Notes on Galois Theory

Tianyu Tao
April 23, 2015

1 Facts from field extensions

References are Dummit Foote and Basic Algebra.
Algebraic Extension and transcendental extension:
notations and definitions.

First a couple of

We write E/F for E a field containing F and call it extension of F .

Let S be an subset of E, there two different things
F (S), which is the subfield generated by F and S, which is of course
contained in E;
F [S], which is the subring generated by F and S, also is contained in
E.
a basic fact is F (S)(T ) = F (S T ), when S = {u1 , . . . , un }, we denote F (S)
by F (u1 , . . . , un ). And in general F [u] F (u).
To investigate the structure of F (u) for arbitrary u E/F , we recall the mapping
property of the polynomial ring F [x]: for any commutative ring S (here we take
S = F [u]) which contains u, and is a homomorphism from F to S, there is an
unique extension of from F [x] to S which sends x to u:
: F [x] F [u]
and is identity on F , sends x to u, let K be the kernel of the map, then we have:
F [u] ' F [x]/K
there are two cases: either
1

Notes on Galois theory

Tianyu Tao

K = (f (x)) for some nonzero polynomial f (x) F [x], since E/F is a filed,
F [u] E/F is a domain, so f (x) has to be prime, but F [x] is a PID because
F is a field, so (f (x)) is maximal and F [x]/(f (x)) is a field, then we are
forced to have
F [u] ' F (u)
Here u is called Algebraic over F , and f (x) is called (if monic), the minimum
polynomial of u, and F (u) a simple extension of F , and the dimension of
F (u), considered as a vector space over F , denoted by [F (u) : F ], is finite (so
we call the extension algebraic), because f (u) = 0, the dimension is precisely
the degree of f .
K = 0, then F [x] ' F [u], and this isomorphism can be extended to an
isomorphism on the field of fractions F (x) of F [x], which are the rational
functions with one indeterminant and coefficient in F to F (u), so
F (x) ' F (u)
this says an element in F (u) is just obtained by plug in x for u in the
rational functions from F (x).
Here u is called transcendental over F , and the degree [F (u) : F ] = , so u
is algebraic over F iff [F (u) : F ] < .
The following is very important, although trivial to see.
Theorem 1.1 Let K/E be an extension of E, E/F itself an extension of F ,
then we can view K as an extension of F , K/F , and the degree has the following
relation: [K : E][E : F ] = [K : F ].
Proof. Omitted.
Splitting field: We want to study solvablity of polynomials by radicals in F [x]
and the extension (E/F )[x], we form the following definition:
Definition 1.2 Let f F [x] be a monic polynomial, an extension E/F is a
splitting field of F if f (x) splits, i.e.
n
Y
f (x) =
(x ri )
i=1

in E[x], and E is just generated by F and the roots:

E = F (r1 , . . . , rn ).
2

Notes on Galois theory

Tianyu Tao

Theorem 1.3 For each monic f (x) F [x], a splitting field for it exists and is
unique up to isomorphism, and the number of isomorphism is bounded by [E : F ],
equality holds when f has distinct roots in the target extension.
sketch. The existence part is typically done by induction on the degree of f , it is
easy to form an extension of f containing one root: if fi (x) is an irreducible factor
of f , let K = F [x]/fi (x), then the image of x in the quotient is a root for f (x) in
the extension K[x].
The uniqueness part goes like this: one first have a field isomorphism
: F 0 F
then one has the splitting field E/F of f (x) F [x] and 0 E/0 F of 0 f (x) F 0 [x] (0 f
has coefficient which are the image of f s), and we want to extend on E, this is
also done by induction: one first try to extend on F (r) for r E algebraic with
some minimum polynomial g F [x], one can do this precisely when 0 g 0 F [x] has
a root, then one goes on from there.
Multiple roots, derivative, separable polynomials and perfect fields
Since the number of distinct roots determines the different types of automorphism
between splitting fields of monic f (x) F [x], we are also aware of the possibility
of f to have repeated roots, hence we want to have a method to tell exactly when
this happens:
First, a reduction: note there is no loss of generality to assume f (x) is a product
of distinct irreducible factors, because if two polynomials p(x), q(x) are relatively
prime (being different irreducible(=prime in PID)), then there is a(x), b(x) such
that
a(x)p(x) + b(x)q(x) = 1,
so that no roots of p shall be that of qs, and vice versa.
The exact object which we will be employing to decide the multiplicity of
the roots is the derivative, which is just the usual derivative from Calculus for
polynomials, but this is an algebraic object, without the notion of any topology,
limits!
And we have
Theorem 1.4 Let f (x) F [x] be monic, then f has no repeated roots in any
splitting field E/F if and only if f and f 0 are relatively prime:
(f, f 0 ) = 1.
Proof. Omitted, this is product rule working in the background.
3

Notes on Galois theory

Tianyu Tao

Then we make the following definition:

f (x) F [x] is separable if all its irreducible factors have distinct roots in
some extension of F .
A field F is called perfect if all f (x) F [x] is separable.
These and above theorem allows us to characterise: Let f (x) F [x] be irreducible, then if f (x) has repeated roots, by the above criterion, (f 0 , f ) 6= 1, but
this means f divides f 0 which is impossible unless f 0 = 0. So there are different
cases:
If the characteristic of F is 0, then f 0 = 0 means f is a constant, i.e. f (x)
F , then f cannot be irreducible (contradict definition), hence in a filed F of
zero characteristic, f 0 6= 0 for irreducible f (x) F [x], and more importantly
every irreducible f (x) are separable. Thus such a field is perfect.
P
If the characteristic of F is p, then if f (x) = ni=0 ai xi , then f 0 (x) = 0 if and
only if ai = 0 for each i not divisible by p, so
f (x) = b0 + b1 xp + + bm xmp = g1 (xp )
now g1 (x) is clearly irreducible, since any factorization of it leads to a factorization of f , but it may or may not be separable, if not, repeating the above
procedure we arrive at g2 (x) such that g1 (x) = g2 (xp ), since the degree is
finite one must stop at some gk (x) such that
k

f (x) = gk (xp )
this gk must be separable by maximality of its power. and in this case there
might be irreducible polynomials f (x) which are not separable, i.e. has
repeated roots in extensions of F .
In a field of characteristic p, the Frobenius endormorphism a 7 ap is a
monomorphism on F and its image F p .
The following lemma is important:
Lemma Let F be a field with characteristic p and a F , then either xp a
is irreducible or it is a pth power in F [x].
Proof. Factor xp a as g(x)h(x) one one hand, on the other hand find a pth
root of a in some extension of F and thus xp a = (x b)p , assume xp a
is not irreducible will force b F .

Notes on Galois theory

Tianyu Tao

Then the following theorem is immediate:

Theorem A field F with character p 6= 0 is perfect iff F = F p .
Proof. Suppose F 6= F p , then pick a F F p , then xp a is not a pth power
in F [x], hence irreducible in F [x], but (xp a)0 = pxp1 = 0. so (f, f 0 ) 6= 1
and f has repeated roots, F is not perfect.
Conversely, if f (x) F [x] is irreducible and not separable, then (f, f 0 ) 6= 1,
k
write f (x) = gk (xp ), and some coefficient of gk , say a F is not a pth power
for otherwise f will be a pth power in F [x], which is mutually exclusive to
irreducible, hence F 6= F p .
This implies every finite field is perfect, since the Frobineus endormorphism
is injective, and hence surjective by finiteness, forces F p = F .

2 Galois Theory
Basic notions in Galois theory The Fundamental theorem of Galois theory
gives a one-to-one correspondence between the set of subsets of E/F for E the
splitting field of a separable polynomial f (x) F [x], with the set of subgroups of
the group of automorphism of E/F .
Given F field, consider a extension E/F , let G denote the set of automorphisms of E/F , these are automorphism of E which fixes elements of F , then
we denote G by
Gal E/F
and we call it the Galois group of E over F .
Given any field E, let G be a subgroup of the automorphisms of E, we denote
Inv G = {a E | (a) = a, G}
which is a subfield of E, whose elements are fixed by all automorphism in G.
(also called the fixed field of G in E).
Then there are two maps
G Inv G
F Gal E/F

Notes on Galois theory

Tianyu Tao

that went between the set of subgroups of automorphisms of E and set of

subfields of E. It is easy to see that these map are inclusion reversing:
G1 G2 = Inv G1 Inv G2
F1 F2 = Gal E/F1 Gal E/F2
Also, these are inverse operations of each other, at least in one direction:
Inv (Gal E/F) F
Gal (E/Inv G) F
Use our previous result on splitting fields, we have the following:
Theorem 2.1 Let f (x) F [x] and E/F its splitting field, then
|Gal E/F | [E : F ]
and equality holdsQprecisely when f (x)s has distinct rootsQin E. Since E is the
splitting field for i pi (x)li iff it is the splitting field for i pi (x), we can always
take f to be a product of distinct irreducibles, then if f is separable, we have
|Gal E/F | = [E : F ]
Remark: Dummint and Foote choose to use Aut E/F here in general and
reserve Gal E/F to the case when it is equal to [E : F ], why? I dont know, life is
hard, confusing, always full of mysteries... fuck
The next result concerns about the fixed field of a subgroup of automorphisms
of G, first we need some tricky idea:
Definition A character of G (group) with respect to F (field) is a homomorphism between G and the invertible elements of F :
: G F
so (g1 g2 ) = (g1 )(g2 ) for all g1(2) G and (g) 6= 0 on G.
Then, due to Artin, we have his
Theorem 2.2 (Independence of characters) Distinct characters i (i = 1, 2, . . . , n)
of G with respect to F are linearly independent over F : if a1 , . . . , an are elements
from F such that
a1 1 + an n = 0
then a1 = = an = 0 in F , the 0 is of course the 0 homomorphism.
6

Notes on Galois theory

Tianyu Tao

sketch. Assume a minimal counter example, multiply and subtract will get a
smaller counter example.
and further
Theorem 2.3 Let G = {1 , . . . , n } be a finite subgroup of the group of automorphisms of a field E, then
[E : Inv G] = |G| = n
Proof. see Andersons notes.
Now even more fucking definitions:
An algebraic extension E/F is that any u E is algebraic over F .
A separable extension E/F is that any u E has separable minimum polynomial. In particular, a perfect field E is separable, if its algebraic.
A normal extension E/F is that any u E has its minimum polynomial
factors completely into linear factors in E; i.e. every irreducible polynomial
in F [x] which has a root in E factors completely in E[x].
A Galois extension (really?) E/F is normal + separable. In this case we
have that every irreducible polynomial of F [x] which has a root in E is a
product of distinct linear factors in E[x].
Now we have our major results of the section:
Theorem 2.4 Consider an extension E/F , then TFAE
(1) E is a splitting field over F of a separable polynomial f (x);
(2) F = Inv G for some finite group of automorphisms of E;
(3) E is finite dimensional Galois over F .
And we have: if E and F are as in (1) G = Gal E/F , then F = Inv G and if G
and F are as in (2), then G = Gal E/F .
Proof. (2) = (3) Let F be as in (2), by assumption, G is a subgroup of Gal E/F ,
so
|G| |Gal E/F |.
Let F1 = Inv (Gal(E/F )), invoke theorem 2.3, we know that |G| = [E : F ]
and [E : F1 ] = |Inv (Gal(E/F ))|, but
[E : F ] = [E : F1 ][F1 : F ]
7

Notes on Galois theory

Tianyu Tao

so [E : F ] [E : F ]/[F1 : F ], we must have |G| = |Gal E/F |, so G =

Gal E/F .
To see E is Galois, let f (x) F [x] be irreducible which has a root r in
E. Enumerate G = {1 = id, 2 , . . . , n }, put ri = i r, in particular ri are
distinct since G is the group of automorphism on E/F . Since f (r) = 0 in
E, then
0 = i f (r) = f (i r) = f (ri )
precisely because i fix F , thus
f (x) =

Y
(x ri )

in E[x].
(3) = (1) Already [E : F ] is finite, so let E = F (r1 , . . . , rn ) with ri algebraic
over F . Let fi (x) be the minimum polynomial of ri over F , since fi (ri ) = 0
so each fi factors completely into distinct linear factors in E[x], then E is
the splitting field of the product of fi .
(1) = (2) Put G = Gal E/F and F 0 = Inv G, use Theorem 2.1, we have
|G| = [E : F ],
while F 0 is a subset of E containing F , E is still the splitting field for f (x)
over F 0 , again we have
|G| = [E : F 0 ]
and
[E : F ] = [E : F 0 ][F 0 : F ]
so [F 0 : F ] = 1 and F 0 = F .
Now finally the fundamental correspondence:
Theorem 2.5 Let E/F be an extension satisfy one of the condition in previous
theorem, let G = Gal E/F , let = {H}, set of subgroups of G, let = {K}, set
of subfields of E containing F . Then the map
H Inv H, H
K Gal E/K, K
are inverses of each other, hence are bijections between and . And we have
(a). H1 H2 iff Inv H1 Inv H2 ;
8

Notes on Galois theory

Tianyu Tao

(b). |H| = [E : Inv H], [G : H] = [Inv H : F ];

(c). H is normal in G iff Inv H is a Galois extension over F , and we have
Gal ((Inv H)/F ) ' G/H
Proof. (a). We first show the map
H Inv H = K Gal E/(K)
is identity on .
Let H be a subgroup of G, we have F = Inv G, and F Inv H = K, so
F K E, and
Gal E/K = H
by theorem 2.4. :)
On the other hand, we need to show the map
K Gal E/K = H Inv H
is identity on . Let K be a subfield of E/F , put H = Gal E/K, then
H G = Gal E/F
Also, E is the splitting field over K for some separable polynomial, since any
f (x) F [x] is in K[x]. Hence
Inv H = Inv (Gal E/K) = K
showing the desired thing.
Next, we consider subgroups H1 H2 of G, clearly
Inv H1 Inv H2 ;
On the other hand, if Inv H1 = K1 Inv H2 = K2 holds, we have
H1 = Gal E/K1 Gal E/K2 = H2
by what we have just proved.
(b). By theorem 2.1 and that Gal (E/Inv H) = H we have
|H| = |Gal (E/Inv H)| = [E : Inv H];
9

Notes on Galois theory

Tianyu Tao

on the other hand, we have

|G| = |Gal E/F | = [E : F ] = [E : Inv H][Inv H : F ],
and by Lagrange theorem:
|G| = |H|[G : H] = [E : Inv H][G : H]
(Yes, the notation is terrible, we have to distinguish between group index
and degree of extension) so [G : H] = [Inv H : F ].
(c). Let H and K = Inv H, then we have the relation
Inv H 1 = (K) = (Inv H)
or
Gal E/(K) = (Gal (E/K)) 1
since (k) = k iff ( 1 )((k)) = (k). Thus H is a normal subgroup of G
iff (K) = K for each G, recall the definition...
So suppose (K) = K for all K, every G then has a restriction on K
which is an automorphism of K/F . We have the restriction homomorphism
: ,
between G = Gal E/F and Gal K/F . Let G denote the image of G under
this map, then
Inv G = F,
so that
G = Gal K/F.
Put L = ker , then
L = { G | |K = 1K },
so L = H = Gal E/K. Hence by first isomorpism theorem:
Gal K/F = G ' G/H,
and since F = Inv G, we have K is Galois over F by previous theorem.

Conversely, suppose K is Galois over F , let a K and f (x) F [x] be its

minimum polynomial, it factors completely in K[x] as linear factors:
f (x) =

m
Y
(x ai ),
i=1

10

Notes on Galois theory

Tianyu Tao

where a1 = a. If G, then f ((a)) = 0, so (a) = ai for some i, and

(a) K. Since a is arbitrary, we have (K) K. But then
H 1 = Gal (E/Inv H 1 ) = Gal (E/(K)) Gal E/K = H
by the correspondence, so H is normal.

3 Solvable groups, derived series

Some results from group theory, we say
1 = Gs+1 / Gs / GS / / G1 = G
is a normal series for G, as indicated, Gi is a normal subgroup of Gi1 , but not a
normal subgroup in higher chains, necessarily.
By normality, we can factor, so there is the associated sequence of factors.
Gs ' Gs /Gs+1 , , G1 /G2 .
We say G is solvable if Gi /Gi+1 is abelian in the above normal sequence for
each i. Some examples:
Theorem 3.1 A pgroup (group of order pn ) P is solvable.
Proof. Induction on n, by the class equation, P has non-trivial center, mod it out,
you get a smaller pgroup, so...
Let us introduce the important notation commutator of two elements g, h G,
we which we write as
g 1 h1 gh := [g, h]
(some other sources write [g, h] = ghg 1 h1 , anyway fuck them).
Some properties of commutators:
[g, h]1 = [h, g]
gh = hg[g, h]
The derived subgroup G0 , also written as [G, G], is the group generated by all
commutators, i.e. [G, G] = h{[g, h], g, h G}i
Note a general elements in G0 could be written as (g1 , h1 )(gk , hk ) for gi , hi G.
Theorem 3.2 G0 is a normal subgroup of G.
11

Notes on Galois theory

Tianyu Tao

Proof. If is a homomorphism from G to another group, then we have

([g, h]) = [(g), (h)] = (G0 ) ((G))0 ,
if happens to be surjective, the we have equality above. Let K / G, any inner
automorphim (act by conjugation) Ia := x 7 (axa1 ) where a G arbitrary
induces an endomorphism of K, so Ia (K 0 ) (Ia (K))0 . So K / G = K 0 / G, put
K = G we have G0 / G.
We may now form a chain G00 = (G0 )0 , . . ., so we have a series:
00

/ G / G0 / G
this has the nicer property than just a normal sequence in that G(k) / G for all
k by the proof in above theorem.
More properties of G0 :
Theorem 3.3 G/G0 is abelian and if G/K is abelian then G0 K.
Proof. Let (g), (h) G/G0 , then (g)(h)(g)1 (h)1 = (ghg 1 h1 ) = ([g, h]) = (1),
so (g)(h) = (h)(g). If g, h G and K / G, then [gK, hK] = [g, h]K as sets, so
(gK, hK) = K = (1) iff [g, h] K, thus G/K is abelian iff G0 K.
Theorem 3.4 G is solvable iff G(k) = 1 for some k.
Proof. If G(k) = 1 for some k, we have the derived series
1 = G(k) / G(k1) / / G0 / G
serve perfectly as a normal series.
On the other hand, if G is solvable, let
1 = Gs+1 / Gs / / G1 = G
be a normal series, then G/G2 is abelian, so G2 contains G0 , and similarly G3
contains G00 , and so on, since Gs+1 = 1. so 1 = Gs+1 G(s) = G(s) = 1.
More properties of solvable groups
Theorem 3.5 Subgroup and homomorphic images of solvable group is solvable, if
K / G and both K, G/K are solvable then G is solvable.
Proof. If H G, then H 0 G0 and so on, then G(k) = 1 = H (k) = 1 so H is
solvable.
If is a homo, then (G0 ) = ((G))0 by above so (G(k) ) = (1) = 10 = 1 =
((G))(k) provided you know that k is
Finally let K be as stated, let be the natural projection so (G) = G/K,
and (G(k) ) = ((G))(k) , so (G(k) ) = 1 = K for some k so G(k) K, then use
the normal series for K to patch together one series for G.
12

Notes on Galois theory

Tianyu Tao

A simple group G is that the only normal subgroup of G is 1 and G, for abelian
group the only simple group are the those cyclic group with prime order, since
otherwise we just find the p-groups and the are of course normal by abelian-ness.
Examples of simple groups An for n 5, P SLn (Fm ) except m = 2, 3.
Thus Sn is not solvable, since An is not abelian.

4 Galois group as permutation group of the roots

Gf , isomorphic to Gal E/F , is the subgroup of Sn on a proper ordering of the
roots {r1 , . . . , rn } {1, . . . n}.
Theorem 4.1 The discriminant is
Y

d = D2 :=

(ri rj )2 ,

1i<jn

the subfield of E/F corresponding to G An is precisely F (D) in the fundamental

pairing, and the Galois group of f (x) over F is a subgroup of An iff d is a square
in F .
Theorem 4.2 If f (x) F [x] has no multiple roots, then f (x) is irreducible in
F [x] iff Gf is transitive on the roots ri .

4.1 General equation of the nth degree

The Galois group of general equation of the nth degree is precisely Sn .

5 Cyclotomic extensions
We study cyclotomic extensions over Q, as an example for primitive nth root of
unity n , we will study the subgroups of G = Gal E/F and subfields of Q(n ) in
the correspondence established in the Fundamental theorem of Galois theory.

6 Algebraic Closures
The notion of algebraic closure of a base field F is a straight forward generalization
of the splitting field of a polynomial, namely, it is a field extension E of F such
that all the roots of every f (x) F [x] are contained in E, more precisely, we have
Definition 6.1 are the following:
13

Notes on Galois theory

Tianyu Tao

An algebraic closed field E is a field such that every f (x) E[x] has a root
in E.
A field extension E/F is called an algebraic closure of F if first: E/F is an
algebraic extension, and second: every f (x) F [x] splits completely in E.
(so, in particular, if u is algebraic over F in some extension field over F ,
since the minimal polynomial f (x) of u is in F [x], we have f (u) = 0 so u is
a root of f , and u E.)
Proposition Let E/F be an algebraic closure of F , then E is algebraically closed
Proof. Let f (x) E[x], we must show that it has a root in E, well, it always has
a root in some extension, let r be such a root, now r is algebraic over E, having
minimal polynomial dividing f (x).
Next, by definition, E is algebraic over F , we claim that r is algebraic over
F . As f (r) = 0 and f (x) E[x], let ai be the coefficients of f , we know ai E,
consider F (ai ), this is a subfield of E, and
[F (ai ) : F ] <
because each ai is algebraic over F by definition of E, join ai successively yield
the result. Now consider the extension F (ai , r), since
[F (ai , r) : F ] = [F (ai , r) : F (ai )][F (ai ) : F ],
note [F (ai , r) : F (ai )] < because F (ai ) have the coefficients of f , so f (r) = 0
and f (x) F (ai )[x], we conclude that F (ai , r) is algebraic over F since its a finite
extension over F , since r F (ai , r), we have r is algebraic over F , as desired.
Having obtained that r is algebraic over F , we conclude, by remark in the
second definition, that r E, done.
Let us prove uniqueness of algebraic closure of a given base field F first, we
shall use
Theorem 6.2 (Hausdorff Maximal principal, HMP) If X is a partially ordered
set, then X has a maximal linearly ordered subset.
Recall, a partial order on a set S is a relation like : any x, y, z S satisfies
x x; x y plus y x forces x = y; and if x y, y z, we will have x z.
A linear order on a set S is a partial order which also satisfies the condition
that for any x, y S, one of x y and y x must occur.
HMP is equivalent to Axiom of Choice.
Theorem 6.3 Let Ei , i = 1, 2 be algebraic closure of F , then E1 ' E2 are isomorphic with an isomorphism which is the identity on F .
14

Notes on Galois theory

Tianyu Tao

Recall how the uniqueness of splitting field was proved, we have a lemma which
uses a similar result:
Lemma 6.4 Let F, Ei , i = 1, 2 be as above, suppose K is that F K ( E1 , and
: K L E2 is an isomorphism such that is identity on F .
Then, there is a subfield K 0 E1 , with K ( K 0 and an isomorphism 0 of K 0
into E2 such that 0 extends .
proof of lemma. ...quotes page 227 of Basic...
Proof. Let X be the set consists of all pairs (K, ) where K is a subfield of E1 ,
with F K and : K E2 is an isomorphism which is identity on F , define a
partial ordering on X as follows:
(K, ) (K 0 , 0 )
if K K 0 and = 0 |K . Use HMP to get a maximal linear ordered subset L of
X. Put
[
H=
K,
K:,(K,)L

the union of all field K such that (K, ) L for some , then H is a subfield of
E1 , check directly... also (F, id) L shows F H.
Now we may define an isomorphism of H into E2 as follows: if a H, then
a K for some K for which there is such that (K, ) L, put
(a) := (a),
this is well defined, since, if (K 0 , 0 ) L, then either K K 0 and 0 |K = or the
other way around, anyway and 0 agrees on the smaller set. To see is indeed an
isomorphism, let a1 , a2 H, so ai Ki with corresponding i , wolg, say K1 K2 ,
2 extends 1 , so a1 , a2 K2 , we have
(a1 + a2 ) = 2 (a1 + a2 ) = (a1 ) + (a2 ) = (a1 ) + (a2 ),
and is identity on F , similarly check other conditions... so (H, ) X.
Next, we claim that (H, ) L, by definition, for every K such that there is
with (K, ) L, we have K H, furthermore, if (K, ) is in L, then (a) =
(a) for all a K, this follows from the fact that L is linearly ordered by its
definition and the remark of the well-definedness of . Hence (K, ) (H, ) for
each (K, ) L. Then, we must have (H, ) L, because if it werent, L fails to
be a maximal linearly ordered subset of X, as L {(H, )} is potentially a bigger
linearly ordered set.

15

Notes on Galois theory

Tianyu Tao

Finally, we claim that H = E1 . Suppose it is not the case, then F H ( E1 ,

and is an isomorphism from H to E2 which is identity on F , by the lemma we
obtain an extension H 0 of H with H ( H 0 and 0 extends which is an isomorphism
that sends H 0 into E2 , then
(H, ) < (H 0 , 0 )
since (H, ) is maximal in L, the set L {(H 0 , 0 )} is linearly ordered and properly
contains L, contradiction.
Therefore, (H, ) = (E1 , ) X, and is an isomorphism of E1 into E2 such
that is identity on F .
It then follows that, (E1 ) = E2 , because, since E1 is an algebraic closure of F ,
and is an isomorphism of E1 which fixes the element of F , it must be that (E1 )
is an algebraic closure of F , too, since (E1 ) E2 , they must be equal: let r E2 ,
we know r is algebraic over F so let f (x) F [x] be its minimal polynomial, then
f (x) (E1 )[x], so f (x) splits there since (E1 ) is also an algebraic closure of F ,
then r is equal to a root of f in (E1 ), so r (E1 ). Thus, we showed, E1 is
isomorphic to E2 via , which is the identity on F , assuming HMP.

16