F-automorphisms of an extension
We continue to work as we did in the last section with fields of characteristic zero.
9.2.1 Definition. Let K an extension of a field F. A monomorphism of K in ℂ which fixes the elements of F is said F-monomorphism of K over F. □
Example 9.2.2. Let F = ℚ, K = ℚ(α), where α is the third real root of x3 − 3, we have the following F-monomorphism φ from K to ℂ which sends α to ωα (with ω primitive third root of unity) and every a ∈ F to itself.
We indicate with I(K,F) the set of all F-monomorphisms. ■
9.2.3 Proposition. If F = ℚ, every monomorphism of K is a ℚ-monomorphism.
Proof. Let φ a monomorphism of K. Then it results
φ(0), φ(1) = 1,
φ(n) = φ(1 + 1 + ... + 1) = φ(1) + φ(1) + ... + φ(1) = 1 + 1 + ... + 1 = n
φ(m/n) = φ(m)φ(n)−1 = mn−1 = m/n. □
Let K a finite extension of F. We shall find all the F-monomorphism of K. Since K is a finite extension we have K = F(α) for some α ∈ K (see Proposition 8.8.2). If [K:F] = n, every element k ∈ K can be written in the form
k = c0 + c1α + c2α2 + ... + cn − 1αn −1, ci ∈ F
Let φ a F-monomorphism of K. Then
φ(k) = φ(c0) + φ(c1α) + φ(c2α2) + ... + φ(cn − 1αn −1), = c0 + c1 φ(α) + c2φ(α2) + ... + cn − 1φ(αn −1)
where φ is completely determined once we know φ(α). Further φ(α) is conjugated to α over F (see definition 8.7.1. Indeed we have the following proposition
9.2.4 Proposition. Let K = F(α) a finite extension of F and let φ a F-monomorphism of K. Then φ(α) is a conjugate of α over F, that is φ(α) and α have the same minimal polynomial over F.
Proof. Let p(x) = a0 + a1 x + a2 x2 + ... + anxn, (ai ∈ F) the minimal polynomial of α. We shall prove that φ(α) is also a root of p(x). We have the following equalities:
p(φ(α)) = a0 + a1 φ(α) + ... + anφ(α)n
= φ(a0) + φ(a1) φ(α) + ... + φ(an)φ(αn)
= φ(a0 + a1α + anαn) = φ(0) = 0. □
Thus the possible transformed element of a root α of an irreducible polynomial with coefficients in F can be found among the roots of the polynomial itself. For example consider the extension K = ℚ(√3) of ℚ. If f(x) = x2 − 3, then the transformed √3 through of K can only be either √3 or −√3. As
φ(x2 − 3) = φ(0)
φ(x2) − φ(3) = 0
φ(x)φ(x) = 3
φ(x) = ±3
As a consequence of the previous results we have the following theorem.
9.2.5 Theorem. Let K a finite extension of F, with degree n. Then there exist at most n F-monomorphism of K. In other words:
|I(K,F)| ≤ [K:F] □
Before proving that there exist exactly n F-monomorphism, we introduce the following lemma.
9.2.6 Lemma. Let F a field and let F(α) an extension of degree n. Let θ a monomorphism from F to ℂ, and let p(x) = a0 + a1x + a2x2 + ... + anxn (ai ∈ F) the minimal polynomial of α. Let pθ the polynomial with coefficients in ℂ, pθ = θ(a0) + θ(a1)x + ... + θ(an)xn with β a root of pθ, there exists a monomorphism, extension of θ to F(α) which sends α to β.
Proof. The proof of this lemma is analogous to that of theorem Theorem 8.2.13, but we repeat it for sake of completeness. Every element s ∈ F(α) can be written univocally as
s = a0 + a1α + a2α2 + ... + an −1αn − 1, ai ∈ F
We define φ: F(α) ⟶ ℂ as
φ(s) := θ(a0) + θ(a1) β + ... + an −1βn − 1
It is obvious that φ is an extension of θ which sends α to β, preserving addition. We show that it preserves the product as well. Let t = b0 + b1 α + ...+ bnαn−1 another element of F(α). We consider the following polynomial associate respectively to s and t:
f(x) := a0 + a1x + ... + an−1xn−1, g(x) := b0 + b1x + ... + bn−1xn−1
Dividing f ⋅ g by p
f(x) g(x) = p(x) ⋅ q(x) + r(x) r(x) = r0 + r1x + ... + rn−1xn−1
the value in α of this relation, knowing that p(α) = 0, is
st = f(α) g(α) = r(α)
For every polynomial v(x) = ∑mi=0 cixi, we indicate with vθ(x) the polynomial ∑mi=0 θ(ci)xi, we have the following chains of equalities
φ(st) = ρθ(β) qθ(β) + rθ(β)
=(p ⋅ q + r)θ (β) = (f ⋅ g)θ(β) = fθ(β) ⋅ gθ(β) = φ(s) ⋅ φ(t).
Thus φ preserves the product as well: it is a homomorphism. Further, since φ(1) = 1, φ is not the null homomorphism, hence being a non-null homomorphism between fields, it is a monomorphism. □
We are now able to prove that there are exactly n F-monomorphisms.
9.2.7 Theorem. Let K an extension of F, of degree n, and let α ∈ K such that K = F(α). Let α1 = α, α2, ..., αn the conjugates elements of α over F. Then
α, α2, ..., αn are all distinct;
for every j = 1, ..,n there exits a unique F-monomorphism φj of K in ℂ such that φj(α) = αj;
φj of the previous points cover all the possible F-monomorphisms from K to ℂ.
Proof. a) α1, α2, ..., αn are distinct since zeros of an irreducible polynomial with characteristic zero. (see proposition 8.3.6).
b) If in lemma 9.2.5 we consider as θ the mapping that fixes all the elements of F, then pθ = p, the lemma says there exists a F-monomorphism φj which sends α to αj given by
φj(a0 + a1α + ... + an − 1 αn − 1) = a0 + a1αj + ... + an − 1αjn−1.
c) We found n distinct F-monomorphisms because the αi are all distinct and thus there cannot be any other, owing to theorem 9.2.5. □
Among the F-monomorphism, there exists a class of particular importance.
9.2.8 Definition. Let K an extension of a field F. A F-automorphism σ of K is a F-monomorphism of K over K itself. We indicate with G(K,F) the set of all F-automorphism of K. □
Not every F-monomorphism is a F-automorphism, e.g. in Example 9.2.2 the image through φ of α is not contained in ℚ(α). (α is real, while φ(α) = ωα is complex and not real).
The important fact regarding the F-automorphisms is that with respect to the product of functions the set G(K,F) is a group.
9.2.9 Proposition. The set G(K,F) of all F-automorphisms of an extension K of F is a subgroup of the group of all automorphisms of K.
Proof. By definition of F-automorphism we have
G(K,F) := {σ: K ⟶ K | σ(a) = a ∀a ∈ F} (9.2.1)
Let σ, τ ∈ G(K,F). Then their product σ ∘ τ is an automorphism which fixes F, because
(σ ∘ τ)(a) = σ(τ(a)) = σ(a) = a ∀a ∈ F
And σ−1 is an automorphism and
a = σ−1(σ(a)) = σ−1(a)
Hence G(K,F) is a subgroup of the group of all automorphisms of K. □
9.2.10 Definition. The group G(K,F) of all utomorphisms of K which fixes the elements of F is known as Galois group of the extension K over F. □
9.2.11 Theorem. Let K a finite extension of a field F and let n its degree. Then
|G(K,F)| ≤ |K:F|
Proof. It is an immediate consequence of theorem 9.2.5, observing that G(K,F) ⊆ I(K,F). □
9.2.12 Definition. Let K a field and G a group of automorphisms of K. We define
KG := {k ∈ K | σ(k) = k ∀σ ∈ G}
which is known as field fixed by G. □
The following proposition show that KG is indeed a field.
9.2.13 Proposition. Let K a field and G a group of automorphisms of K. The set KG = {k ∈ K | σ(k) = k ∀σ ∈ G} is a subfield of K.
Proof. Let k1,k2 ∈ KG. Then for every σ ∈ G.
σ(k1 ± k2) = σ(k1)±σ(k2) = k1 ± k2 ⇒ k1 ± k2 ∈ KG
σ(k1 k2−1) = σ(k1)σ(k2)−1 = k1 k2−1 ⇒ k1 k2−1 ∈ KG □
Let now K an extension of a field F. If G(K,F) is the Galois groups of the extension K over F, it results as immediate consequence of the definition of G(K,F), that
KG(F,K) ⊇ F
9.2.14 Example. Let K = ℂ, field of the complex numbers and let F = ℝ.
If σ ∈ G(ℂ, ℝ), since ℂ = ℝ(i). A field automorphism, is such that σ(x+y) = σ(x) + σ(y), and σ(xy) = σ(x) σ(y), so
σ(a + bi) = a + b σ a, b ∈ ℝ
σ(−1) = σ(i2) = σ(i)σ(i). Since σ(−1) = −1, the square of σ(i), must be −1, that is σ(i) must be a root of x2 + 1, i.e. ±i. Now that we know that σ(i) = ± i, it's clear that can have only two possible mappings
σ1 : ℂ ⟶ ℂ
a + ib ⟶ a + ib
and
σ2 : ℂ ⟶ ℂ
a + ib ⟶ a − ib
are respectively the identity automorphism and the automorpism which maps an element to its complex conjugate. Thus
G(ℂ, ℝ) = {σ1, σ2} ≃ ℤ2
We evaluate now the field fixed by G(ℂ, ℝ), that is
ℂG(ℂ, ℝ) = {k ∈ ℂ | σ(k) = k ∀ σ ∈ G(ℂ, ℝ)}
Since that the identity fixes all the elements in ℂ, it suffices to consider the element k ∈ ℂ that are fixed by σ2. Thus
ℂG(ℂ, ℝ) = {k ∈ ℂ | σ2(k) = k} = {a + ib | σ2(a + ib} = {a + ib | a − ib = a + ib} = ℝ
In this case, KG(F,K), which as we know must contain F, is equal to F. □
9.2.15 Example. Let K = ℚ(∛5) and F = ℚ. By Eisenstein and Gauss x3 − 5 is irreducible in ℚ[x], so [ℚ(∛5): ℚ] = 3 and a basis is {1, ∛5, ∛52}. So we can express K as
ℚ(∛5) = {a0 + a1 ∛5 + a2 ∛52 | a0, a1, a2 ∈ ℚ}
The action of σ on a element of K is such that
σ(a0 + a1 ∛5 + a2 ∛52) = σ(a0) + σ(a1 ∛5) + σ(a2 ∛52) = a0 + a1 σ(∛5) + a2 σ(∛52)
So to find all possible automorphisms of ℚ(∛5) (which fix ℚ), by proposition 9.2.4, we only need to find the images σ(∛5). Since the minimal polynomial of α over ℚ is x3 − 5, σ(α) must be another root of the same polynomial. Apart from α the other two roots are complex, hence not in ℚ(∛5) ⊂ ℝ. It follows that the only possibility is σ(α) = α, that is the unique automoprhism of G(ℚ(∛5), ℚ) is the identity automorphism.
G(ℚ(∛5), ℚ) = {id}
It is obvious that the field fixed by G(ℚ(∛5), ℚ) is all G(ℚ(∛5)), hence in this case
KG(F,K) = ℚ(∛5)G(ℚ(∛5), ℚ) = ℚ(∛5) ⊇ ℚ
Notice that in this case, respect to the previous example we have
1 = |G(ℚ(∛5), ℚ)| < |ℚ(∛5) : ℚ| = 3. □
9.2.16 Example. Let K = ℚ(∛5), ω), where ω is a primitive third root of unity. Let's calculate ℚ(∛5), ω) as vector space over ℚ. We start by ℚ(∛5) (already calculated in the previous exercise) and then adjoin ω. ω is root of the irreducible polynomial x2 + x + 1 over ℚ(∛5) (as over ℚ, see exercise 7.d), so
[ℚ(∛5,ω) : ℚ] = [ℚ(∛5,ω) : ℚ(∛5)] = [ℚ(∛5) : ℚ] = 2 ⋅ 3 = 6.
Since a basis of ℚ(∛5,ω) over ℚ(∛5) is given by {1, ω} and a basis of ℚ(∛5) over ℚ is {1, ∛5, ∛52} we have that a basis of ℚ(∛5,ω) over ℚ is given by
1, ∛5, ∛52, ω, ∛52ω, ∛52ω
The roots of x3 − 5, are ∛5, ∛5ω, ∛5ω2.
ω is the root of x3 − 1 which is reducible as (x − 1) (x2 + x + 1), only the last is irreducible over ℚ with roots −1/2 ± √3i/2. There are in total six automoprhisms
σ1 = id | ||
σ2: | σ2(ω) = ω2 | σ2(∛5) = ∛5 |
σ3: | σ2(ω) = ω | σ2(∛5) = ω ∛5 |
σ4 = σ32: | σ4(ω) = ω | σ4(∛5) = ω2 ∛5 |
σ5 = σ2 ∘ σ3: | σ5(ω) = ω2 | σ5(∛5) = ω2 ∛5 |
σ6 = σ2 ∘ σ32: | σ(ω) = ω2 | σ6(∛5) = ω ∛5 |
Notice that in this case
6 = |G(ℚ(∛5,ω), ℚ| = |(ℚ(∛5,ω)| = 6
it results G(K, ℚ) ≃ S3. ■
9.2.17 Example. Let ζ a primitive nth root of unity. We calculate ℚ(ζ). We know ζ satisfies the nth cyclotomic polynomial Φn(x) which is irreducible over ℚ, thus the minimal polynomial of ζ over ℚ. It follows that
[ℚ(ζ) : ℚ] = φ(n)
Every element of G(ℚ(ζ), ℚ), maps ζ to another root of Φn(x), i.e. an element of the form ζr with 1 ≤ r ≤ n and (n,r) = 1. Conversely if r is an integer such that 1 ≤ r ≤ n and (n,r) = 1, the mapping σr which maps ζ to ζr is an automorphism of ℚ(ζ). This fact suggests the possibility to define the following mapping between the group U(ℤn) of the invertible elements of ℤn and G(ℚ(ζ), ℚ):
Ψ : (U(ℤn),⋅) ⟶ (G(ℚ(ζ), ℚ), ∘)
k̄ ⟶ σk
This is a bijective mapping. We prove it is a group homomorphism. If ī, j̄ ∈ U(ℤn), since we are dealing with a multiplicative group we have that ij ≡ k (mod n). G(ℚ(ζ), ℚ) is a group under composition of function so we show that σi σj = σk, indeed:
σiσi(ζ) = σi(ζj) = ζij = ζk = σk(ζ)
We have to confirm that ζij = ζk. Now since ij ≡ k (mod n) means that there exists m such that ij = k + mn, therefore,
ζij = ζk + mn = ζk ζmn = ζk (ζn)m = ζk 1m = ζk
which proves that
G(ℚ(ζ), ℚ) ≃ U(ℤn) □
In the examples above, it results
|G(K,F)| ≤ [K:F] and KG(K,F) ⊇ F
In some cases the equality sign holds and as we shall prove soon it happens when K is a normal extension.
Through definition 9.2.10 it is possible to associate to every extension K of a field F a group, its Galois group G(K,F). We shall now establish a relatin between the set of intermediate fields of an extension K of F and the set of subgroups of G(K,F).
Let K an extension of F and let G(K,F) its Galois group. For every intermediate field, T of K containing F we define
G(K,T) := {σ ∈ G(K,F) | σ(t) = t ∀t ∈ T} (9.2.2)
For every subgroup H of G(K,F) we define
KH := {k ∈ K | σ(k) = k ∀σ ∈ H}
The following properties hold true.
9.2.18 Proposition. Let K an extension of F and let G(K,F) its Galois group. For every subgroup H of G(K,F) and every intermediate subfield T such that F ⊆ T ⊆ K, we have
G(K,K) = id;
if T1, T2 are fields such that F ⊆ T1 ⊆ T2 ⊆ K, then
G(K,T) is a subgroup of G(K,F);
Kid = K;
If H1, H2 are subgroups of G(K,F) such that H1 ⊆ H2 ⊆ G(K,F), then
KH2 ⊆ KH1;
KH is an intermediate field: F ⊆ KH ⊆ K;
H ⊆ G(K,KH);
T ⊆ KG(K,T).
G(K,T2) ⊆ G(K,T1)
Proof.
G(K,K) = {σ ∈ G(K,F) | σ(k) = k ∀k ∈ K}, and this is exactly what σ = id means.
Supose that T1 ⊆ T2 and let σ ∈ G(K,T2). Then σ fixes every element of T2 and so certainly fixes every element of T1. Hence σ ∈ G(K,T1)
G(K,T) is made by the automorphisms that fix the elements of T, which contains F, but all the automorphism that fix T must also fix F.
Kid = {k ∈ K | σ(k) = k ∀σ ∈ {id}} = K.
The elements k ∈ K that are fixed by the automorpshims of H2, must be fixed as well by the automorphisms of H1, because H1 is a subgroup of H2. Thus KH2 ⊆ KH1
By definition of KH is clear that KH ⊆ K. The other inclusion F ⊆ KH derives from the fact that H is a subgroup of G(K,F) which by definition the group of automorphisms that fix the elements of F.
H is a subgroup of G(K,F)
G(K,KH) is made of the automorphisms that fix all the elements in KH, and we proved in f) that it contains F. G(K,KH) must contain all the automorpshisms of H, because this latter is a subgroup of G(K,F) which fixes only the element of F.
The automorphisms in G(K,T) fix all the elements of T, and KG(K,T) contains all the elements of K which are fixed by the elements of G(K,T), thus it contains T. □
9.2.19 Proposition. In the field ℝ of all real numbers, the identity mapping of ℝ is the only automorphism.
Before proving it we introduce two lemmas.
9.2.20 Lemma. Let ϕ: ℝ → ℝ be an automorphism of the field of real numbers ℝ, then for any positive real number x , we have ϕ(x) > 0
Proof. ϕ(x2) = ϕ(x) ϕ(x) ≥ 0. Note that since ϕ(0) = 0 and ϕ is bijective, ϕ(x) ≠ 0 for any x ≠ 0. Thus, it follows that ϕ(x) > 0 for each positive real number x. □
9.2.21 Lemma. For any x,y ∈ ℝ such that x > y, we have ϕ(x) > ϕ(y). Thus, in any automorphism of the field of real numbers the order of the elements is preserved.
Proof. Since x > y, we have x−y > 0 and it follows from the previous lemma that
0 < ϕ(x−y) = ϕ(x) − ϕ(y).
Hence, ϕ(x) > ϕ(y). □
Proof of 9.2.19. Let σ an automorphsim of ℝ, and let a,b ∈ ℝ such that a < b. Then by the previous lemma σ(a) < σ(b). Suppose that there exists a ∈ ℝ such that σ(a) > a. Let q a rational number, such that σ(a) > q > a, we have σ(q) = q. This leads to a contradiction since q > a but σ(a) > σ(q) = q. □
Exercise 9.2.22. Determine the Galois group G(ℚ(∛2), ℚ) and its fixed field.
All the images of ∛2 are to be searched among the roots of x3 − 2. Since ℚ(∛2) has only one root, the only automorphism is the identity, G(ℚ(∛2), ℚ) = {id}, whereas I(ℚ(∛2), ℚ) has three elements. It is obvious that its fixed field is the entire ℚ(∛2). ■