitive if in addition:
3. For all x, y ∈ X there exists g ∈ G such that g(x)=y, or equivalently, that there is
only one orbit.
When we look at the group Sn or one of its subgroups, we see that the group action on
a set is nothing more than a permutation of the elements of that set. For example, taking
X = {x1 , x2 , x3 } and g = (132) ∈ S3 , we see that g(x1 ) = x3 , g(x2 ) = x1 , g(x3 ) = x2 . We
know that Sn acts transitively on a set, but not all of its subgroups do this. For exam-
ple, the subgroup of S4 consisting of {(12), (34), (12)(34), id} does not act transitively on
X = {x1 , x2 , x3 , x4 }, since there is no element in it that will send x1 to x3 .

Now that we understand the basics of group theory, we can look at rings. Rings and
groups have similar structures. In fact, a ring has two different maps which put two
different kinds of structures on the group. The formal definition is the following:

3
Definition 2.4 (Ring). A ring R is a set with two mappings, from R×R → R which we
denote by + and ∗. Moreover, it has two identity elements, respectively 0 and 1, which are
both in R, and it satisfies the following conditions:

1. R is an abelian group under +.

2. R is associative under ∗.

3. For all a, b, c ∈ R we have that a ∗ (b + c) = a ∗ b + a ∗ c and (b + c) ∗ a = b ∗ a + c ∗ a.

4. For all a, b ∈ R, a ∗ b = b ∗ a.

A ring is called a field if in addition:

5. 1 6= 0 and for all non-zero a ∈ R there exists an inverse a−1 ∈ R such that a∗a−1 = 1.

Looking at the definitions above, we see that Z with the regular addition and multipli-
cation is a ring, but not a field. Q on the other hand, is a field since for every a ∈ Q, we
know we can write a = pq , with p, q ∈ Z. Therefore a−1 = pq ∈ Q, so Q is a field. Further
on, Z/pZ is a field for all primes p. In this thesis we will only use fields with characteristic
0, which means that the intersection of all subfields of the field is Q. This means that fields
like Z/pZ will not be used, which makes the proof of some theorems a lot easier. Further
on, we will only use ground fields that are subfields of C.

2.2 Polynomial rings

If R is a ring, R[x] forms the set of polynomials in x with coefficients in R. It is easy to
check that R[x] is a ring as well. Therefore, we can look at the ring of polynomials with
coefficients in Q, namely Q[x]. Note that we can also look at polynomials in Z[x], since
the roots will not change if we multiply f by a constant. Considering the fact that Q is a
field and Z is not, we usually prefer to look at Q[x]. Let f ∈ Q[x], so f is a polynomial
with rational coefficients. Let n be the degree of f . The fundamental theorem of algebra
tells us the following:

Theorem 2.5 (Fundamental theorem of algebra). Every non-zero, single variable, degree
n polynomial with complex coefficients has, counted with multiplicity, exactly n complex
zeroes.

So we know that f has n roots in C and that some of them might be real numbers or even
integers. We call these roots α1 , α2 , ..., αn . For some purposes, we want to look at the field
in which α1 lies as well as Q. We need to extend Q with the element α1 and make sure it is
still a field. We call this construction a field extension, and denote it with Q(α1 ). If f is
irreducible, we can write Q(α1 ) explicitely as Q(α1 )={a0 + a1 α1 + a2 α12 + ... + an−1 α1n−1 |ai ∈
Q ∀ 0 ≤ i < n}. Since α1 is a root of f , we know that f (α1 )=0 and therefore that
α1n + a1 α1n−1 + ... + an−1 α1 + an =0. So we can express α1n in terms of lower powers of α1
and therefore we do not need to include powers higher than n in Q(α1 ).

4
 We can then extend Q(α1 ) in the same manner with the element α2 to get Q(α1 , α2 )
etc. The smallest field which contains all roots of the polynomial f is called the splitting
field of f over Q, and is denoted by Q(α1 , ..., αn ).
In this case,
√ we use Q as the ground field, but in general we can also use other ground
fields, like Q( 2). Let us look at an example of such a splitting field. We will consider the
polynomial f = x2 + 1 ∈ Q[x]. We know that its zeroes are i, −i so we need to extend the
ground field Q by the element i. We see that Q(i) = {a + bi | a, b ∈ Q} , which is also the
splitting field of the polynomials x2 + 4, x2 + 9 etc. We can always create such a splitting
field, by extending our ground field K ⊂ C by the roots α1 , ..., αn , and then choosing the
smallest field possible.

From now on, we will look at monic irreducible polynomials only. We call a polynomial
monic if the coefficient of the highest degree is 1. We call a polynomial irreducible if it
cannot be written as a product of polynomials that all have degree ≥ 1. We work in Q and
we know Q[x] is a unique factorization domain, that means, every non-zero element
which is not a unit can be written uniquely (up to units and ordering of factors) as a
product of irreducible elements. This makes it a lot easier to figure out if a polynomial is
irreducible, but still not easy. We will list a few helpful theorems and tricks here.

Theorem 2.6. Let R be a unique factorization domain and K its quotient field. Let
f (x) ∈ R[x] be a monic polynomial of degree n, then f (x) is irreducible in K[x] if and only
if it is irreducible in R[x].

So if we want to prove that a monic polynomial with integer coefficients is irreducible

in Q[x], we only have to look at factorization in Z[x], since we know Z is a unique
factorization domain. For low degree polynomials, say n < 6, we can look for linear
or quadratic factors to quickly see if a polynomial is irreducible. Let us write f (x) =
an xn + an−1 xn−1 + ... + a1 x + a0 , then we know that if f is reducible, say f (x) = b(x)c(x)
, then b0 c0 = a0 . Look for linear factors by inserting all integers that a0 is divisible by. If
none of them gives 0, f has no linear factors, since there is no b0 such that f (b0 ) = 0 and
thus f (x) does not contain a factor (x − b0 ). We can do the same for quadratic factors,
although this will take us a little more effort.

Another very useful theorem is Eisenstein’s criterion. We call a polynomial f (x) =

an xn + an−1 xn−1 + ... + a1 x + a0 in a unique factorization domain R[x] an Eisenstein
polynomial with respect to π if there is an irreducible element π ∈ R such that π - an , π |
ai for all 1 ≤ i < n but π 2 - a0 .

Theorem 2.7 (Eisenstein). An Eisenstein polynomial with respect to a prime π in Z[x] is

irreducible in Q[x].

Proof: Assume f (x) is an Eisenstein polynomial and reducible. Then we can write
f (x) = b(x)c(x), where b(x) = bm xm + ... + b1 x + b0 and c(x) = cq xq + ... + c1 x + c0 . If we
look at these polynomials modulo π we see that f (x) = an xn (mod π) and from that we

5
know that b(x) = bm xm (mod π) and c(x) = cq xq (mod π). So all coefficients except the
first one are 0 (mod π) and in particular we see that b0 = 0 (mod π) and c0 = 0 (mod π).
Now look at a0 = b0 c0 , since π | b0 and π | c0 we have that π 2 | a0 , hence we have a
contradiction and f (x) is irreducible.

It usually does not take much time to see whether a polynomial is an Eistenstein
polynomial or not. If it is, this is by far the easiest way to prove that the polynomial is
irreducible. If it is not and we want to work with a high degree polynomial, we will use a
program like Mathematica to factor it for us.

2.3 Galois groups

Consider an irreducible polynomial f ∈ Q[x] of degree n, with roots α1 , α2 , ..., αn . We
now know that f has a splitting field, which we denote by L. Now look at the map
φ : Q[x1 , x2 , ..., xn ] → L that sends a polynomial h(x1 , ..., xn ) 7→ h(α1 , ..., αn ). We call this
the evaluation homomorphism. Note that f (xi ) will be sent to zero by φ, since αi is a root
of f so f (αi ) = 0. The set of all polynomials that will be sent to zero by φ is called the
kernel of φ, which we also denote by Ker(φ). The polynomials inside the kernel tell us a
lot about the relations between the roots.

When examining a certain polynomial g(x1 , ..., xn ) that lies in the kernel of φ, we could
ask ourselves what would happen with φ(g) if we would exchange two roots. Does φ(g)
still lie in the kernel or not? And what happens when we exchange two other roots, or
three? More generally speaking, what happens when we let the elements of Sn act on the
polynomials, such that if σ ∈ Sn , then σ(g(x1 , ..., xn )) = g(xσ(1) , ..., xσ(n) ). It turns out
that some polynomials stay in the kernel, no matter which permutation acts on them, and
that others only stay there when certain permutations act on them. These permutations
have some special properties which we will examine, starting with the following definition.

Definition 2.8 (Galois group). Let f be a polynomial in Q[x] and its splitting field be L.
The Galois group is the group of permutations of the variables xi that respects the relations
between the roots of f . Formally, we define the Galois group by:

Gal(L/Q) = {σ ∈ Sn |σ(Ker(φ)) = Ker(φ)}.

Let us prove that the above definition makes the Galois group an actual group. We call
it G and denote Ker(φ) = I. We know that G ⊂ Sn since all elements of G have to be
permutations of order ≤ n. Therefore we know that G satisfies condition 1. Condition 3 is
obvious as well, since id(I) = I. If σ ∈ G, then I = id(I) = σ −1 σ(I) = σ −1 (σ(I)) = σ −1 (I),
so σ −1 (I) = I and σ −1 ∈ G. Now we only need to prove that G is closed, since it is a
subgroup of Sn (not necessarily a proper subgroup). Take σ, τ ∈ G, then we need to prove
στ ∈ G. We see that στ (I) = σ(I) = I so στ ∈ G. We know that permutations are not
abelian, so G forms a group, but not an abelian group.

6
 Let us look at an example of computing a Galois √ √ group.
√ We √ will√use √ the polynomial
√ √
f (x) = x4 − 10x2 + 1. We find√that the √ roots are 2 + 3, 2 − 3, − 2 + 3, − √2 −√ 3,
so we need to extend Q by 2 and 3 to construct our splitting field: L = Q( 2, 3).
Since deg(f )=4, we need to look for a subgroup of S4 that respects the relations between
these roots. This means its elements send a polynomial to the kernel if and only if that
polynomial was already in the kernel.
Let us look at elements of√the kernel
√ √ of φ,√that√ is, functions
√ √h(x1√ , x2 , x3 , x4 ), such that
φ(h) = h(α1 , α2 , α3 , α4 ) = h( 2 + 3, 2 − 3, − 2 + 3, − 2 − 3) = 0. We see that
h1 = x1 + x4 lies in the kernel, just like h2 = x2 + x3 . If we look a little further, we can also
see that h3 = x1 x2 +1, h4 = x1 x3 −1 and similar functions are elements of the kernel. Since
the elements of the Galois group must send an element of the kernel to another element of
the kernel, we find that for σ ∈ Gal(L/Q) we have the following permutations:
√ √ √ √
σ : ±√2 ± √3 → ∓√2 ± √3
τ :± 2± 3→± 2∓ 3

That way we have four different permutations, namely {σ, τ, στ, id}, which give us ex-
actly the four-group of Klein.

We have seen that the splitting field L of Q is the smallest field which contains all the
roots of a certain polynomial. Generally speaking, we can extend any field by one or more
elements and form a new field. We call the latter one a field extension of the original field.
We denote the degree of a field extension L over a field K by [L : K].
We call a finite field extension L/K a Galois extension if it is a splitting field of a
polynomial f (x) ∈ K[x], where f has deg(f )=n distinct zeroes. In that case, we say f is
separable. Note that f does not have to be irreducible.
Let us look at some properties of Galois extensions and field extensions in general.

Theorem 2.9 (Tower relation for fields). Let K be a field, L a finite field extension of K
and M a finite field extension of L. Then [M : K] = [M : L][L : K].

The proof of this theorem is not relevant for the course of this thesis, which is why it
will not be discussed here. A good proof can be found in Rings and Galois theory by Frits
Beukers [2, Thm 6.4.1]. We will however use this theorem to prove the following important
theorem.

Theorem 2.10. Let K a field and L a finite field extension. Then |Gal(L/K)| ≤ [L : K],
and if |Gal(L/K)| = [L : K], then L/K is a Galois extension.

Proof: We will prove this theorem by induction in n = [L : K]. We see that for
n = 1 we have L = K and the theorem holds. Assume the theorem holds for all ex-
tensions of degree i where 1 ≤ i ≤ n and let us now look at an extension of degree
n + 1. Take α in L but not in K and let f be its minimal polynomial of degree d. Let
G = Gal(L/K) and H = Gal(L/K(α)). Note that H is a subgroup of G and look at
the right coset decomposition, G = ri=1 gi H where gi ∈ G ∀ 1 ≤ i ≤ r and all gi H are
S

7
pairwise disjoint. Since α ∈ L and gi ∈ Gal(L/K) we have that gi (α) is a root of f . We
see that all gi (α) are distinct, since if gi (α)=gj (α) for i 6= j we would get gi−1 gj (α) = α
and gi−1 gj ∈ H, which is in contradiction with our right coset decomposition. So we
have r different elements out of the set of d roots, which means that r ≤ d. Therefore,
|Gal(L/K)| = |G| = r|H| ≤ d|H| ≤ [K(α) : K][L : K(α)] = [L : K] so our hypothesis is
proved. Moreover, if r = d for every minimal polynomial of an element in L , which means
that every f has precisely d distinct zeroes, then |Gal(L/K)| = [L : K].

Using the first two theorems, we can prove the following theorem, which is used a lot
in all kinds of Galois theory.
Theorem 2.11. Let L/Q be a Galois extension over Q. Let α ∈ L such that σ(α) = α for
all σ ∈ Gal(L/Q). Then α ∈ Q.
Proof: Since σ(α) = α for all σ ∈ Gal(L/Q) we know that Gal(L/Q) = Gal(L/Q(α)).
Therefore we can write [L : Q] = |Gal(L/Q)| = |Gal(L/Q(α))| ≤ [L : Q(α)]. Using the
theorem above, where Q(α) is the intermediate field, we also see that [L : Q] = [L :
Q(α)][Q(α) : Q]. It immediately follows that [Q(α) : Q] = 1 and therefore that α ∈ Q.

There is a third way of stating that a field extension is Galois, namely that L/K is
a Galois extension if and only if it is a normal and separable extension. Since we work
in characteristic 0, all field extensions are separable. We call a finite field extension L/K
normal if every irreducible polynomial in K[x] that has one zero in L, has all of its ze-
roes in L. Note that this is more difficult to prove, since it means that if we have a field
extension L/K we have to find a way to check all polynomials in K[x]. It can however
be of help when we already know a field extension is Galois and have found one zero of a
polynomial, to find the other zeroes of that polynomial. For a proof of this statement, see
[2, Thm. 7.3.6] where we have to keep in mind that since we work with subfields of C, all
extensions are separable.

Now that we have a basic understanding of Galois groups and their properties, we can
start using them. The next section will give us an idea of how to figure out which group
is the Galois group of a certain polynomial.

2.4 Theorems that help find the Galois group

In the next chapter we will look at an algorithm used to compute Galois groups for every
polynomial of degree up to 7. Before we will be able to fully understand that algorithm,
there are a few special properties of Galois groups we need to cover. In this section we will
cover these theorems, and add some other useful theorems. Some of these theorems might
seem to have no connection to each other at this point. However, they are widely used to
determine the Galois group corresponding to a certain polynomial.
Theorem 2.12 (Dedekind). Let f ∈ Q[x] be an irreducible polynomial of degree n with n
distinct roots and p a prime that does not divide the discriminant of f . Then the degrees

8
of the k irreducible factors of f mod (p) correspond to the cycle types of Gal(f ). So, if
f mod (p) = g1 (x) · g2 (x) · ... · gk (x) with di = deg(gi ) then Gal(f ) contains an element with
cycle lengths (d1 , d2 , ..., dk ).

This theorem is proven by van der Waerden, in his well-known book Algebra [3]. We
will not give the proof here but we will look at the purpose of this theorem. Note that we
can use the theorem above for all possible primes p which do not divide the discriminant.
This gives us infinitely many primes to test with, although usually only the first few times
we find new relevant information. Therefore, we can eliminate quite some groups with the
information we find here.

Let us look at an example of how to use Dedekind’s theorem. We take the polynomial
f (x) = x4 + 4x3 − 2x − 1. We can calculate by hand or using a computer program
the discriminant of f , Disc(f ) = −3632 = 24 · 227 where 227 is prime. This means we
do not need to worry about the primes dividing the discriminant, except for not using
p = 2, p = 227. We see that:

x4 + 4x3 − 2x − 1 (mod 3) = (x2 + 1)(x2 + x + 2)

x4 + 4x3 − 2x − 1 (mod 5) = x4 + 4x3 + 3x + 4
x4 + 4x3 − 2x − 1 (mod 7) = (x3 + 5)(x + 4)
x4 + 4x3 − 2x − 1 (mod 11) = (x3 + x2 + 8x + 7)(x + 3)
x4 + 4x3 − 2x − 1 (mod 13) = (x2 + 5x + 11)(x + 10)(x + 2).

It now follows from Dedekind’s theorem that the Galois groups has elements with cycle
types (2,2),(4),(1,3),(2,1,1). Let us look at the five transitive subgroups of S4 and the cycle
lengths they contain next to (1,1,1,1) from the identity, namely:

• C4 =< (1234) >, which contains cycle types of lengths 2 and 4

• V4 = {id, (12)(34), (13)(24), (14)(23)} which contains cycle types (2,2)

• D4 =< (1234), (13) > which contains cycle types (4),(2,2) and (2,1,1)

• A4 which contains all even permutations, so only the cycle types (2,2) and (3,1).

So we see that there is no proper transitive subgroup of S4 containing all the desired
cycle types and therefore we can conclude that the Galois group of f is S4 . Note that for
higher degree polynomials this is not a waterproof method to determine the Galois group.
It does, however, help to eliminate sets of subgroups and can therefore be very helpful in
the process.

Another theorem that will tell us something about the cycle structure of the Galois
group is Frobenius’s theorem. Where Dedekind’s theorem tells us which cycle types appear
in the Galois group, the theorem of Frobenius tells us how often certain cycle types will

9
occur. The notion of natural density will be used in this theorem. Let X be a subset of
the natural numbers, and n ∈ N. Denote by a(n) the number of elements in X smaller or
equal to n. We say that the natural density of X is α if a(n)
n
→ α for n → ∞

Theorem 2.13 (Frobenius). Let f ∈ Q[x] an irreducible polynomial and let its Galois
group have degree n over Q. Denote by η the proportion of g ∈ Gal(f ) that have cycle type
(d1 , d2 , ..., dk ). Then the primes p such that f mod (p) splits in factors of degree d1 , d2 , ..., dk
have natural density equal to η.

Again, the proof of this theorem is too complex to be treated here. For those who wish
to look into it, a proof can be found here [4]. This theorem thus tells us that every cycle
type that is an element of the Galois group will eventually occur, and that the frequency
with which it occurs goes to the density of that cycle type in the Galois group.
On top of these two rather advanced theorems which we can apply on Galois groups,
there are two simpler theorems left that will help us when looking for the Galois group.
Both theorems work with elementary properties of the Galois group, but the theorems turn
out to be quite useful.

Theorem 2.14. Let f ∈ Q[x] be a monic irreducible polynomial of degree n. Then its
Galois group is a subgroup of the alternating group An if and only if the discriminant of f
(disc(f )) is a square.
2
Proof:
Y Let αi be the roots of f , then its discriminant is given by: disc(f ) = D where
D= (αj − αi ) . If we take σ ∈ Gal(f ) then we know that σ(D) = ± D. Now if
1≤i<j≤n
Gal(f ) ⊂ An then all permutations in the Galois group are even, which means that they
will give a plus sign and thus that σ(D) = D for all σ ∈ Gal(f ). Theorem 2.11 then tells
us that D ∈ Z and thus that disc(f ) is a square.
Conversely, if disc(f ) is a square, then we know that D ∈ Z so σ(D) = D and thus
that all permutations have to be even. Therefore, Gal(f ) ⊂ An .

Theorem 2.15. Let f ∈ Q[x] be a monic polynomial of degree n. Then f is irreducible if

and only if the Galois group is a transitive subgroup of Sn .

Proof: Let f be an irreducible polynomial of degree Qm n. Now assume the zeroes

Qmm ≤ n. NowQ
α1 , ..., αm are in the same orbit, with look at i=1 (x−αi ). Since α1 , ..., αm are
m
in the same orbit we know Qm that σ( i=1 (x − α i ))= i=1 (x − αi ) for all σ ∈ Gal(f ). Theorem
2.11 now tells us that i=1 (x − αi ) ∈ Z[x] and therefore that it forms a non-trivial factor
of f . But since f is irreducible, there does not exist such a factor for m < n and we can
conclude m = n. This means that the Galois group has only one orbit, and therefore that
it is transitive.
Conversely, let f be an irreducible polynomial, then we know that there do exist non-
trivial factors and therefore that the Galois group is not transitive.

10
2.5 The fundamental theorem of Galois Theory
We would like to get as much information as possible from the results we have already
found. For example, if we have calculated the splitting field and the Galois group of a
certain polynomial, we would like to use this information to determine the Galois group
corresponding to a subfield of the original field. Before looking at the theorem which tells
us this is possible, we need to introduce a notation.
Let L be a field and H a group that permutes the elements of L. Then we denote the
fixed field under H by

LH = {x ∈ L | σ(x) = x ∀ σ ∈ H}.

So we see that the fixed field consists of the elements that are left unchanged by the action
of a group. For example, when looking at the field C and the group that consists of the
identity and the action of complex conjugation, we see that the fixed field will be R.
We can now look at the theorem that tells us how subfields of a splitting field are
related to subgroups of the corresponding Galois group.

Theorem 2.16 (Fundamental theorem of Galois Theory). Let L/K be a Galois extension.
Then the map from intermediate fields M of L/K and subgroups of Gal(L/K) given by M
7→ Gal(L/M) is a bijection. Its inverse map is given by H 7→ LH .

The proof of this theorem is given in [2, Thm 8.2.1].

Let us look at an example that illustrates this theorem. We consider the polynomial
4 2
√ x√− 10x + 1 in Q[x], of which we already calculated that the splitting field is
f =
Q( 2, 3). We know its Galois group is isomorphic to the fourgroup of Klein, and we
denote the different permutations by:
√ √ √ √
σ1 ( 2, 3) = ( 2, 3)
√ √ √ √
σ2 ( 2, 3) = (− 2, 3)
√ √ √ √
σ3 ( 2, 3) = ( 2, − 3)
√ √ √ √
σ4 ( 2, 3) = (− 2, − 3).
√ √ √ √ √
Note that has Q( 2, 3) has three intermediate fields: Q( 2), Q( 3), Q( 6). These
fields are the fixed fields under (σ1 , σ3 ), (σ1 , σ2 ) and (σ1 , σ4 ) respectively. So, we see that
these are the subgroups of Gal(L/Q) that correspond to these subfields M of L/Q and
Theorem 2.16 now tells us that there exists a bijection between each of these subfields and
the corresponding subgroup of the Galois group.
In this case, we have even created three new Galois extensions of Q by looking at the
subfields of our original field extension. This is a consequence of the fact that C2 occurs
as a normal subgroup of V4 , three times to be precise. Therefore, all three of the subfields
are Galois extensions themselves. The following theorem states this:

11
Theorem 2.17. Let L/K be a finite Galois extension. Then an intermediate extension M
with K ⊂ M ⊂ L is normal over K if and only if Gal(L/M) is a normal subgroup of Gal
(L/K). Moreover, we have the isomorphism:

Gal(M/K) ∼
= Gal(L/K)/Gal(L/M ).

For a proof, again see [2, Thm. 8.2.2]. This theorem is going to be extremely useful in
the next chapters. We have seen in section 2.3 that a field extension is Galois if and only if
it is normal. Therefore, this theorem tells us that every normal subgroup of the group we
are looking at automatically gives us an intermediate field which is also a Galois extension
over Q. When constructing polynomials in Chapter 4, we thus have to make sure we take
into account all normal subfields.

12
3 An algorithm to find the Galois group
In this chapter we will examine the algorithms that Cohen describes in his book ”A course
in computational algebraic number theory” [1]. He there states algorithms to find the
Galois group corresponding to a given polynomial. These algorithms work excellently for
degrees up to 7, but there is little theoretical clarification of why they work. We will
here try to illustrate why and how these algorithms work. Most of the theory needed is
explained in the first chapter, but there are some applications of Galois theory that we
still need to look into. We will start by looking into invariants, which are needed to create
resolvents. Resolvents in turn are the main tools used in these algorithms. Next, we will
see how Tschirnhausen transformations are used, and in the fourth section we will unite
these tools to create a general algorithm that works for all irreducible polynomials.

3.1 Invariants
We have seen that if we have an element that is invariant under all permutations of the
Galois group, that element belongs to the ground field. It turns out that certain com-
binations of the roots of a polynomial f ∈ Q[x] are invariant under all permutations
of the Galois group of f . We can use this information to decide if the Galois group of
f is a (not necessarily proper) subgroup of the group we are looking at. For example,
look at D4 . We can see D4 as the group which permutes the corners of a square by
rotation or reflection. We choose the following representation: D4 = h(1234), (13)i =
{id, (1234), (13)(24), (1432), (13), (14)(23), (24), (12)(34)}.

We can see that y = x1 x3 + x2 x4 is invariant under the permutation (1234), and also
under (13). Because D4 is generated by these two permutations, we know it is invariant
under D4 . The only transitive subgroup of S4 containing D4 is S4 itself, and we can easily
see that y is not invariant under all permutations of S4 . We call such an invariant a gen-
erating invariant of a group G ⊂ Sn : a function in the roots xi that is invariant under
all permutations of G, but not under all permutations of any other transitive subgroup of
Sn containing G. We call it a generating invariant since it generates the invariant field of
G, which is a field extension of the invariant field over Sn that we will see later on.

Note that if we would have chosen another representation for D4 , so a conjugate of

this representation, we would have gotten another invariant. For example, if we choose
D40 = h(1324), (12)i (we conjugate D4 by (23)), we have that y 0 = x1 x2 + x3 x4 is invariant
under D40 . We will therefore look at groups up to conjugation, which means that we do
not distinguish between D4 and D40 .
We have now constructed a generating invariant for the dihedral group, so we know
that if a polynomial of degree 4 has a Galois group that is D4 or a subgroup of D4 , the
combinations of the roots given by y must be invariant under the permutations of D4 . The-
orem 2.11 then tells us that y(α1 , α2 , α3 , α4 ) will give us a rational number, since we know
σ(y) = y and thus evaluating y gives σ(φ(y)) = φ(σ(y)) = φ(y) and therefore φ(y) ∈ Q

13
. By finding generating invariants for each group of a certain order, we can thus decide
whether or not the given Galois group can be a subgroup of the group we are looking at.

We will now present a list of all groups that have order up to 12, and the corresponding
generating invariants. Note that for every Sn , the only invariant polynomials are the ele-
mentary symmetric polynomials, since Sn contains all possible permutations of n variables.
To keep the list as uncomplicated as possible, we only present one elementary symmetric
polynomial in that case.
Note that when looking at Cp =< (123...p) > , with p prime, not only x1 x22 + ... +
xp−1 x2p + xp x21 is an invariant, but also x1 x23 + ... + xp−1 x21 + xp x22 etc. When p is not a
prime, we see that x1 x23 +...+xp−1 x21 +xp x22 etc. are still invariant under Cp , but also under
larger groups, and we therefore they are not generating invariants. Look for example at
y1 = x1 x23 + x2 x24 + x3 x21 + x4 x22 . We see that y1 is also invariant under D4 which is a group
properly containing C4 , so this is in contradiction with our conditions for a generating
invariant.

order group generating invariant polynomial

2 C2 = S 2 x1 + x2 , x 1 x2
3 C3 = A 3 x1 x22 + x2 x23 + x3 x21
4 C4 x1 x22 + x2 x23 + x3 x24 + x4 x21
V4 x1 x2 x23 + x21 x3 x4 + x1 x2 x24 + x22 x3 x4
5 C5 x1 x22 + x2 x23 + x3 x24 + x4 x25 + x5 x21
6 C6 x1 x22 + ... + x5 x26 + x6 x21
7 C7 x1 x22 + ... + x6 x27 + x7 x21
8 C8 x1 x22 + ... + x7 x28 + x8 x21
D4 x1 x3 + x2 x 4
9 C9 x1 x22 + ... + x8 x29 + x9 x21
10 C10 x1 x22 + ... + x9 x210 + x10 x21
D5 x1 x 2 + x2 x3 + x3 x4 + x4 x5 + x5 x1
11 C11 x1 x22 + ... + x10 x211 + x11 x21
12 C12 x1 x22 + ... + x11 x212 + x12 x21
D6 x1 x3 x 5 + x2 x4 x6
A4 x1 x2 x3 + x2 x23 x31 + x3 x21 x32 + x2 x24 x33 +
2 3

x4 x21 x33 + x4 x23 x32 + x3 x22 x34 + x1 x23 x34 +

x2 x21 x34 + x1 x24 x32 + x4 x22 x31 + x3 x24 x31
We have now created invariants for all groups up to order 12, which can be checked by
simply letting the elements of the group act on the invariant. A question that may arise is
if we can always create such an invariant. The answer is yes, and we will prove that using
the following theorem.
Theorem 3.1. Let f (x) = xm + a1 xm−1 + ... + am−1 x + am be monic with coefficients ai
in a unique factorization domain R. If f (β) = 0 for some β ∈ Q(R) (the quotient field of
R), then β ∈ R.

14
 Proof: Let β ∈ Q(R) such that f (β) = 0. Then we know we can write β = cb for some
b, c ∈ R, such that the greatest common divisor of b and c is 1. Since f ( cb ) = 0, we know
cm f ( cb ) = bm + a1 cbm−1 + ... + am−1 cm−1 b + am cm = 0. If we look at this modulo c, we find
that bm = 0(mod c) and we can thus conclude that c|bm . Since gcd(b, c)=1, we know that
c has to be a unit in R and therefore that ab ∈ R, and thus that β ∈ R.

We will now use this theorem to prove that for every subgroup of Sn , there exists a
generating invariant, which is in fact the same as a generator of the invariant field.

Theorem 3.2. Let H be a subgroup of Sn . Then there exists a polynomial F ∈ Q[x1 , ..., xn ]
such that ∀h ∈ H we have that h(F ) = F (xh(1) , ..., xh(n) ) = F (x1 , ..., xn ). Moreover, H is
the biggest group for which this applies, so for all g ∈
/ H we have that g(F ) 6= F .

Proof: Let us look at the elementary symmetric functions, given by s1 = x1 + x2 + ... +

xn , s2 = x1 x2 +x1 x3 +...+xn−1 xn , ... , sn = x1 x2 ...xn . Let f = (x−x1 )(x−x2 )...(x−xn ) =
xn − s1 xn−1 + s2 xn−2 − ... + (−1)n sn be the general polynomial of degree n and let L be
its splitting field over K = Q(s1 , s2 , ..., sn ). Then we need to extend K by x1 , x2 , ...xn , so
L = Q(x1 , ..., xn ) and we see that [L : K] = n! We denote the fixed field of L under H by
LH = {x ∈ L | σ(x) = x ∀ σ ∈ H}, and Artin’s theorem then tells us that L/LH is a finite
Galois extension with Galois group H, from which it follows that |H| = [L : LH ].
Since H is a finite group and LH is a Galois extension, and thus separable, we know
that there exists a α ∈ LH such that LH = K(α) (see [2, Thm 7.3.6, Thm 8.4.1]). Let
g(x) ∈ K[x] be the minimal polynomial of α and let us write g(x) = am xm + am−1 xm−1 +
n!
...+a1 x+a0 . We know that m = |H| and that ai ∈ K = Q(s1 , s2 , ..., sn ), and without loss of
generality we can assume that ai ∈ Q[s1 , ..., sn ]. Let us now multiply g by am−1 m and write it
in terms of y = am x to find F (y) = y m + am−1 y m−1 + am−2 am y m−2 + ... + a1 am−2 m−1
m y + a0 am .
Note that F is monic with coefficients in Q[s1 , ..., sn ] ⊂ Q[x1 , ..., xn ] and that it is the
minimal polynomial of am α. So we have found an element am α in L = Q(x1 , ..., xn ) which
is invariant under H. We now use the theorem above, with R = Q[x1 , ..., xn ] and β = am α
to state that am α ∈ Q[x1 , ..., xn ] and that am α is therefore the invariant polynomial F we
were looking for since K(F ) = K(am α) = K(α) = Q(x1 , ..., xn )H = LH .

This theorem tells us that we can always find a polynomial F ∈ Q[x1 , ..., xn ] which is
invariant under a subgroup of Sn , and which is not invariant under all other permutations
of Sn . Unfortunately, it does not tell us how to create such a polynomial. For the groups
covered in this thesis, we have stated the invariants needed above, and in the next sections
we will use them to find the Galois group.

3.2 Resolvents
Resolvents are elementary combinations of invariants, that turn out to have special prop-
erties. Resolvents are the main tool for the method we will use in Section 3.4, which is
called the descendant method or even the resolvent method. In this section, we will define

15
resolvents, look at their properties and look into several ways to calculate them.

Cohen assumes that all polynomials are monic with integer coefficients. When we
encounter monic polynomials in Q[x] we can transform them into monic polynomials in
Z[x] whose roots generate the same Galois extension. Let a be the least common multiple of
all the denominators of the coefficients, then f (x) → an f ( xa ) gives the desired polynomial.
Since Q(α) = Q(kα) ∀k ∈ Q it is easy to see that this transformation does indeed give us
the same Galois extension.
Since we will only be looking at irreducible polynomials, we know that the Galois group
is a transitive subgroup of Sn up to conjugacy. Two groups A and B are equal ”up to
conjugacy” if there is an element σ in the larger group, in this case Sn such that σAσ −1 = B.

Given a group G ⊂ Sn of which we know the Galois group is in it, we can look at H ⊂ G
and ask ourselves whether the Galois group could be a subgroup of H or not. We know
the answer is no if the evaluated invariant does not give us a rational number. We have
to keep in mind that we need to look at all the conjugates of H in G, since we randomly
choose an order of the roots, and we have to correct for that. Therefore, we will look at
the polynomial that has all conjugates of the evaluated invariant as its roots.
Definition 3.3 (Resolvent polynomial). Let f ∈ Z[x] an irreducible polynomial of degree
n and G a subgroup of Sn containing Gal(f ). Let H be a transitive subgroup of G and
let h ∈ Z[x1 , ..., xn ] be the generating invariant of H in G. We then define the resolvent
polynomial as
Y
RG (h, f )= (x − h(ασ(1) , ασ(2) , ...ασ(n) ))
σ∈G/H

where G/H denotes the set of left coset representatives of G modulo H.

Notice that Theorem 2.11 tells us that RG (h, f ) ∈ Q[x], since an element of the Galois
group only permutes the (x − h(ασ(1) , ασ(2) , ...ασ(n) )) but the product of all these elements
will stay the same. Moreover, since f ∈ Z[x] , we know that RG (h, f ) ∈ Z[x]. When we
need to calculate resolvents in the next section, we will see that for higher degree poly-
nomials with few symmetries it will be dreadful to calculate them algebraically. We can
do this numerically, and if we work with enough precision, there will be no errors. The
reason for this is the following. When we calculate the roots numerically, there will be
a certain error, and when we use these roots to compute the resolvent, the error is still
there. We can overcome this by redefining the resolvent, by rounding all coefficients to
the nearest integer. We are allowed to do so since we know the resolvent has only integer
coefficients. Note that we do need a certain accuracy to start with, if the coefficients of the
first resolvent are not close to integers, we need to use a higher accuracy when computing
the roots. If we look at even higher degree polynomials, there will come a moment that we
cannot calculate the roots with high enough accuracy to use the method mentioned above.
We then need to look at p-adic approximations, which we will not cover here.

16
 In the section above we created invariants for groups of order up to 12. We know that if
the invariant of a group A does not give us an integer number for some ordering of the roots,
that would mean that the Galois group is not a subgroup ofA. It turns √ out√the converse
√ √ is
4 4 4 4 4
not always true, for example the roots of the polynomial x − 5 are 5, i 5, − 5, −i 5
and the invariant for V4 gives us x1 x2 x23 + x21 x3 x4 + x1 x2 x24 + x22 x3 x4 = 0 but the Galois
group of x4 − 5 is D4 which is clearly not a subgroup of V4 . This inconvenience is caused
by the fact that some resolvents have a double root.

Note that if RG (h, f ) has a double root, that means that there exist σi , σj ∈ G/H such
that h(ασi (1) , ασi (2) , ...ασi (n) ) = h(ασj (1) , ασj (2) , ...ασj (n) ), then RG (h, f ) has a factor (x − a)2
which means that when we differentiate RG (h, f ) to x, we still have a factor (x − a) in
each term. Therefore, we compute the greatest common divisor (gcd) of RG (h, f ) and
its derivative, and if it is constant, we have a squarefree resolvent polynomial. In the
following theorem, we will need the notion of squarefree.

Theorem 3.4. Use the notation used in Definition 3.3. Then if RG (h, f ) is squarefree, it
has a rational root if and only if Gal(f ) is conjugate under G to a subgroup of H.

Proof: We defined the Galois group of a polynomial f ∈ Q[x] as the set of permutations
in Sn that send an element of the kernel of the evaluation homomorphism φ to its kernel
again. Note that σ ∈ Sn works on Q[x1 , ...xn ] by permuting the indices of the variables,
which means that xi → xσ(i) . Now let F be a polynomial in n variables and let us denote
F σ = F (xσ(1) , ..., xσ(n) ). Note that (F σ )τ = F τ σ . Let G ⊂ Sn be a group such that
Gal(f /Q) is a subgroup of G, and let F be an invariantSunder a subgroup H of G. Let us
look at the left coset decomposition of G by H: G = σH.
σ∈G/H
Since RG (h, f ) is squarefree the zeros of RG (h, f ) are all distinct and so we know that
F (α1 , ...αn ) = F τ (α1 , ..., αn ) if and only if σ ∈ τ H. Let us now assume F σ (α1 , ..., αn ) ∈
σ

Z. Then we know that for g ∈ Gal(f ) we have F gσ (α1 , ...αn ) = F σ (αg(1) , ...αg(n) ) =
F σ (α1 , ...αn ), and thus that F gσ = F σ . Using the fact that RG (h, f ) is squarefree, we
conclude that gσ ∈ σH and thus that there exists a h ∈ H such that gσ = σh and
therefore that g = σhσ −1 . So we conclude that Gal(f ) is conjugate under G to a subgroup
of H.
Conversely, when we know the Gal(f ) is conjugate under G to a subgroup of H, Theo-
rem 2.11 tells us that one of the conjugates of the invariant under F will give us an integer,
and thus that RG (h, f ) has an integer root.

We will make good use of this theorem, since it makes it a lot easier for us to compute
Galois groups. This theorem tells us that we only need to check for linear factors in the
resolvent to figure out if the Galois group is a subgroup of the group we are looking at.
Even though that might not always be the fastest way, we will see in Section 3.4 that this
theorem enables us to always find the Galois group.

17
3.3 Tschirnhausen transformations
We cannot use the theorem above if we have a resolvent polynomial that is not squarefree.
Since we need to use this theorem to find the Galois group, we have to transform the given
resolvent into another resolvent, while preserving the given information. Since a resolvent
is a function of two polynomials and changing h does not help, we have to change f into
another polynomial in Q[x]. The best way to do this is by using a Tschirnhausen trans-
formation.

When we want to know what the Galois group of a certain polynomial is, we actually
want to know what the Galois group of its splitting field over Q is. When the polynomial
we are interested in cannot be used in the algorithm because its resolvent is not squarefree,
we can create a new polynomial whose splitting field is the same as the one from the orig-
inal polynomial. This way, we will get exactly the same Galois group with the advantage
that we can use the algorithm.

Let f (x) be our original function of degree n and let α be a generator of the splitting
field of f over Q. Now let β be a function of α that generates the same field over Q, for
example β = α + k for k ∈ Q. We call this function g, so β = g(α). There is not a
foolproof method on how to choose β or g, we have to choose g based on our knowledge
of the original polynomial f . Now, since α is generator of the splitting field, the roots can
be expressed in terms of α. Let h(α) ∈ Q(α) be one of the roots of f (x), then we know
that f (h(α)) = 0. Since we also know that β = g(α) we can combine these two equations
to eliminate α.

There are several ways to do this. We can use linear algebra and expand all powers of
β up to β n−1 in terms of α. This, together with the equation f (h(α)) = 0 gives us n + 1
0

equations with powers of α up to n, which we can solve and thus find a polynomial j of
degree n or lower such that j(β) = 0. This is then our new minimal polynomial of the
given splitting field, and we can check if it is squarefree. If it is, we can use this polynomial.
If it is not, we have to use another β and try again.

It is however a tedious task to expand all powers of β in terms of α. A less time-

consuming approach is expanding all terms of g(x) of powers 0 up to n − 1, and look at
them modulo f (x). This gives us the same result since Q(α) = Q(x)/f (x), but takes less
time. We will then use these n + 1 equations to eliminate α and find a minimal polynomial
of β. A third method uses the Euclidean algorithm in x for the two functions −y + g(x)
and f (x), which is just another way of doing the same, but might come in handy if we
want a computer to do this for us.

18
3.4 General algorithm
In the previous sections we have looked into resolvents and Tschirnhausen transformations.
In this section, we will use all this information to form a general algorithm to find the Ga-
lois group. In his book, Cohen [1] gives us different algorithms degrees up to 7. In this
section we will state a general algorithm that works for all degrees. The underlying idea of
this algorithm is the following: we know that the Galois group of a polynomial has to be
a subgroup of Sn . To figure out which subgroup it is, we take a subgroup of Sn , call it H,
and try to figure out if the Galois group is also a subgroup of H. This way, we can peel off
the different layers inside Sn until we have found a group A of which we know the Galois
group is inside it, but this does not hold for any subgroup of A. Then we can conclude
that the Galois group is conjugate to A.

Since we want to be as efficient as possible, we want to take for H the group with
the highest probability that the Galois group is inside of it. Therefore, we take a maximal
subgroup of Sn and to increase the efficiency we can start with the one that has the highest
number of transitive subgroups. For small n, there are not that many transitive subgroups
and we only need to use the algorithm once or twice. For larger n, we will have to use
the algorithm multiple times and it can therefore take some time to find the Galois group.
The algorithm goes as follows:

Theorem 3.5. Let f ∈ Z[x] be a monic irreducible polynomial of degree n. Then the
following algorithm gives us the Galois group.

1. Let G = Sn .

2. Let H be a maximal subgroup of G.

3. Look for an invariant polynomial under H and call that polynomial F.

4. Compute G/H, the left coset decomposition.

5. Compute the roots of f in C.

6. Compute RG (F, f ) and round the coefficients to the nearest integer.

0
7. Compute gcd(RG (F, f ), RG (F, f )). If it is constant, go to 8. If it is not constant,
use a Tschirnhausen transformation on f and go back to 5.

8. Factor RG (F, f ) over Z.

9. If the resolvent has an integer root, conclude the Galois group is a subgroup of H and
go back to 2, where the new G is the old H.
If it has no integer root, choose a new maximal subgroup of G and go back to 3.
If there is no maximal subgroup left, conclude the Galois group is G.

19
 To clarify the general algorithm, let us take a look at the one Cohen suggests for n = 4
in [1]. He uses S4 as subgroup containing the Galois group (there is no other option since
this is the first time running the algorithm), and C4 as H. We know the polynomial
F = x1 x22 + x2 x23 + x3 x24 + x4 x21 is invariant under C4 , and this is the one he uses for the
resolvent. He follows the general algorithm up to step 9. He then starts looking at the
cycle types of the factorized resolvent, and uses them to conclude what the Galois group
is. This is in principle a good method, expect for the fact that when we want to use the
algorithm for degree higher than 7, we first need to compute all the different cycle types
using polynomials of which we already know the Galois group, and then hope there is only
one group that corresponds to the cycle type we found.

The advantage of the algorithm stated here is that we do not need to compute the
cycle types of the factorized resolvent for all transitive subgroups of Sn . We just go over
all maximal transitive subgroups of Sn and then use a descend method to find the Galois
group. We can of course compute all these cycle types, but for degrees > 10 this will
become tedious. Therefore it will be easier to let the algorithm run several times and
use Theorem 3.4 which tells us that the resolvent has a linear factor (and thus an integer
root) if and only if Gal(f ) is conjugate under G to a subgroup of H. This means that
if the resolvent has a linear factor, we can restart the algorithm with G0 = H and H 0 a
maximal subgroup of H. In fact, this is the most general form of the algorithm and al-
though Cohen’s algorithm is based on this theory it is not explicitly mentioned in his book.

Note that this algorithm is not necessarily the fastest way to compute the Galois group.
For specific degrees, for example n = 7, there exists faster algorithms to find the Galois
group. We decided to use this algorithm since it works for all n, even though it might
take lots of time. We can reduce the amount of time by using the properties of Galois
groups that we stated before. For example, if we know that disc(f ) is a square, we can
skip roughly half of the groups before even starting the algorithm, because we only have
to consider even groups. The same goes for Dedekind’s theorem. If we find out that our
Galois group has an element of a certain rare cycle structure, we only have to consider
the groups that have this cycle structure. By using the knowledge we have wisely, we can
considerably reduce the time this algorithm takes.

20
4 Inverse Galois Theory
In the two previous chapters we have looked at polynomials and their corresponding Galois
groups. We have discussed several options to find the Galois group, or to at least reduce
the number of possibilities. In this chapter, we would like to do the opposite. We want to
know whether all groups can appear as a Galois group of a certain polynomial, and how
we can find these polynomials for each group. This question has yet to be answered, and
most of the theory is far beyond the range of this thesis. We will however look at the basics
of inverse Galois theory, and state a method to find polynomials of degree 3 and 4 that
have the desired Galois group. We will see that there are different approaches, depending
on which group we are looking at, and that some group theory is needed to make sure we
create the right polynomial.

4.1 Galois groups for degree 3 polynomials

For a degree three polynomial, we know the Galois group is either A3 = C3 or S3 , since
these are the only two transitive subgroups of S3 . Theorem 2.14 tells us that the Galois
group is A3 if and only if the discriminant of the polynomial is a square. Let f be a degree
three polynomial, we write f = x3 + ax2 + bx + c, with a, b, c ∈ Q. Note that we can also
choose a, b, c ∈ Z, depending on what we are looking for. Let α1 , α2 , α3 be the zeroes of
f , then we have defined the discriminant as ((α1 − α2 )(α1 − α3 )(α2 − α3 ))2 . We can also
express the discriminant in terms of a, b, c to get
disc(f ) = a2 b2 − 4b3 − 4a3 c − 27c2 + 18ab
It will be difficult to find general a, b, c such that this term becomes a square. We therefore
start by setting one of a, b, c to zero. Note that we cannot set c = 0 since in that case x = 0
will be a root for our polynomial, which means that the Galois group is not transitive any
more and therefore cannot be A3 . So we can only simplify this discriminant by taking
a = 0 or b = 0.

Let us start with a = 0. We then see that the discriminant becomes disc(f ) = −4b2 −
27c2 . When b, c are rational we know that there exists a l ∈ Q such that b = −l · c. If we
want b, c to be integers, we limit our possibilities by assuming b = −l · c, where l ∈ Z. We
will then find
disc(f ) = −4(−lc)3 − 27c2
= c2 (4l3 c − 27)
Since we know the product of two squares is a square again, we only need to make sure
the second term is a square. Note that for integer coefficients, 4l3 c will always be even, so
since 27 is odd, we know this term will be odd. We therefore solve 4l3 c − 27 = (2k + 1)2
with k ∈ Z to find
k2 + k + 7
c=
l3

21
 2 2
This will give us the polynomial f (x) = x3 − k +k+7 l2
x + k +k+7
l3
. Note that for l = 1
we will always find that f ∈ Z[x] and we have thus found infinitely many solutions. For
l > 1 this is not the case. We can make the coefficients integer by using the transformation
f (x) → l3 f ( xl ), but this will give us the same result as using l = 1. For specific l we can
find polynomials in Z[x], for example, l = 3, k = 4 gives us f (x) = x3 − 3x + 1 which has
integer coefficients, and so do all k of the form 4 + 27d for d ∈ Z when l = 3.

If we are looking for polynomials with rational coefficients this will give us an endless
list of possibilities. We further generalize this process by simply taking 4l3 c − 27 = k 2 ,
where k ∈ Q and will then find
k 2 + 27
c=
4l3
k2 +27 k2 +27
which gives us the polynomial f (x) = x3 − 4l2
x + 4l3
which is clearly in Q[x].

A general way to find polynomials with b = 0 that have Galois group A3 is by using the
transformation x → x1 . In that case, we see that x3 + ax2 + c goes to 1 + ax + cx3 which
turns in to x3 + ac x + 1c . This gives us a polynomial of the form x3 + bx + c and we have
just seen how to create infinitely many polynomials of that form.
Moreover, we can choose any polynomial we just created and transform it by looking
at f (x − p) instead of f (x). We can therefore also create polynomials for which neither a
nor b is zero. We can now conclude that we can make infinitely many polynomials of both
forms with Galois group A3 .

4.2 Galois groups for degree 4 polynomials

We will look at the transitive subgroups of S4 in this section. We have already encountered
polynomials that have some of these groups as Galois group, for example x4 − 10x2 + 1
with Galois group V4 . Nevertheless, we will in this section state some more general ways
of finding polynomials with our desired Galois groups.

We will start with the group we already encountered, namely V4 . We know that
V4 ' C2 × C2 and therefore that √ if √
the Galois group is isomorphic to V4 , the splitting
field will be of the form L = Q( a, b) where a, b ∈ Z. We do have to be careful here,
since when a or b are squares, this is not a degree four extension, and neither is that the
case when ab or ab is a square in Q.

The most evident way of creating such a splitting field is by looking at (x2 − a)(x2 − b),
but since this is a reducible polynomial it does not have a transitive √Galois group.
√ We
thus have to look at the polynomial whose√ zeroes
√ are √ combinations
√ of a and b. Let us
look at the polynomial whose roots are a ± b, − a ± b, so that we have four different
roots. This polynomial can be written as f (x) = x4 − 2(a + b)x2 + (a − b)2 and gives us
infinitely many possibilities since we can use any integers for a and b. Of course, we can

22
 √ √ √ √
generalize this by looking at the polynomial whose roots are a ± b + c, − a ± b + c,
for a, b, c ∈ Z, where a and
√ b√are both not square, and neither is their quotient. Note that
this again gives L = Q( a, b) as splitting field and that adding a term c to the roots
does not change the Galois group. We find that the corresponding polynomial is:

f (x) = x4 − 4cx3 + (6c2 − 2(a + b))x2 + 4c(a + b − c2 )x + (a − b)2 + c2 (c2 − 2(a + b))

Proof that this polynomial has Galois group V4 : We first prove that f is irre-
ducible. We know that none of the roots is an integer, so we do not need to look for linear
factors. When looking for quadratic factors, we need the product and the sum of two √ roots
√
to be an integer. Since for all possible products of two roots, we always have factors a, b
left in the product, we know that they are never integers. We therefore know that f is
irreducible. We know that [L : Q] = 4, so our Galois group can be either√ V4√or C4 . √Note
that the splitting field L has three subfields of degree 2, namely Q( a), Q( b), Q( ab).
Theorem 2.16 now tells us that the Galois group has to have three subgroups of order 2.
We know that C4 =< (1234) > has only one subgroup of order 2, namely the one gener-
ated by the permutation (13)(24) whereas V4 = {id, (12)(34), (13)(24), (14)(23)} has three
subgroups of order two, which are generated by its nontrivial elements. We can therefore
conclude that the Galoisgroup of this polynomial has to be V4 .

Now let us look at D4 . Note that |D4 | = 8, and that it has normal subgroups of order
4. We therefore know that if the Galois group of a polynomial is D4 , there is at least one
quadratic extension
√ of Q in the splitting field. Let us assume this quadratic extension is of
the form Q( b) for some b ∈ Z that is not zero, nor a square. In order to√find a polynomial
with Galois group D4 , we need to find a field extension of degree 4 of Q( b). We also want
to generate a degree 4 polynomial with rationalpcoefficients. Let us look at the polynomial
√ p √ √ √
with the zeroes a + b ± c + b, a − b ± c − b, for some a, c ∈ Z. Explicitly, this
polynomial turns out to have rational coefficients, since:
√ √ √ √
q q
f (x) = (x − (a + b + c + b)) · (x − (a + b − c + b))·
√ √ √ √
q q
(x − (a − b + c − b)) · (x − (a − b − c − b))
= x4 − 4ax3 + (6a2 − 2b − 2c)x2 + (4ab + 4ac − 4a3 − 4b)x +
(a4 − b + 4ab − 2a2 b + b2 − 2a2 c − 2bc + c2 )
√ p √
To create its splitting field, we need to extend Q first by b and then by c + b and
p √ p √ √
c − b. We will here only look at b, c such that c + b ∈ /√Q( b). This will give us a
c+√b
√
field extension of degree 4 or 8, depending on whether or not c− b is a square in Q( b). If it
√ √ √ √ √ p √
c+ b c+ b
is, that means that √ √ ∈ Q( b) and thus that √ √ is also an element of Q( c − b),
c− b c− b
p √ √ √ p √ p √
and therefore that c + b = √c+√b · c − b ∈ Q( c − b). This tells us that we only
c− b

23
 √ p √
need to extend Q( b) by c − b to find our splitting field, and therefore that [L : Q] = 4.
√
c+√b
√ √ √
Note that when c− is a square in Q( b), that also means that (c+ b)(c− b) = c2 −b
√ b
2
√ 2 2
is a square in Q( b). This means, that we can write√ 2 c − b = (u + v b) Note that c − b
is a rational number, which tells us that (u + v b) also has to be rational. This is only
possible when either u or v are zero, and therefore when c2 − b = u2 or when c2 − b = bv 2 .
The following theorem will tell us what the Galois group is under those circumstances.
Theorem 4.1. Let a, b, c ∈ Z such that b is not a square and
√ √ √ √
q q
f (x) = (x − (a + b + c + b))(x − (a + b − c + b))
√ √ √ √
q q
(x − (a − b + c − b))(x − (a − b − c − b))
= x4 − 4ax3 + (6a2 − 2b − 2c)x2 + (−4a3 − 4b + 4ab + 4ac)x+
(a4 − b + 4ab − 2a2 b + b2 − 2a2 c − 2bc + c2 )

Let L be its splitting field, then Gal(L/Q) is

√
c+√b
√
• D4 if c− b
is not a square in Q( b).

• V4 if b = c2 − u2 for some u ∈ Z.
c2
• C4 if b = 1+v 2
for some v ∈ Z.
Proof: We know that f is irreducible since all the roots√are clearly not integers, and
all the different
√
products of two roots√
still contain factors of b and other roots. We know
c+√b
that when c− b is not a square in Q( b), that [L : Q]=8, which means that D4 has to be
the Galois group, since there is no other transitive subgroup of S4 with order 8.
√
c+√b
√
When c− b
is a square in Q( b), it is not so easy to see what the Galois group is. Note
that since V4 and C4 are the only transitive subgroups of S4 of order 4, we know that in
both cases one of them has to be the Galois group. In the appendix, it is shown that when
c2
b = 1+v 2 , substituting the roots of f in the invariant of C4 and evaluating this for all 6

coset permutations gives us a rational number twice. Furthermore all 6 coset permutations
of this invariant give us different outcomes, which means that the resolvent is squarefree.
We can therefore use Theorem 3.4 which tells us that the Galois group is a subgroup of
C4 , but since the Galois group has order 4 we know it has to be C4 itself.

When we look at b = c2 − u2 we encounter the difficulty that the resolvent is not

squarefree when we insert the roots into the invariant for V4 . We therefore take another
approach. We know that when the Galois group is V4 , there are three subgroups of order
2, which are generated by the individual non-trivial elements of V4 . This tells us that the
roots must be invariant under x1 x2 + x3 x4 , x1 x3 + x2 x4 and x1 x4 + x2 x3 , and therefore
that all three of these must give us a rational number. Since this is the case, and we

24
find no rational number when we evaluate the roots in the invariant for C4 we know that
the Galois group is V4 . The explicit calculations of this proof are given in the appendix.
p √
Note that we can make this theorem even more general by replacing c ± b by
p √ √
c+d√b
c ± d b with d ∈ Z. In that case, the Galois group will be D4 if if c−d b is not a square
√ 2 2 2
in Q( b), V4 if b = c d−u
2 for some u ∈ Z and C4 if b = d2c+v2 for some v ∈ Z. The proof of
this version of the theorem is completely analogous to the proof above, and will therefore
not be repeated.

With this theorem, it is possible to create infinitely many polynomials with Galois
group C4 , V4 or D4 , since we can choose for a any integer we want, then choose u or v in
case we want the Galois group to be V4 or C4 and choose c such that b will be a non-square
integer, or choose a and b randomly for D4 and then pick c such that the field extension
has order 8. Note that the formula given for V4 on page 23 still works, and that we can
use both theorems to create a polynomial that has Galois group V4 .

The only transitive subgroup left is A4 . To create a polynomial whose Galois group
is a subgroup of A4 , we know we need to ensure that its discriminant is a square. For a
general degree four polynomial, f (x) = x4 + ax3 + bx2 + cx + d, the discriminant is given
by

Disc(f ) = 256d3 − 192acd2 − 128b2 d2 + 144bc2 d − 27c4 + 144a2 bd2 − 6a2 c2 d − 80ab2 cd +
18abc3 + 16b4 d − 4b3 c − 27a4 d2 + 18a3 bcd − 4a3 c3 − 4a2 b3 d + a2 b2 c2

Since it’s impossible to algebraically make this discriminant a square, we will set a = 0
(we can always do this by transforming f (x) → f (x − a)) , and look at two possibilities:
First, we will consider b to be zero, and c = λd, where λ ∈ Q. We then find

disc = 256d3 − 27λ4 d4 = u2

1 u
256 − 27λ4 = ( 2 )2 = v 2
d d
256
d= 2
v + 27λ4
256 256
This gives us the polynomial f (x) = x4 + λ v2 +27λ 4 x + v 2 +27λ4 which has rational coefficients

for all λ, v ∈ Q.
The second possibility is setting c = 0, b = µd for some µ ∈ Q, which gives us:

disc = 256d3 − 128µ2 d4 + 16µ4 d5 = u2

u2
16 − 8µ2 d + (µ2 d)2 = 3
d

25
Since we can choose u ∈ Q randomly, let us choose u = 0. We then find
16 − 8µ2 d + (µ2 d)2 = 0
(µ2 d − 4)2 = 0
4
d= 2
µ
This gives us the polynomial f (x) = x4 + µ4 dx2 + µ42 , which has rational coefficients for
all µ ∈ Q. Unfortunately, this polynomial always has Galois group V4 (since the three
invariants for D4 give us three rational values, namely µ4 , − µ4 and − 4d
µ
). The same thing
2 1 16
happens if we choose for example u = 3µd and find d = µ2 , d = µ2 . Therefore, this
polynomial does not give us the desired solution.

We have found a general way to create a polynomial whose Galois group is a subgroup
of A4 . Since V4 is the only transitive subgroup of A4 that can correspond to a degree four
polynomial, we need to check that the polynomials we create this way do not have Galois
group V4 . The fastest way to do that is by either computing the invariant for V4 for all
cosets of A4 /V4 , or by computing the values of x1 x3 + x2 x4 , x1 x2 + x3 x4 , x1 x4 + x2 x3 ,
where we know that the Galois group is V4 if and only if all three of these give a rational
number. Unfortunately there is no general way to do this, so after choosing the variables λ
and v it takes a quick check to make sure we have really created a polynomial with Galois
group A4 and not one with Galois group V4 .

4.3 Symmetric functions

After looking at Theorem 4.1 we could ask ourselves: Can every polynomial with Galois
group V4 , C4 or D4 be constructed this way? To answer this question, we will create in
the most general way a polynomial that has one of these groups as Galois group. We
will therefore look at the elementary symmetric functions. Since we want to examine the
groups that correspond to degree 4 polynomials, we have n = 4 and
s1 = x1 + x2 + x3 + x4
s2 = x1 x2 + x1 x3 + x1 x4 + x2 x 3 + x2 x4 + x3 x4
s3 = x1 x2 x 3 + x1 x2 x4 + x1 x3 x4 + x2 x3 x4
s4 = x1 x2 x3 x4
We know that any polynomial f with roots x1 , x2 , x3 , x4 has these elementary symmetric
functions as coefficients, namely f = (x − x1 )(x − x2 )(x − x3 )(x − x4 ) = x4 − s1 x3 + s2 x2 −
s3 x + s4 . Let us now look at the invariants of D4 and its conjugates, namely
F1 =x1 x3 + x2 x4
F2 =x1 x2 + x3 x4
F3 =x1 x4 + x2 x3

26
When constructing the resolvent of S4 over D4 , we use these functions to create
RS4 (D4 , f ) = (y − F1 )(y − F2 )(y − F3 ) = y 3 − s2 y 2 + (s1 s3 − 4s4 )y − (s23 − 4s2 s4 + s21 s4 ).
Now, we know that the Galois group of f is a subgroup of D4 if and only if the resolvent
has a rational solution and all three F ’s are distinct. Since this is a sufficient condition, we
can choose whichever rational root we like. Let us take y = 0 and make sure this is a root
of the resolvent. When y = 0, we see that RS4 (D4 , f )(0) = (s23 − 4s2 s4 + s21 s4 ), which has
to be zero. We can now choose s1 , s2 , s3 ∈ Q randomly, and since the resolvent is linear in
s4 , we can always find a rational solution for s4 . We can do this for all y ∈ Q, and find
infinitely many solutions.

We have now created a polynomial with rational coefficients (since s1 , s2 , s3 , s4 are ra-
tional) that has a subgroup of D4 as Galois group. We can find the zeros numerically, and
then compute F1 , F2 , F3 . We know that V4 has three subgroups of order 2, and therefore
all three F ’s have to give distinct rational outcomes if the Galois group of the polynomial
f is V4 . If we have one rational outcome and two irrational ones, we have constructed a
polynomial with either C4 or D4 as Galois group. We can now check whether this polyno-
mial has Galois group C4 or D4 by using an invariant for C4 . Before doing this, we have
to renumber the roots such that F1 is the invariant that has a rational outcome. We know
F10 = x1 x22 + x2 x23 + x3 x24 + x4 x21 is invariant under C4 and that D4 /C4 = {id, (13)}. We
therefore only need to check F 10 and F20 = x3 x22 + x2 x21 + x1 x24 + x4 x23 . If at least one of
them gives us a rational number, and F10 6= F20 , we know the Galois group is C4 , and if
both are irrational it is D4 .

We would however like to know beforehand which of the three possible groups will be
the Galois group of a polynomial we construct this way, without having to compute the
roots and using resolvents. We know at least one of the three F1 , F2 , F3 has to be rational.
Let us say F1 is always rational.
When we want the Galois group to be V4 , we need that the other two invariants for
D4 are also rational numbers. Note that they are not allowed to give the same number,
since this would mean our resolvent is not squarefree and we cannot use Theorem 3.4.
Therefore, we want F1 − F2 = (x1 − x4 )(x2 − x3 ) 6= 0, F1 − F3 = (x1 − x2 )(x3 − x4 ) 6= 0 and
F2 − F3 − (x1 − x3 )(x2 − x4 ) 6= 0. This is the case if and only if the discriminant of f is not
zero. We therefore only consider polynomials whose discriminant is not zero. This is the
same as only considering irreducible polynomials, which fortunately was already our plan.

It turns out just choosing F2 , F3 ∈ Q does not give a general rational solution for our el-
ementary symmetric functions, which we want to be rational since they are the coefficients
of the polynomial we are looking for. Neither does choosing only F2 ∈ Q. We therefore try
an alternative approach. We can always set s1 = 0 without loss of generality, since we can
just use the transformation (x − c) for some c ∈ R. We know that F1 + F2 + F3 = s2 (since
this is the coefficient of y 2 in the resolvent), which has to be rational in order to create a
polynomial f with rational coefficients. Since we also know F1 is rational, we know F2 + F3
is rational for any polynomial that has Galois group D4 , C4 or V4 . If we can now find a

27
way to make sure F2 − F3 ∈ Q, we know both of them are rational, and the Galois group
is V4 .

The way to do this is the following. We can express (F2 − F3 )2 in terms of our rational
functions, namely
(F2 − F3 )2 = −s22 + 4s1 s3 − 16s4 − 2s2 F1 + 3F12
Since we want (F2 − F3 ) to be a rational number, we want (F2 − F3 )2 to be a square. We
also want the resolvent of D4 expressed in terms of elementary symmetric functions to be
zero when we evaluate it at the point y = F1 . We now have two equations in the same
variables and can therefore use the resultant to eliminate the variable s4 , which occurs only
linearly in both equations. We then find that in order for (F2 − F3 )2 to be equal to u2 and
for RS4 (D4 , f )(F1 ) = F13 − s2 F12 + (s1 s3 − 4s4 )F1 − (s23 − 4s2 s4 + s21 s4 ) to be zero, we need
to satisfy
4(s32 + 4s23 + s22 F1 − s2 F12 − F13 + s2 u2 − F1 u2 ) = 0
Note that we can also write this as (s2 + F1 )(s22 − F12 ) + (s2 − F1 )u2 + 4s23 = 0, which
gives us (s2 − F1 )((s2 + F1 )2 + u2 ) + 4s23 = 0. We can now choose u, s3 ∈ Q and also
(s2 + F1 ) ∈ Q, and then find some (s2 − F1 ) ∈ Q since this term occurs only linearly. Note
that the only choices we are not allowed to make are u = 0, since then F2 = F3 , or u, s2
such that s2 ± u = 3F1 since then F1 = F2 (in the + case) or F1 = F3 (in the − case). We
can then find s2 and F1 using s2 + F1 and s2 − F1 . This way, we have found s2 , s3 , F1 , u
and can find F2 , F3 by using F1 + F2 + F3 = 0 and F2 − F3 = u, and use the formula for
(F2 − F3 )2 to find s4 . We have now found the most general polynomial that has Galois
group V4 , since we know f = x4 + s1 x3 + s2 x2 + s3 x + s4 and any polynomial that has
Galois group V4 has to meet the necessary and sufficient condition that all F are rational
and distinct.

To create the general polynomial with Galois group C4 we can do something simi-
lar. Note that F2 − F3 = (x1 − x3 )(x2 − x4 ), which is not invariant under all permu-
tations of C4 . The same goes for (x1 − x2 + x3 − x4 ), but the product of these two,
g = (x1 − x3 )(x2 − x4 )(x1 − x2 + x3 − x4 ) is invariant under all permutations of C4 . Since
it is not invariant under D4 (the action of (13) turns g into −g), nor under V4 (since
(12)(34)g=−g), we know that if we make sure the resolvent is squarefree, we created the
most general polynomial with Galois group C4 . We therefore again need our polynomial f
to have a non-zero discriminant and when running the algorithm the second time there are
only two cosets, {id, (13)}. We can use g as generating invariant, and then want to make
sure g 6= −g. We therefore need that g 6= 0, so (x1 − x3 )(x2 − x4 )(x1 − x2 + x3 − x4 ) 6= 0.
We know the first part is not zero since the discriminant of f is non-zero, and on top of
that we put the condition that x1 + x3 6= x2 + x4 .

Now, we can write g 2 = (x1 −x3 )2 (x2 −x4 )2 (x1 −x2 +x3 −x4 )2 in terms of the elementary
symmetric functions and F1 to find g 2 = (−s22 +4s1 s3 −16s4 −2s2 F1 +3F12 )(−s21 +4s2 −4F1 )

28
We again want that RS4 (D4 , f )(F1 ) = F13 − s2 F12 + (s1 s3 − 4s4 )F1 − (s23 − 4s2 s4 + s21 s4 ) is
zero, and that g = (x1 − x3 )(x2 − x4 )(x1 − x2 + x3 − x4 ) is part of the ground field. We
therefore want g 2 to be a square in Q, and thus want g 2 − u2 to be zero for some non-zero
rational number u. Since we work in Q we can just as well set g 2 − 4u2 = 0. Taking the
resultant of these two equations to eliminate s4 and again setting s1 = 0 gives us that

16(s2 − F1 )(s32 + 4s23 + s22 F1 − s2 F12 − F13 + u2 ) = 0

This means that either s2 = F1 which means that F2 = F3 = 0, and is in contradiction

with our conditions or (s32 + 4s23 + s22 F1 − s2 F12 − F13 + u2 ) = 0. We can also write the latter
equation as (4s23 + u2 + (s2 − F1 )(s2 + F1 )2 = 0. We can now choose s3 , u, (s2 + F1 ) ∈ Q\{0}
to find (s2 − F1 ) ∈ Q \{0}. With (s2 + F1 ) and (s2 − F1 ) we can find s2 and F1 and
therefore all variables. We know all s1 = 0, s2 , s3 and can find s4 by g 2 = (−s22 + 4s1 s3 −
16s4 − 2s2 F1 + 3F12 )(−s21 + 4s2 − 4F1 ) = 4u2 . We can now conclude that we have found
the most general polynomial with rational coefficients that has Galois group C4 . We might
even say we have also done so for D4 , since all polynomials that are constructed in the way
we explained at the beginning of this section, and do not satisfy the conditions for V4 or
C4 automaticcaly have Galois group D4 .

4.4 Subgroups and subfields of order 2

√
As we have seen, the p Galois group of the degree 4 polynomial with roots a + b ±
p √ √ √ 2 2
c + d b, a − b ± c − d b , where a, b, c, d ∈ Q is V4 if we can write b = c d−u 2
2
for some u ∈ Q, C4 if we can write b = d2c+v2 for some v ∈ Q and D4 in all other cases.
Now we know that both V4 and D4 have three subgroups of order 2, and in this section
we will see what the corresponding subfield look like, using the knowledge we already have.
2 2
Let us first look at V4 . We know that if the Galois group is V4 , b = c d−u
2 . We also know
p √ p √
that the splitting field L has degree 4 over Q and contains both c + d b√and c − d b.
Therefore it is obvious that one of the intermediate fields is of the form Q( b). Let us now
look at the other two intermediate fields, and denote them by M1 , M2 . We know they are
quadratic extensions of the ground field Q, which means that the square of the generator
has to be a rational number. Also, there has to be a degree 2 polynomial withpcoefficients
p √ √
in M which generates L (which is equal to generating both c + d b and c − d b),
such that Gal(M/Q) = C2 .
√ √
We now claim those two intermediate fields are Q( c + u) and Q( c − u). Note that
since c and u are rational numbers, these clearly are two degree 2 extensions of Q. On the
other hand, we see that

29
 √ √ q p q p
2c + 2u ± 2c − 2u = 2c + 2 c − (c − u ) ± 2c − 2 c2 − (c2 − u2 )
2 2 2
s r s r
c 2 − u2 c2 − u2
= 2c + 2 c2 − d2 ( ) ± 2c − 2 c 2 − d2 ( )
d2 d2
√ √
q q
= 2c + 2 c − d b ± 2c − 2 c2 − d2 b
2 2
r
√ q √ √ √
= (c + d b) + 2 (c + d b)(c − d b) + (c − d b)±
r
√ q √ √ √
(c + d b) − 2 (c + d b)(c − d b) + (c − d b)
rq rq
√ √ √ √
q q
= ( c + d b + c − d b) ± ( c + d b − c − d b)2
2

√ √ √ √
q q q q
= c + d b + c − d b ± ( c + d b − c − d b)
√
q
=2 c ± d b
√ √ p √ p √
Therefore, we can conclude that√ Q( 2c + 2u, √ 2c − 2u) = Q( c + d b, c − d b) = L,
and it is then obvious that Q( c + u) and Q( c − u) only need a degree 2 field extension
to form L, which means these are the two fields we are looking for.

Let us now focus on the subfields of degree 2 when the Galois group is D4 . We will
investigate this using the symmetric functions. We denote the field containing the ele-
mentary symmetric functions in 4 variables by K, so K = Q(s1 , s2 , s3 , s4 ), and know that
when the Galois group of a polynomial f is D4 , the field of invariants is K(F1 ), where
F1 = x1 x3 + x2 x4 , the invariant of D4 . We are now looking for three degree 2 field exten-
sions of K(F1 ) that are intermediate fields of K and L = Q(x1 , x2 , x3 , x4 ). We will thus
look for field extensions of Q in terms of the xi .

We now claim these field extensions are M1 = Q(x1 − x2 + x3 − x4 ), M2 = Q((x1 −

x3 )(x2 − x4 )) and M3 = Q((x1 − x3 )(x2 − x4 )(x1 − x2 + x3 − x4 )). We can see this by noting
that (x1 − x2 + x3 − x4 )2 is an element of K(F1 ) since it is not changed by the permutations
of D4 , and the same goes for (x1 − x3 )2 (x2 − x4 )2 . Since we cannot express one of those in
terms of the other, we know they are two separate field extensions, and the third quadratic
field extension will be formed by the product of these two.
To make this more explicit, we can write

(x1 − x2 + x3 − x4 )2 = s21 + 4(s2 − F1 )

(x1 − x3 )2 (x2 − x4 )2 = s22 − 3F12 − 4s1 s3 + 2F1 s2 + 16s4

From here, we can see that both squares clearly are elements of the ground field K(F1 ),

30
since we can express them in terms of the symmetric functions and F1 .

When we have found a polynomial with Galois group D4 , we can thus compute the
roots
p and use them to calculate
p the degree 2 splitting fields over Q, namely
Q(ps21 + 4(s2 − F1 )), Q( s22 − 3F12 − 4s1 s3 + 2F1 s2 + 16s4 ) and
Q( (s21 + 4(s2 − F1 ))(s22 − 3F12 − 4s1 s3 + 2F1 s2 + 16s4 )). Note that the (x1 −x2 +x3 −x4 )
and (x1 − x3 )(x2 − x4 ) are elements of the invariant field of V4 , wheras their product
(x1 − x2 + x3 − x4 )(x1 − x3 )(x2 − x4 ) is an element of the invariant field of C4 , just like
we used in last section. This means we can construct a D4 field extension either by one of
the first two C2 extensions and a V4 extension on top of that, or by extending the ground
field Q first by the product of the two roots, and then create a C4 extension of that field.

We have now used group theoretic properties to state the type of field extensions that
will occur in Galois groups. This relation between group theory and field theory is a very
special one between two subjects that may not have very much in common when first
introduced. It is also a very useful one, since groups are at some points better understood
than fields, and so we can use our group theoretic knowledge to find out more about fields,
which occur in i.a. analysis, number theory and topology.

4.5 Overview of developments in Inverse Galois Theory

Although Évariste Galois submitted his paper on what would later be called Galois theory
in 1830 (at that time he was 18 years old), it took until the end of the 19th century before
people started acknowledging an understanding the meaning and importance of his ideas.
This was mainly due to the fact that the paper was rejected for publication combined with
the fact that he died two years after submitting it. By the time Galois theory became a
well-known subject, the following question arose:

Is every finite group realizable as the Galois group of a Galois extension of Q?

This question has not been solved yet, and there is no general conjecture on whether the an-
swer should be yes or no. Nevertheless, there have been quite some important results since
this question has emerged, and we will describe a summary of this evolution. Note that
most of the results are based on Hilbert’s Irreducibility theorem and the rigidity method,
which are topics far above the reach of this thesis. Notwithstanding, we have already seen
that for degree 3 and 4 we managed to find general polynomials for some groups, and have
therefore proven much more than only the existence problem.

The first theorem is named after Leopold Kronecker and Heinrich Martin Weber and
uses the definition of a cyclotomic field. This is a field extension of Q by a nth root of
2πi
unity, which means that we extend Q by ζn = e n . So the nth -cyclotomic field is denoted
by Q(ζn ).

31
Theorem 4.2 (Kronecker-Weber). Every finite abelian group G occurs as a Galois group
over Q. To be more specific, if G is a finite abelian group, it is the Galois group of a
subfield of the cyclotomic field Q(ζn ) for some n ∈ N.

There are several ways to prove this theorem, a more elementary one is given by M.J.
Greenberg [5]. This theorem tells us that there is a set of groups, namely the set of finite
abelian groups, for which we know that the Inverse Galois theorem is true. Hilbert proved
at the end of the 19th century that the same is true for Sn and An for every n ∈ N, and
in 1937 A. Scholz and H. Reichard proved that for every odd prime p, every finite p-group
occurs as a Galois group over Q. A p-group is a group in which each element has a power
of p as its order. For example, C4 is a p group for p = 2 since all non-trivial elements have
order 2 or 4. So, before the second world war, there were already quite some groups for
which the Inverse Galois Problem was solved.

After World War II, some improvements were made, and in 1989 Shafarevich proved
that every solvable group occurs as a Galois group over Q. A group is called solvable if
it can be constructed by abelian extensions of an abelian group. Alternatively defined, a
group G is called solvable if there exists a set of subgroups id = G0 < G1 < ... < Gk = G
such that Gj−1 is normal in Gj , and Gj /Gj1 is an abelian group. All finite p-groups are
solvable, so in a way this is a generalization of Scholz and Reichard’s theorem. In the last
two decades, there are several individual groups, like the Matthieu groups or the Monster
group, for which mathematicians have proven that the Inverse Galois problem is true.
Nevertheless, it is still not clear whether or not it is true for all groups, and this problem
remains one of the most interesting ones in abstract algebra.

32
5 Conclusion
In the first chapter we have seen the basics of Galois theory as taught in most under-
graduate courses. Moreover, we have looked into ways to reduce the number of options
for the Galois group, for example by using Dedekind’s theorem or by testing whether or
not the discriminant is a square. At the end of this first chapter we have looked into the
fundamental theorem, which we used a lot later on. This chapter contained of results and
theorems which have been known since a long time, but were required to advance in this
topic.

In the second chapter we have investigated the theory behind Cohen’s algorithms. We
have stated a number of invariants and seen that is always possible to find a generating
invariant. We used that to find out how resolvents work and why we need to use them
instead of just computing invariants (Thm. 3.4). We then used this knowledge together
with a small description of Tschirnhausen transformations to create a general algorithm to
find the Galois group for polynomials of all degrees. When using this algorithm in practice
we only need to make a list of all transitive subgroups of Sn and their generating invariants,
and will then find the Galois group.

In the third chapter we created infinitely many polynomials with rational coefficients
for each of the groups A3 , V4 , C4 and A4 . For A3 , we found the polynomials f (x) =
2 k2 +27
x3 − k 4l+27
2 x + 4l3
for all k, l ∈ Q. When looking at degree 4, we saw that f (x) =
x − 4cx + (6c − 2(a + b))x2 + 4c(a + b − c2 )x + (a − b)2 + c2 (c2 − 2(a + b)) has Galois group
4 3 2

V4 , where we can choose a, b, c ∈ Z to find a polynomial with integer coefficients if needed,

and otherwise choose a, b, c ∈ Q. The polynomial f (x) = x4 − 4ax3 + (6a2 − 2b − 2c)x2 +
(−4a3 − 4b + 4ab + 4ac)x + (a4 − b + 4ab − 2a2 b + b2 − 2a2 c − 2bc + c2 ) with a, b, c ∈ Q
turned out to have Galois group V4 if we can write b = c2 − u2 for some u ∈ Q, C4 if we
c2
can write b = 1+v 2 for some v ∈ Q and D4 in all other cases. Moreover, we found that the
256 256
polynomial f (x) = x4 + λ v2 +27λ 4 x + v 2 +27λ4 has a square discriminant for all v, λ ∈ Q and

therefore either has Galois group A4 or V4 .

By looking at elementary symmetric functions we found out more about the structure
of the field extensions created when the Galois group is one of V4 , C4 or D4 . We first
constructed the most general polynomials that have one of these groups as Galois group.
We then used this method to look deeper into the field extensions we construct while
executing this process. It turned out we were able to describe what the degree 2 subfields
of V4 and D4 extensions looked like, and moreover which degree 2 extension of the ground
field corresponded with the invariant field of C4 and which two to the invariant fields of
V4 , when creating a D4 extension. We ended this thesis by describing the developments in
Inverse Galois Theory, which will continue being an interesting field, since it has enough
unresolved problems for the next few years.

33
6 Appendices
6.1 Appendix 1: Proof of Theorem 4.1 for V4
In[106]:= g@x_D := Hx - Ha + Sqrt@bD + Sqrt@c + Sqrt@bDDLL Hx - Ha + Sqrt@bD - Sqrt@c + Sqrt@bDDLL
Hx - Ha - Sqrt@bD + Sqrt@c - Sqrt@bDDLL Hx - Ha - Sqrt@bD - Sqrt@c - Sqrt@bDDLL;
b = c^2 - u^2 ;
Expand@g@xDD
Out[108]= a4 - 2 a2 c + 4 a c2 - 2 a2 c2 - 2 c3 + c4 + u2 - 4 a u2 + 2 a2 u2 + 2 c u2 - 2 c2 u2 + u4 - 4 a3 x +
4 a c x - 4 c2 x + 4 a c2 x + 4 u2 x - 4 a u2 x + 6 a2 x2 - 2 c x2 - 2 c2 x2 + 2 u2 x2 - 4 a x3 + x4

In the cell above we defined the function g(x) and have put the restricion on b. We see that we get a
function with integer coefficients. In the cell below we will define the roots and the three different
invariants that correspond to the three subgroups of V4 of order 2. We know the Galois group is a
subgroup of V4 if and only if all three give us an integer number, and that is the case. (Note that

c- c2 - u2 c+ c2 - u2 is nothing else than u2 and therefore the absolute value of u,

which is an integer.)
In[109]:= x1 = a + Sqrt@bD + Sqrt@c + Sqrt@bDD;
x2 = a + Sqrt@bD - Sqrt@c + Sqrt@bDD;
x3 = a - Sqrt@bD + Sqrt@c - Sqrt@bDD;
x4 = a - Sqrt@bD - Sqrt@c - Sqrt@bDD;
Inv1 = x1 x3 + x2 x4;
Inv2 = x1 x2 + x3 x4;
Inv3 = x1 x4 + x2 x3;
Expand@Inv1D
Expand@Inv2D
Expand@Inv3D

Out[116]= 2 a2 - 2 c2 + 2 u2 + 2 c- c2 - u2 c+ c2 - u2

Out[117]= 2 a2 - 2 c + 2 c2 - 2 u2

Out[118]= 2 a2 - 2 c2 + 2 u2 - 2 c- c2 - u2 c+ c2 - u2

We now evaluate the invariants for C4, and see that all of them have factors c2 - u2 that will not
disappear, so none of the invariants give us an integer number. Also, all of the invariants give a
different expression which means that the resolvent is squarefree and we can therefore conclude
that Galois group is definitely not a subgroup of C4.

34
2 Galois group D4.nb

In[119]:= Invariant1 = x1 x2 ^ 2 + x2 x3 ^ 2 + x3 x4 ^ 2 + x4 x1 ^ 2;
Invariant2 = x2 x1 ^ 2 + x1 x3 ^ 2 + x3 x4 ^ 2 + x4 x2 ^ 2 ;
Invariant3 = x3 x2 ^ 2 + x2 x1 ^ 2 + x1 x4 ^ 2 + x4 x3 ^ 2 ;
Invariant4 = x4 x2 ^ 2 + x2 x3 ^ 2 + x3 x1 ^ 2 + x1 x4 ^ 2;
Invariant5 = x1 x3 ^ 2 + x3 x2 ^ 2 + x2 x4 ^ 2 + x4 x1 ^ 2;
Invariant6 = x1 x2 ^ 2 + x2 x4 ^ 2 + x4 x3 ^ 2 + x3 x1 ^ 2;
Expand@Invariant1D
Expand@Invariant2D
Expand@Invariant3D
Expand@Invariant4D
Expand@Invariant5D
Expand@Invariant6D

Out[125]= 4 a3 - 4 c2 + 4 a c2 + 4 u2 - 4 a u2 - 4 c2 c- c2 - u2 + 4 u2 c- c2 - u2 -

2 c2 - u2 c- c2 - u2 - 4 c2 c+ c2 - u2 + 4 u2 c+ c2 - u2 +

2 c2 - u2 c+ c2 - u2 -4a c- c2 - u2 c+ c2 - u2

Out[126]= 4 a3 - 4 c2 + 4 a c2 + 4 u2 - 4 a u2 - 4 c2 c- c2 - u2 + 4 u2 c- c2 - u2 -

2 c2 - u2 c- c2 - u2 + 4 c2 c+ c2 - u2 - 4 u2 c+ c2 - u2 -

2 c2 - u2 c+ c2 - u2 +4a c- c2 - u2 c+ c2 - u2

Out[127]= 4 a3 - 4 c2 + 4 a c2 + 4 u2 - 4 a u2 + 4 c2 c- c2 - u2 - 4 u2 c- c2 - u2 +

2 c2 - u2 c- c2 - u2 + 4 c2 c+ c2 - u2 - 4 u2 c+ c2 - u2 -

2 c2 - u2 c+ c2 - u2 -4a c- c2 - u2 c+ c2 - u2

Out[128]= 4 a3 + 4 a c - 4 c2 - 4 a c2 + 4 u2 + 4 a u2 + 8 c2 - u2 c- c2 - u2 c+ c2 - u2

Out[129]= 4 a3 + 4 a c - 4 c2 - 4 a c2 + 4 u2 + 4 a u2 - 8 c2 - u2 c- c2 - u2 c+ c2 - u2

Out[130]= 4 a3 - 4 c2 + 4 a c2 + 4 u2 - 4 a u2 + 4 c2 c- c2 - u2 - 4 u2 c- c2 - u2 +

2 c2 - u2 c- c2 - u2 - 4 c2 c+ c2 - u2 + 4 u2 c+ c2 - u2 +

2 c2 - u2 c+ c2 - u2 +4a c- c2 - u2 c+ c2 - u2

35
Printed by Wolfram Mathematica Student Edition
6.2 Appendix 2: Proof of Theorem 4.1 for C4
In[38]:= g@x_D := Hx - Ha + Sqrt@bD + Sqrt@c + Sqrt@bDDLL Hx - Ha + Sqrt@bD - Sqrt@c + Sqrt@bDDLL
Hx - Ha - Sqrt@bD + Sqrt@c - Sqrt@bDDLL Hx - Ha - Sqrt@bD - Sqrt@c - Sqrt@bDDLL;
b = c ^ 2  H1 + v ^ 2L;
Expand@g@xDD
c4 c2 4 a c2 2 a2 c2 2 c3
a4 - 2 a2 c + c2 + - + - - -
I1 + v2 M
Out[40]=
2 1 + v2 1 + v2 1 + v2 1 + v2

4 c2 x 4 a c2 x 2 c2 x2
4 a3 x + 4 a c x - + + 6 a2 x2 - 2 c x2 - - 4 a x3 + x4
1 + v2 1 + v2 1 + v2

In the cell above we defined the function g(x) and have put the restricion on b. We see that we get a
function with rational coefficients which we can, if needed, make integers. In the cell below we will
define the roots and the different invariants according to the left coset decomposition of S4 over C4.
We then see that all evaluations of the invariant give us different values, and that the third and the
fourth one give us integer values. (Note that

c2 c2 c2 c2 v2 cv
1+v2
c- 1+v2
c+ 1+v2
= 1+v2 1+v2
= 1+v 2

which is an integer. )
In[41]:= x1 = a + Sqrt@bD + Sqrt@c + Sqrt@bDD;
x2 = a + Sqrt@bD - Sqrt@c + Sqrt@bDD;
x3 = a - Sqrt@bD + Sqrt@c - Sqrt@bDD;
x4 = a - Sqrt@bD - Sqrt@c - Sqrt@bDD;

Invariant1 = x1 x2 ^ 2 + x2 x3 ^ 2 + x3 x4 ^ 2 + x4 x1 ^ 2;
Invariant2 = x2 x1 ^ 2 + x1 x3 ^ 2 + x3 x4 ^ 2 + x4 x2 ^ 2 ;
Invariant3 = x3 x2 ^ 2 + x2 x1 ^ 2 + x1 x4 ^ 2 + x4 x3 ^ 2 ;
Invariant4 = x4 x2 ^ 2 + x2 x3 ^ 2 + x3 x1 ^ 2 + x1 x4 ^ 2;
Invariant5 = x1 x3 ^ 2 + x3 x2 ^ 2 + x2 x4 ^ 2 + x4 x1 ^ 2;
Invariant6 = x1 x2 ^ 2 + x2 x4 ^ 2 + x4 x3 ^ 2 + x3 x1 ^ 2;
Expand@Invariant1D
Expand@Invariant2D
Expand@Invariant3D
Expand@Invariant4D
Expand@Invariant5D
Expand@Invariant6D

c2
4 c2 c-
1+v2
4 c2 4 a c2 c2 c2
Out[51]= 4 a3 - + -2 c- - +
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c+
1+v2
c2 c2 c2 c2
2 c+ - -4a c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

36
2 Galois group C4.nb

c2
4 c2 c-
1+v2
4 c2 4 a c2 c2 c2
Out[52]= 4 a3 - + -2 c- - -
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c+
1+v2
c2 c2 c2 c2
2 c+ + +4a c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c-
1+v2
4 c2 4a c2 c2 c2
Out[53]= 4 a3 - + +2 c- + -
1+ v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c+
1+v2
c2 c2 c2 c2
2 c+ + -4a c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

4 c2 4 a c2 c2 c2 c2
Out[54]= 4 a3 + 4 a c - - +8 c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

4 c2 4 a c2 c2 c2 c2
Out[55]= 4 a3 + 4 a c - - -8 c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c-
1+v2
4 c2 4a c2 c2 c2
Out[56]= 4 a3 - + +2 c- + +
1+ v2 1 + v2 1 + v2 1 + v2 1 + v2

c2
4 c2 c+
1+v2
c2 c2 c2 c2
2 c+ - +4a c- c+
1 + v2 1 + v2 1 + v2 1 + v2 1 + v2

We now evaluate the three different invariants for D4, where we know that the Galois group is only
a subgroup of D4 if all three give us a rational number. We see that the first and third one do not
give a rational number, and we can therefore conclude the Galois group is not a subgroup of D4.

37
Printed by Wolfram Mathematica Student Edition
 Galois group C4.nb 3

In[57]:= Inv1 = x1 x3 + x2 x4;

Inv2 = x1 x2 + x3 x4;
Inv3 = x1 x4 + x2 x3;
Expand@Inv1D
Expand@Inv2D
Expand@Inv3D

2 c2 c2 c2
Out[60]= 2 a2 - +2 c- c+
1 + v2 1 + v2 1 + v2

2 c2
Out[61]= 2 a2 - 2 c +
1 + v2

2 c2 c2 c2
Out[62]= 2 a2 - -2 c- c+
1 + v2 1 + v2 1 + v2

38
Printed by Wolfram Mathematica Student Edition
References
[1] Henri Cohen: A course in computational algebraic number theory.
Section 6.3: Computing Galois Groups.
Graduate texts in mathematics, Springer. Fourth printing, 2000.

[2] Prof. Frits Beukers: Rings and Galois Theory.

Department of Mathematics, Utrecht University, 2017.

[3] Bartel Leendert van der Waerden: Moderne Algebra.

1930, Springer.

[4] Prof. Hendrik Lenstra: The Chebotarev density theorem.

http : //websites.math.leidenuniv.nl/algebra/Lenstra − Chebotarev.pdf
Mathematical Institute, Leiden University. Used 18-10-2017.

[5] M. J. Greenberg: An Elementary Proof of the Kronecker-Weber Theorem

The American Mathematical Monthly
Vol. 81, No. 6 (Jun. - Jul., 1974), pp. 601-607
https : //www.jstor.org/stable/2319208?seq = 1#page scan tab contents

39