2

Algebraic Extensions

1. Let K be a field and E an extension of K. One writes this assumption in short

as
Let E=K be a field extension;
and the word field is often omitted when it can be inferred from the context.
An element of E is called algebraic over K if there exists a polynomial
f .X / 0 in KX  such that
f ./ D 0:
If is not algebraic over K, we say that is transcendental over K.
Remarks. (a) If K D and E D , the elements of E algebraic over K are called
simply algebraic numbers, and the elements of Eptranscendental over K are
3
called transcendental numbers. Example: WD 2 is an algebraic number,
3
since is a root of the polynomial X  2 2 X .
(b) The set of algebraic numbers is countable (since X  is countable and any
nonzero polynomial in X  has finitely many roots in ). Therefore the set
of transcendental numbers must be uncountable. To actually be able to exhibit
a transcendental number is a different (and much harder) matter.
Theorem 1. Let M be a subset of containing 0 and 1. Any point z 2
algebraic over K WD .M [ M /.

M is

The proof will be given later in this chapter. But first we quote a famous result:
Theorem 2 (Lindemann 1882). The number  is transcendental.
Corollary. The quadrature of the circle with ruler and compass is impossible.
Proof. If it were possible, we would have  2 ; by Theorem 1 then  would be
algebraic, which by Lindemanns Theorem is not the case.

Lindemanns Theorem can be proved using relatively elementary algebraic and

analytic arguments, but the proof is on the whole quite intricate. We will go into it
later on (Chapter 17).

16

2 Algebraic Extensions

2. Now we start our study of field theory with the following statement:
F1. Let E=K be a field extension. If 2 E is algebraic over K, then
K./ W K < 1:
Proof. Suppose there exists a nonzero polynomial
(1)

f .X / D X n C an1 X n1 C    C a0 2 KX 

such that f ./ D 0; we have assumed without loss of generality that f is normalized

(has leading coefficient 1). There exists a unique homomorphism of K-algebras '
from the polynomial ring KX  into E such that '.X / D (see page 21); its image
R D im '  E
consists precisely of those elements of E that can be written as polynomial expressions g./ in with coefficients in K. But in writing such an expression we
immediately see from the relation
(2)

n D .an1 n1 C    C a1 C a0 /

that only terms of degree less than n are needed, so in fact

(3)

R D fc0 C c1 C    C cn1 n1 j ci 2 Kg:

Thus, as a vector space over K, the dimension of R is at most n. Since R, being

a subring of E, has no zero-divisors, a simple argument (given a bit further down)
shows that R is actually a field. It follows that K./  R (using the definition of
K./), and therefore that R D K./. From (3) we then get
(4)

K./ D fc0 C c1 C    C cn1 n1 j ci 2 Kg:

In particular,
(5)

K./ W K n.

F2. Let R be an integral domain (that is, a commutative ring with no zero divisors
and with 1 0), and let K be a subfield of R. If R is finite-dimensional as a K-vector
space, R is a field.
Proof. For a given a 0 in R, consider the map h W R ! R given by multiplication
by a, namely, h.x/ D ax for all x in R. Then h is an endomorphism (linear map)
of the K-vector space R. Since R has no zero-divisors, h is injective. Because R is
assumed finite-dimensional over K, it is also surjective. In particular, there exists
b 2 R such that ab D 1.

Remark. It can be proved in an analogous way that an integral domain that has
finite cardinality is a field.

The minimal polynomial

17

3. Let E=K be a field extension, and let 2 E be algebraic over K. Consider

on the K-vector space K./ the endomorphism h defined by multiplication by .
The minimal polynomial of h is called the minimal polynomial of over K, and we
denote it by
MiPoK ./:
This is the lowest-degree normalized polynomial in KX  that has as a zero. (That
there can be only one such polynomial is clear: if f; g are both normalized and of
degree n, the degree of f  g is less than n.) The degree of f D MiPo .K/ is also
called the degree of over K, and is denoted by W K.
Example. Consider E D , K D and D e 2 i=3 . Then is a root of X 3 1. But
X 3  1 D .X  1/g.X /, with g.X / D X 2 C X C 1; since 1, we have g./ D 0.
Let f D MiPoK ./; we claim that f D g. Otherwise necessarily deg f < deg g, so
f could only be of the form f .X / D X  , which is impossible since .
F3. Let E=K be a field extension and let 2 E be algebraic over K, of degree
n WD W K. The elements
1; ; 2 ; : : : ; n1

(6)

of E form a basis of K./ over K. In particular,

(7)

K./ W K D W K D deg MiPoK ./:

Proof. Let f .X / D X n C    C a1 X C a0 the minimal polynomial of over K. We

know that
K./ W K nI
see (5) in the proof of F1. There remains to show that 1; ; 2 ; : : : ; n1 are linearly
independent over K. Suppose there is a relation
(8)

n1
X

ci i D 0

with ci 2 K:

iD0

P
i
Set g.X / WD n1
iD0 ci X . If some ci in (8) were nonzero, g.X / would be a nonzero
polynomial in KX  of degree less than n and vanishing at . Contradiction!

4. Let E=K be a field extension and assume 2 E is algebraic over K. Is it the

case that any 2 K./ is also algebraic over K?
Definition. An extension E=K is called algebraic if every element of E is algebraic
over K. An extension E=K is called finite if E W K < 1.
Remarks. = is a finite extension, since W D 2. The extension = is not
algebraic; see Remark (b) in Section 2.1.
An extension E=K is called transcendental if it is not algebraic.

18

2 Algebraic Extensions

F4. If an extension E=K is finite, it is also algebraic; for each 2 E the degree
W K is a divisor of E W K.
Proof. Let E=K be finite of degree n. Given 2 E, the n C 1 elements 1; ;
2 ; : : : ; n of the n-dimensional K-vector space E are linearly dependent. Therefore
there exist a0 ; a1 ; : : : ; an 2 K, not all zero, such that
a0 1 C a1 C    C an n D 0:
Thus is algebraic over K. By F3, W K D K./ W K, and K./ W K is a divisor
of E W K by the degree formula (Chapter 1, F7).

We now can easily answer in the affirmative the question asked at the beginning
of this section.
F5. Let E=K be a field extension. If 2 E is algebraic over K, the extension
K./=K is algebraic.
Proof. If is algebraic over K, we know from F1 that K./=K is finite. But every
finite field extension is algebraic, by F4.

Together, F1 and F4 afford the following criterion:

F6. Let E=K be a field extension. An element of E is algebraic over K if and only
if K./=K is finite.
Now it is a cinch to prove Theorem 1, which we can reformulate as follows:
Theorem 1. Let M be a subset of containing 0 and 1. Let K D .M [ M /. The
field extension M =K is algebraic.
Proof. Take z 2 M . From F9 of Chapter 1 we know that K.z/ W K < 1. Then F6
says that z is algebraic over K.

Remark. The converse of F4 is not true: Not every algebraic extension is finite.
This will soon become obvious. In fact a counterexample comes up naturally in our
context: If E D f0; 1g is the field of all numbers constructible from f0; 1g with
ruler and compass, the field extension E= is algebraic but not finite. (With what
we know so far this is not very easy to prove, but its worth thinking about; see 2.5
in the Appendix.)
Among algebraic extensions, finite extensions can be characterized thus:
F7. Let E=K be a field extension. The following conditions are equivalent:
(i) There are elements 1 ; : : : ; m of E, finite in number and algebraic over K,
such that E D K.1 ; : : : ; m /.
(ii) E=K is finite.

Properties of algebraic extensions

19

Proof. (ii) ) (i) is clear; all we need to do is choose a basis 1 ; : : : ; m for E=K.
Then we actually have E D K1 C    C Km , and by F4 all the i are algebraic
over K.
To show (i) ) (ii) we use induction over m. For m D 0 there is nothing to
prove. Assume that (i) holds for some m  1 and set
K 0 D K.1 ; : : : ; m1 /:
Then E D K 0 .m /. Since m is algebraic over K, it is a fortiori algebraic over the
larger field K 0 . By F1 this implies E W K 0 < 1. But by the induction hypothesis,
K 0=K is finite. The degree formula (Chapter 1, F7) then implies that E=K is finite.

5. Let E=K be a field extension. A subfield L of E containing K is called an

intermediate field of the extension E=K.
F8. Let E=K be a field extension. The subset
F D f 2 E j is algebraic over Kg
is an intermediate field of E=K. It is called the algebraic closure of K in E. In
particular, the set of all algebraic numbers is a subfield of .
Proof. Take ; 2 F . Consider the subfield K.; / of E. By F7 the extension
K.; /=K is finite (prove this again for practice). Now apply F4; all elements of
K.; / are algebraic over K, so
K.; /  F:
The elements C, , and 1= (if 0) lie in K.; /, and thus also in
F . So F really is a subfield of E. Clearly K  F , since any 2 K is a zero of
a polynomial X  2 KX  and therefore algebraic over K. This completes the
proof.

This proof qualifies as easy, but its only easy because we have the right notions
at our disposal. Otherwise, would you be able to write down, at the drop of a hat, a
nontrivial rational polynomial that vanishes at the sum of two numbers, given only
rational polynomials vanishing at one and the other number respectively?
F9 (Transitivity of algebraicness). Let L be an intermediate field of the extension
E=K. If E=L and L=K are algebraic, so is E=K (and vice versa).
Proof. Take 2 E. By assumption is algebraic over L. Let 0 ; 1 ; : : : ; n1
be the coefficients of MiPoL ./; then is also algebraic over the subfield F WD
K.0 ; 1 ; : : : ; n1 /. By assumption all the i are algebraic over K. Therefore we
can apply F7 to conclude that F W K is finite. But F./ W F is also finite, by F6;
therefore the degree formula gives
F./ W K < 1:
Using F4 we see in particular that is algebraic over K.

20

2 Algebraic Extensions

F10. Let E=K be a field extension and A a subset of E. If all elements of A are
algebraic over K, the extension K.A/=K is algebraic.
Proof. Clearly K.A/ is the union of all subfields of the form K.M /, where M
ranges over finite subsets of A. By F7, each K.M /=K is finite and therefore also
algebraic. Thus K.A/ contains only elements algebraic over K. (Of course F10
also follows directly from F8.)

F11. Let E=K be a field extension, and L1 ; L2 intermediate fields of E=K. The field
(9)

L1 L2 WD L1 .L2 / D L2 .L1 /

is called the composite of L1 and L2 in E.

(a) If L1 =K is algebraic, so is L1 L2 =L2 .
(b) If L1 =K is finite, so is L1 L2 =L2 ; moreover L1 L2 W L2 L1 W K.
(c) If L1 =K and L2 =K are algebraic, so is L1 L2 =K.
(d) If L1 =K and L2 =K are finite, so is L1 L2 =K; if , moreover, the extension
degrees n1 D L1 W K and n2 D L2 W K are relatively prime, we have L1 L2 W K D
n1 n2 .
Proof. Part (a) follows from F10, taking (9) into account. Part (c) therefore also
follows, thanks to F9. Let L1 =K and L2 =K be finite. Assuming (b) already proved,
we see from the degree formula that
(10)

L1 L2 W K D .L1 L2 W L2 /.L2 W K/ .L1 W K/.L2 W K/;

which is the first part of (d). Again from the degree formula we obtain that L1 L2 W K
is divisible by n1 and by n2 . If n1 ; n2 are relatively prime, L1 L2 W K is divisible by
n1 n2 , which together with (10) gives the second part of (d).
There remains to prove (b). Consider the set R of all finite sums of products ab
with a 2 L1 ; b 2 L2 . Clearly R is a subring of E containing L1 and L2 . It is also
clear that any basis of L1 =K generates R as an L2 -vector space R, so in particular
R W L2 L1 W K. If L1 W K < 1, this implies that R is a field (see F2). It follows
that R D L1 L2 , which concludes the proof.

http://www.springer.com/978-0-387-28930-4