Chapter 6

Algebraic extension fields

Our goal remains: to show that there exist polynomials that are not solvable by radicals
over Q. In Chapter 4, we finally succeeded in properly formulating the notion of solvabil-
ity by radicals, which we did in the language of field extensions. Additionally, we were
able to catalog many polynomials that we are certain are solvable by radicals. Then, in
Chapter 5, we took a much closer look a polynomials, ultimately building a significant
amount of language and theory to analyze polynomial rings over fields. The conclusion of
the chapter hinted at the tight connection between polynomial rings and field extensions
where we saw that Q[x]/(x2 +1) Q(i). In this chapter, we will clarify this connection and
exploit it to significantly deepen our understanding of extension fields of the form Q(↵)
where ↵ is a root of some polynomial in Q[x].

6.1 Algebraic elements

Recall that a polynomial in F[x] is solvable by radicals over F if all of the roots of the
polynomial lie in some radical extension of F. Our focus is on roots of polynomials, and
the next definition gives us some language to highlight this.

Definition 6.1. Let F be a subfield of E. An element ↵ 2 E is algebraic over F if p(↵) = 0

for some nonzero p(x) 2 F[x]. If ↵ is not algebraic over F, then it is said to be transenden-
tal over F.

To show an element ↵ is algebraic over F, we need only produce a polynomial

p with
coefficients in F for
p which ↵ is a root. For example, the complex number 2 is algebraic
2
over Q because 2 is a root of p(x) = x 2 and p(x) 2 Q[x]. Also, ⇡ is algebraic over R
because ⇡ is a root of q(x) = x ⇡ and q(x) 2 R[x]. However, it is much harder to show
that ⇡ is not algebraic over Q (so ⇡ is transcendental over Q). Incidentally, a set-theoretic
argument shows that almost all elements of C are transcendental over Q; nevertheless,
we will focus on algebraic elements.
p
Problem 6.2. Show that each of the following are algebraic over Q: 11, 3 11, ⇣11 , and i.

Problem 6.3. Suppose that 2 C and 5 =2 2 7. Explain why is algebraic over Q.

CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

p
Problem 6.4. Let ↵ = 2 + i. Show that ↵ is algebraic over Q.

Now, an algebraic element over F is a root of some nonzero polynomial

p over F, but such
an element will be a root of lots of polynomials.
p For example, since 2 is a root
p of p(x)
p =
2
x p 2, we find that for every q(x) 2 Q[x], 2 is a root of q(x)p(x) because q( 2)p( 2) =
2
q( 2)p(0) =p 0. The polynomial x 2 is special because it is a polynomial of smallest degree
for which 2 is a root. In order to formalize this observation, we need to weave together
several results from the last chapter.
Let F be a subfield of E, and suppose that ↵ 2 E is algebraic over F. Then ↵ is the
root of some nonzero p(x) 2 F[x]. Let’s look at the set of all polynomials for which ↵ is
a root: I = {p(x) 2 F[x] | p(↵) = 0}. By Theorem 5.109, “evaluation at ↵” gives rise to a
homomorphism E↵ : F[x] ! E[x], and notice that I is precisely the kernel of E↵ . Thus, by
Theorem 5.116, I is an ideal of F[x], and since F[x] is a PID, it must be that I = (m(x)) for
some m(x) 2 F[x].
Now, m(x) is nonzero since I contains some nonzero polynomial, and this also implies
that m(x) is not a constant polynomial since the only constant polynomial that has roots
is the zero polynomial. Moreover, if m(x) is not monic, we can make it monic by mul-
tiplying by the inverse of the leading coefficient and the result will still generate I by
Theorem 5.103. So, we may assume that I = (m(x)) with m(x) nonconstant and monic.
Also, notice that by Theorem 5.91, m(x) divides every polynomial in I. So if I = (n(x))
for some other monic polynomial n(x), then m(x) and n(x) would divide each other. In
other words, m(x) = a(x)n(x) = a(x)b(x)m(x) for some polynomials a(x), b(x) 2 F[x]. Con-
sidering the degree of both sides of m(x) = a(x)b(x)m(x), we see that a(x) and b(x) must be
constant polynomials. But since m(x) = a(x)n(x) with m(x) and n(x) both monic, the only
conclusion is that a(x) = 1, so m(x) = n(x).
In summary, I = {p(x) 2 F[x] | p(↵) = 0} is an ideal, and there is a unique nonconstant
monic polynomial m(x) that generates I . In fact, more is true.

Lemma 6.5. Let F be a subfield of E, and suppose that ↵ 2 E is algebraic over F. Let
I = {p(x) 2 F[x] | p(↵) = 0}, and suppose that I = (m(x)) for some nonconstant m(x) 2 F[x].
Then m(x) is irreducible.

Combining Lemma 6.5 with our previous discussion, we arrive at the following fact.

Fact 6.6. Let F be a subfield of E, and suppose that ↵ 2 E is algebraic over F. Then there
is a unique irreducible monic polynomial m(x) 2 F[x] such that m(↵) = 0. Moreover, if
p(x) 2 F[x] and p(↵) = 0, then m(x) divides p(x).

The polynomial m(x) from Fact 6.6 gets a special name.

Definition 6.7. Suppose that ↵ 2 E is algebraic over F. The unique irreducible monic
polynomial m(x) 2 F[x] such that m(↵) = 0 is called the minimal polynomial of ↵ over F.
The degree of ↵ over F is defined to be the degree of the minimal polynomial of ↵ over F.

Theorem 6.8. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let
m(x) be the minimal polynomial of ↵ over F. If V = {p(x) 2 F[x] | p(↵) = 0} (i.e the set of
all polynomials that vanish at ↵), then V = (m(x)).

51
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

We will see shortly that the minimal polynomial of ↵ over F is key to understanding
the field extension F(↵). But how do we find the minimal polynomial of ↵ over F? The
first step is to find any monic polynomial p(x) 2 F[x] for which p(↵) = 0 (which also
verifies that ↵ is algebraic over F). If we can show that p(x) is irreducible, then p(x)
is the minimal polynomial, and we’re done. Otherwise, we factor p(x) into irreducible
polynomials in F[x], and by Fact 6.6, the minimal polynomial will be whichever one of
the factors has ↵ as a root. Let’s test this out with some examples.
Problem 6.9. Explain why the minimal polynomial for ⇣3 over Q is not x3 1. Find the
minimal polynomial ⇣3 over Q, and determine the degree of ⇣3 over Q?
Problem 6.10. The polynomial p(x) = x3 11 has three roots in C.
(1) Find the degree over Q of each of the three roots.

(2) Find the degree over R of each of the three roots.

Problem 6.11. The polynomial p(x) = x5 2x3 3x has five roots in C. Find the minimal
polynomial over Q of each of the five roots.
Problem 6.12. Let z = a + bi with a, b 2 Q, and define p(x) = (x z)(x z). Prove that
p(x) 2 Q[x], and then use this to find the minimal polynomial of 2 + i over Q.
p
Problem 6.13. Find the minimal polynomial of 3 5 over Q.

6.1.1 Describing elements of F(↵)

In Chapter 3, we explored extension fields of the form F(↵) where ↵ was chosen from
some larger field E. However, the definition of F(↵) was abstract and often hard to work
with. For example, Q(i) is defined to be the smallest subfield of C containing Q and i, but
this doesn’t tell us much about what the elements of Q(i) actually look like. Nevertheless,
in Problem 3.64, we were able show that Q(i) is precisely the p set of elements p of the form
a + bi with a, b 2 Q, and we also succeeded in showing that Q( 5) = {a + b 5 | a, b 2 Q}.
However,
p we noticed that describing Q(↵) is not always so easy since, for example, if
↵ = 2 + i, then Q(↵) , {a + b↵ | a, b 2 Q}. That said, we’ve learned a lot since Chapter 3,
so let’s take another go at tryingp to describe fields like Q(↵).
Remember that (for ↵ = 2 + i) we were able to show that {a + b↵ | a, b 2 Q} ✓ Q(↵), but
the reverse containment did not hold. And, the reason the reverse containment didn’t
hold was because {a + b↵ | a, b 2 Q} is not a field, which can be seen fairly easily since
↵ 2 < {a + b↵ | a, b 2 Q} (so it’s not closed under multiplication).
So why not add in ↵ 2 and consider something like {a+b↵+c↵ 2 | a, b, c 2 Q}? Remember-
ing that Q(↵) is a field containg Q and ↵ (and is closed under addition and multiplication),
we see again that {a+b↵+c↵ 2 | a, b, c 2 Q} ✓ Q(↵). So maybe {a+b↵+c↵ 2 | a, b, c 2 Q} = Q(↵).
Or, maybe we need to look at {a + b↵ + c↵ 2 + d↵ 3 | a, b, c, d 2 Q}. These are all good ideas
since a0 + a1 ↵ + a2 ↵ 2 + a3 ↵ 3 + · · · + an ↵ n 2 Q(↵) whenever a0 , a1 , a2 , a3 , . . . , an 2 Q. Let’s
formalize this.
Theorem 6.14. Let F be a subfield of E, and let ↵ 2 E. If p(x) 2 F[x], then p(↵) 2 F(↵).

52
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

p p
Problem 6.15. Let p(x) = 2x3 + 7x2 12 . We know that Q( 5) = {a + b 5 | a, b 2 Q} by
p p p p
Problem 3.66. Verify directly that p( 5) 2 Q( 5) by writing p( 5) in the form a + b 5 for
some a, b 2 Q.

We are studying the elements we get when we plug ↵ into polynomials. Let’s frame
this in terms of the evaluation homomorphism introduced in Theorem 5.109, like we did
at the beginning of this chapter. The next theorem generalizes Problem 5.118.

Theorem 6.16. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let
m(x) be the minimal polynomial of ↵ over F. If E↵ : F[x] ! E is the homomorphism
defined by E↵ (p(x)) = p(↵), then

(1) ker (E↵ ) = (m(x)), and

(2) im (E↵ ) ✓ F(↵) and F [ {↵} ⇢ im (E↵ ).

Using Theorems 5.105 and 5.124 (and the First Isomorphism Theorem for Rings), we
can deduce that the image of E↵ is a field. But then, if we can verify that the image of E↵
contains F and ↵, we can conclude that the image of E↵ is actually equal to F(↵) (by the
definition of F(↵)). Let’s put this together.

Theorem 6.17. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let
m(x) be the minimal polynomial of ↵ over F. Then F(↵) F[x]/(m(x)).

Notice that we still haven’t succeeded in providing a nice description F(↵), but we will
if we can find a nice description of F[x]/(m(x)).

Lemma 6.18. Let F be a subfield of E, and suppose that ↵ 2 E is algebraic over F. Let
m(x) be the minimal polynomial of ↵ over F, and let n be the degree of ↵ over F. If
a(x) + (m(x)) 2 F[x]/(m(x)), then a(x) + (m(x)) = r(x) + (m(x)) where r(x) is the remainder
obtained when dividing a(x) by m(x). Consequently,

F[x]/(m(x)) = {r(x) + (m(x)) | deg r(x) < n or r(x) = 0}.

Tying together all of our work, we finally arrive at our desired description of F(↵).

Theorem 6.19. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let n
be the degree of ↵ over F. Then
1
F(↵) = {a0 + a1 ↵ + · · · + an 1 ↵ n | a0 , a1 , . . . , an 1 2 F}.

We had to work hard for Theorem

p 6.19, but using it is fairly easy. Forpexample, sup-
pose wepwant to describe Q( 2). We first need to know the degree of 3 2 over Q. Of
3

course, 3 2 is a root of m(x) = x3 2, and m(x) is irreducible

p by Theorems
p 3.26 and 5.64.
3
Thus, m(x) = x 2 is the minimal polynomial 3 3
p of 2 overpQ, so p2 has degree 3 over Q.
Now we apply Theorem 6.19 to find that Q( 3 2) = {a0 + a1 3 2 + a2 ( 3 2)2 | a0 , a1 , a2 2 Q}.

Problem 6.20. Use Theorem 6.19 to describe Q(⇣3 ).

53
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

Problem 6.21. Let p(x) = x3 x + 1 2 Z3 [x], and let r be a root of p(x).

(1) Prove that p(x) is irreducible in Z3 [x].

(2) Use Theorem 6.19 to describe Z3 (r).

(3) How many elements are in the field Z3 (r)?

p
Problem 6.22. Let ↵ = 2 + i. In Problem 6.4, we saw that ↵ is algebraic over Q since ↵ is
a root of x4 2x2 + 9 2 Q[x]. Let m(x) denote the minimal polynomial for ↵ over Q.

(1) Use Theorem 6.19 (and Problem 3.64) to explain why the degree of m(x) isn’t 1 or 2.

(2) Use Fact 6.6 to explain why m(x) is a factor of x4 2x2 + 9.

(3) Explain why the degree of m(x) isn’t 3.

(4) Explain why m(x) = x4 2x2 + 9.

p
(5) Use Theorem 6.19 to describe Q( 2 + i).

The next problem highlights how we use the minimal polynomial for ↵ to compute
(or rather, simplify) in F(↵).

Problem 6.23. The polynomial p(x) = x5 + 2x + 2 is irreducible in Q[x] (you do not need to
prove this). Let s be a root of p(x). By Theorem 6.19, every element of Q(s) can be written
in the form a0 + a1 s + a2 s 2 + a3 s3 + a4 s 4 for some a0 , a1 , a2 , a3 , a4 2 Q.

(1) Use the fact that p(s) = 0 to write s 5 in the form a0 + a1 s + a2 s 2 + a3 s 3 + a4 s 4 .

(2) Rewrite each of the following elements of Q(↵) in the form a0 +a1 s+a2 s 2 +a3 s3 +a4 s 4 .

(a) (s 3 + 2)(s 3 + 3s)

(b) 3s 4 (2 + s3 )(5 s + s 2 )
p p ⇣ p ⌘
Problem 6.24. Let’s try to describe Q( 2, i). Notice that Q( 2, i) = Q( 2) (i). In words,
p
the field obtained
p by adjoining 2 and i at the same time is equal to the field obtained by
first adjoining 2 and then adjoining i to the result.
p
(1) Use Theorem 6.19 to describe Q( 2).
p
(2) Use Theorem 5.64 to explain why x2 + 1 is the minimal polynomial of i over Q( 2).
⇣ p ⌘ p
(3) Use Theorem 6.19 to describe Q( 2) (i), which is equal to Q( 2, i).
p p
(4) How does your description of Q( 2, i) compare to that for Q( 2+i) in Problem 6.22?

54
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

6.1.2 Eisenstein’s irreducibility criterion

Theorem 6.19 provides a nice description of F(↵) when ↵ is algebraic over F. However,
the description rests on us knowing the minimal polynomial of ↵ over F (or at least its
degree), which in turn rests on us being able to determine when polynomials are irre-
ducible. We now introduce a useful irreducibility criterion for polynomials in Q[x].
Fact 6.25 (Eisenstein’s Irreducibility Criterion (EIC)). Let p(x) = a0 + a1 x + · · · + an xn with
all a0 , a1 , . . . , an 2 Z. Suppose that there is some prime number p 2 Z such that all of the
following conditions are met:
(1) p does not divide an ,
(2) p does divide ai for all i < n, and
(3) p 2 does not divide a0 .
Then p(x) is irreducible in Q[x].
The proof of the EIC is interesting and not too difficult, but as it is more number-
theoretic than algebraic, we will leave it for another time. The idea is to take the poly-
nomial p(x) whose coefficients are all integers and consider it as a polynomial in Zp [x]
by reducing all of the coefficients modulo p. Details can be found in other books or on
Wikipedia.
Problem 6.26. Use the EIC to show that each of the following polynomials are irreducible
in Q[x]. What did you choose as your prime p? Were there other choices?
(1) f (x) = 7x4 + 6x3 + 12x 30
(2) g(x) = x8 6x5 30x3 + 12
Problem 6.27. Let ↵ be a root of p(x) = x5 + 5x4 5. (You don’t need to compute ↵!)
(1) Prove that p(x) is irreducible over Q.
(2) Use Theorem 6.19 to describe Q(↵).
5
Problem 6.28. Consider the polynomial f (x) = x7 2.
(1) Explain why the EIC does not apply to f (x).
(2) Prove that if f (x) is reducible in Q[x], then so is g(x) = 2x7 5.
(3) Use the EIC to show that g(x) is irreducible, and conclude that f (x) is irreducible.
Let’s try to use the EIC to determine the minimal polynomial of some special elements,
namely the ⇣n . We know that ⇣n is algebraic over Q because it is a root of xn 1. However,
xn 1 can not be the minimal polynomial since it has x 1 as a factor. But, we’ve observed
a few times now that
1 2
xn 1 = (x 1)(xn + xn + · · · + x2 + x + 1).
Set n (x) = xn 1 +xn 2 +· · ·+x2 +x +1. Since (⇣n )n 1 = 0, it must be ⇣n 1 = 0 or n (⇣n ) = 0.
Of course, ⇣n 1 , 0, so ⇣n is a root of n (x). Could n (x) be the minimal polynomial?

55
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

Problem 6.29. Show that 4 (x) = x3 + x2 + x + 1 is not the minimal polynomial for ⇣4 over
Q. What is the minimal polynomial?

So, we see that n (x) is not always the minimal polynomial for ⇣n , as it fails for n = 4.
But what about for n = 5? Could 5 (x) = x4 + x3 + x2 + x + 1 be the minimal polynomial
for ⇣5 ? We need to determine if 5 (x) is irreducible, but the EIC does not apply because
there is no prime that divides the non-leading coefficients. Let’s see if we can transform
5 (x) into a related polynomial for which the EIC will apply.

Theorem 6.30. Let F be a field, and let p(x) 2 F[x]. If p(x) is reducible in F[x], then p(x+1)
is also reducible in F[x].

Problem 6.31. Consider the polynomial 4 + x 3 + x 2 + x + 1.

5 (x) = x

(1) Compute 5 (x + 1), and write it in the form a4 x4 + a3 x3 + a2 x2 + a1 x + a0 . Consider

using the fact that x5 1 = (x 1) 5 (x) to help with the computation.

(2) Use the EIC to show that 5 (x + 1) is irreducible.

(3) Explain why 5 (x) is the minimal polynomial for ⇣5 over Q.

(4) Use Theorem 6.19 to describe Q(⇣5 ).

So, why was 5 (x) irreducible while 4 (x) was not? To address the general case, we
could use a similar approach as in Problem 6.31 to analyze n (x + 1) (using the Binomial
Theorem to simplify the expression).

Problem 6.32. Make a conjecture as to when n (x) = xn 1 + xn 2 + · · · + x2 + x + 1 is the

minimal polynomial for ⇣n over Q. That is, try to fill in the blank: “ n (x) is the minimal
polynomial for ⇣n over Q if and only if (something about n) .” If you have the time,
try to prove your conjecture.

To learn more about the minimal polynomial for ⇣n (for any n), try looking up “cyclo-
tomic polynomials” on Wikipedia.

6.2 Extension fields as vector spaces

When we work with complex numbers, we often write them in the form a + bi for a, b 2 R.
So, every complex number can be described using two real numbers: a and b. More-
over, each complex number is described by a unique choice of a and b. This allows us to
associate a complex number a + bi with a vector in R2 via
" #
a
a + bi 7! .
b

Additionally, adding two complex numbers a+bi and c+di corresponds to adding the two
associated vectors [ ba ] and [ dc ], and multiplying a + bi by a real number r corresponds to
multiplying the vector [ ba ] by the scalar r. But, we should be careful with multiplication

56
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

because multiplying a + bi and c + di does not correspond to multiplying the entries in the
corresponding vectors—do you see why? Nevertheless, we find that C is a vector space
over R, and the fact that every complex number can be described by a unique pair of real
numbers is expressing that C is a 2-dimensional vector space over R.
Can we do this for other fields? In Problem 6.27, we saw that if ↵ is a root of x5 +5x4 5,
then each element y 2 Q(↵) is of the form y = a + b↵ + c↵ 2 + d↵ 3 + e↵ 4 for a, b, c, d, e 2 Q. If
we knew that there was a unique choice of a, b, c, d, e for each y, then like before we could
associate each element of Q(↵) with a vector in Q5 via
2 3
66 a 77
66 77
66 b 77
6 7
a + b↵ + c↵ 2 + d↵ 3 + e↵ 4 7! 6666 c 7777 .
66d 77
66 77
4 5
e

We’ll see that these observations generalize to every extension field E of a field F. Let’s
start by properly defining vector spaces. Note that, in the formal definition below, scalar
multiplication by a number c is being written as c v, instead of just cv, but as we continue
on, we will return to writing simply cv.

Definition 6.33. Let F be a field. A vector space over F is a structure (V , +, { c | c 2 F})

consisting of a set V together with a binary operation + and a unary operation c for each
c 2 F (which we call addition and scalar multiplication by c) such that for some element
0 2 V the following axioms hold.

• Addition Axioms: Addition is associative and commutative; the element 0 is an

additive identity; every x 2 F has an additive inverse with respect to 0, denoted x.

• Distributivity Axioms: For all u, v 2 V and all c 2 F, c (u + v) = cu + c v.

• Compatibility Axioms: For all v 2 V and all c, d 2 F,

(1) c+d v = cv + d v,
(2) c ( d v) = cd v, and
(3) 1v = v.

Notice that the distributivity axiom can also be expressed by saying that each c is
a homomorphism from (V , +) to (V , +). The compatibility axioms can also be framed in
terms of a homomorphism, but we will not explore that here.
As mentioned above, we tend to omit writing the , so for example, the distributivity
axiom will be often written as c(u + v) = cu + cv.

Theorem 6.34. Let E be an extension field of F. Then E is a vector space over F where
vector addition for E is just the usual field addition for E and scalar multiplication by an
element c 2 F is just the usual field multiplication by c.

57
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

6.2.1 Degree of a field extension

We can now explore core concepts of linear algebra like linear independence, spanning
sets, bases, dimension, and linear transformations. Basic results from a first course on
linear algebra (over R) transfer to our more general setting, and you should feel free to
use them here.

Definition 6.35. Let V be a vector space over a field F, and let v1 , . . . , vn 2 V . Then

• v1 , . . . , vn are linearly independent if for all c1 , . . . , cn 2 F, c1 v1 + · · · + cn vn = 0 implies

that c1 = · · · = cn = 0;

• v1 , . . . , vn span V if for all w 2 V , there exist c1 , . . . , cn 2 F such that c1 v1 +· · ·+cn vn = w;

• v1 , . . . , vn form a basis for V if they are linearly independent and span V .

Notice that we have defined linear independence and span only for finite sets of vec-
tors, but the concepts can also be defined for infinite sets of vectors in a similar way. The
importance of bases is more-or-less summarized in the following fact.

Fact 6.36. If V is a vector space, then all bases have the same cardinality (“size”). If B
is any basis for V , then every element of V can be expressed as a linear combination of
vectors in B in one and only one way.

This leads to the notion of dimension, which when considering field extensions (as in
Theorem 6.34) we will refer to as the degree of the extension.

Definition 6.37. The dimension of a vector space V over a field F, denoted dim V , is the
cardinality of any basis for V .

Definition 6.38. If E is an extension field of F, then the dimension of E as a vector space

over F is called the degree of E over F, denoted [E : F]. If [E : F] is finite, we say that E is
finite dimensional over F.

Problem 6.39. Let ↵ be a root of p(x) = x5 +5x4 5. In Problem 6.27, we saw that p(x) is the
minimal polynomial for ↵ over Q and that Q(↵) = {a + b↵ + c↵ 2 + d↵ 3 + e↵ 4 | a, b, c, d, e 2 Q}.
By Theorem 6.34, Q(↵) is a vector space over Q.

(1) Explain why the elements 1, ↵, ↵ 2 , ↵ 3 , ↵ 4 span Q(↵) as a vector space over Q.

(2) Assume that c0 + c1 ↵ + c2 ↵ 2 + c3 ↵ 3 + c4 ↵ 4 = 0 for some c0 , . . . , c4 2 Q. Show that if

at least one of c0 , . . . , c4 is nonzero, then ↵ is a root of some nonzero polynomial that
has degree at most 4.

(3) Explain why the elements 1, ↵, ↵ 2 , ↵ 3 , ↵ 4 are linearly independent over Q.

(4) What is the degree of Q(↵) over Q? That is, find [Q(↵) : Q].

Generalizing our work in Problem 6.39, we obtain a crucial theorem.

58
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

Theorem 6.40. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let n be
the degree of ↵ over F. Then {1, ↵, ↵ 2 , . . . , ↵ n 1 } is a basis for F(↵) over F, and [F(↵) : F] = n.
Problem
p 6.41. Use Theorem 6.40
p and the results of Problem 6.10 to find a basis for
3 3
Q( 11) over Q and compute [Q( 11) : Q].
Problem 6.42. Use Theorem 6.40 and the results of Problem 6.21 to find a basis for Z3 (r)
over Z3 and compute [Z3 (r) : Z3 ] where r is a root of p(x) = x3 x + 1 2 Z3 [x].
Problem 6.43. Use Theorem 6.40 and the results of Problem 6.31 to find a basis for Q(⇣5 )
over Q and compute [Q(⇣5 ) : Q].
Theorem 6.40 (which built o↵ of Theorem 6.19) completes our goal of describing F(↵),
but we may want to adjoin more than one element to a field. For example, when we
defined solvability by radicals in Chapter 4, we needed to ensure that all roots of the
polynomial lived in some radical extension, often built by adjoining several elements.
Our approach to this will be as in Problem 6.24. Suppose we want a basis for F(↵, )
over F. We can first find a basis for F(↵) over F, and then find a basis for F(↵, ) over F(↵).
If 1, ↵, . . . , ↵ m 1 is a basis for F(↵) over F and 1, , . . . , n 1 is a basis for F(↵, ) over F(↵),
then we have that
1
F(↵) = {a0 + a1 ↵ + · · · + am 1 ↵ m | a0 , . . . , am 1 2 F}; and
n 1
F(↵, ) = {b0 + b1 + · · · + bn 1 | b0 , . . . , bn 1 2 F(↵)}.
But what we want is a basis for F(↵, ) over F, so we want to express elements of F(↵, )
using coefficients from F (not F(↵)). However, notice that since each bi is in F(↵), we
can write each bi in terms of 1, ↵, . . . , ↵ m 1 , using only coefficients from F. Doing this for
each bi and simplifying, we see that every element of F(↵, ) can be written as a linear
combination of the elements
1, ↵, . . . , ↵ m 1 ,
1
, ↵ , . . . , ↵m ,
..
.
n 1 n 1 1 n 1
,↵ , . . . , ↵m
using only coefficients from F. And moreover, it can be shown that these are linearly
independent, so we found a basis for F(↵, ) over F. The basis has size mn.
In fact, this process generalizes in a straightforward way to any chain of field ex-
tensions F ✓ K ✓ L. In words, a basis for L as a vector space over F can be found by
multiplying a basis for K over F by the elements of a basis for L over K. This also yields
an extremely useful multiplicative property for the degrees in a chain of field extensions,
namely that [L : F] = [L : K][K : F]. The next fact summarizes our findings.
Fact 6.44. Let F ✓ K ✓ L be fields. If {u1 , . . . , um } is basis for K over F and {w1 , . . . , wn } is
basis for L over K, then
{u1 w1 , . . . , um w1 , u1 w2 , . . . , um w2 , . . . , u1 wn , . . . , um wn }
is a basis for L over F. In particular, [L : F] = [L : K][K : F].

59
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

p
Problem 6.45. Let’s find a basis for Q( 4 2, ⇣3 ) over Q.
p
(1) Use the EIC to show x4 2 is the minimal polynomial of 4
2 over Q.
p
(2) Use Theorem 6.40 to find a basis for Q( 4 2) over Q.
p
(3) Use Theorem 5.64 to show x2 + x + 1 is the minimal polynomial of ⇣3 over Q( 4 2).
p p
(4) Use Theorem 6.40 to find a basis for Q( 4 2, ⇣3 ) over Q( 4 2).
p p
(5) Use Theorem 6.44 to find a basis for Q( 4 2, ⇣3 ) over Q and determine [Q( 4 2, ⇣3 ) : Q].

When exploring degrees of extensions, the following fact from linear algebra often
comes up: if W is a subspace of V , then dim W = dim V if and only if W = V . Applying
this to field extensions yields the following.

Fact 6.46. Let F ✓ K ✓ L be fields. Then [K : F] = [L : F] if and only if K = L.

p p
Problem 6.47. Let’s revisit the fields Q( 2 + i) and Q( 2, i).
p p
(1) Explain why Q( 2 + i) ✓ Q( 2, i).
p
(2) Use Theorem 6.40 and Problem 6.22 to determine [Q( 2 + i) : Q].
p
(3) Use Theorem 6.44 and Problem 6.24 to determine [Q( 2, i) : Q].
p p
(4) Use the previous parts to show that Q( 2 + i) = Q( 2, i).

The next couple of problems illustrate the power of the multiplicative property of
field degrees for a chain of fields.
p
Problem 6.48. Let’s take a look at Q( 4 2, ⇣3 ) over Q(⇣3 ).

(1) Explain why the EIC can not be used to show that x4 2 is irreducible in Q(⇣3 )[x].
p
(2) Use that Q ⇢ Q(⇣3 ) ⇢ Q( 4 2, ⇣3 ) together with p the multiplicative property of field
degrees (from pTheorem 6.44) to determine [Q( 4 2, ⇣3 ) : Q(⇣3 )]. Remember that you
computed [Q( 4 2, ⇣3 ) : Q] in Problem 6.45.
p p
(3) Use the fact that [Q( 4 2, ⇣3 ) : Q(⇣3 )] equals the degree of 4 2 over Q(⇣3 ) (by Theo-
rem 6.40) to explain why x4 2 is irreducible in Q(⇣3 )[x].
p p
Problem 6.49. Let’s show that 3 2 < Q( 2).
p p p p
(1) Explain why 3 2 2 Q( 2) would imply that Q( 3 2) ✓ Q( 2).
p
3
p
(2) Use the multiplicative
p property of fieldpdegrees to show that 2 2 Q( 2) would
3
imply that [Q( 2) : Q] is a divisor of [Q( 2) : Q].
p p p p
(3) What is [Q( 3 2) : Q]? What is [Q( 2) : Q]? Prove that 3 2 < Q( 2).
p p
Theorem 6.50. Let m, n 2 Z with 2  m < n. If p, q 2 Z are prime, then n p < Q( m q).

60
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

p
Problem 6.51. Let’s find the degree of Q( 5 3, ⇣5 ) over Q.
p
(1) Show that [Q( 5 3) : Q] = 5.
p p
(2) Use that ⇣5 is a root of x4 + x3 + x2 + x +p1 to explain why [Q( 5 3, ⇣5 ) : Q( 5 3)]  4.
Then Use Theorem 6.44 to show that [Q( 5 3, ⇣5 ) : Q]  20.
p
(3) Use Problem 6.31 and Theorem 6.44 to show [Q( 5 3, ⇣5 ) : Q] is divisible by 4.
p
(4) Explain why [Q( 5 3, ⇣5 ) : Q] = 20.

Let’s wrap up this section with a couple more results. The first says that if E is a finite
dimensional extension of F, then every element of E is in fact algebraic over F. To prove
this, we need to take an arbitrary r 2 E, and show it is a root of some nonzero polynomial
F[x]. But how do we find such a polynomial? To explore this, let’s let n = [E : F] (which
we are assuming is finite). This means that every basis for E over F consists of n elements.
And by a result from linear algebra, a set of n + 1 vectors must be linearly dependent. If
we apply this to the set {1, r, r 2 , . . . , r n }, we see that there are elements a0 , a1 , a2 , . . . , an 2 F
that are not all zero such that

a0 + a1 r + a2 r 2 + · · · + an r n = 0.

This implies that r is a root of the nonzero polynomial p(x) = a0 + a1 x + a2 x2 + · · · + an xn ,

so r is indeed algebraic over F. Also, r 2 E implies that F(r) ✓ E, so we now have that
[E : F] = [E : F(r)][F(r) : F]. Thus [F(r) : F] divides [E : F], so as [F(r) : F] equals the degree
of r over F (by Theorem 6.40), we also get that the degree of r over F divides [E : F]. Here
is the summary.

Fact 6.52. Let E be an extension field of F. Assume that [E : F] is finite. If r 2 E, then r is

algebraic over F, and the degree of r over F divides [E : F].

Combining this with Theorem 6.40, we obtain the following characterization of alge-
braic elements in terms of the fields they generate.

Corollary 6.53. Let E be an extension field of F, and let r 2 E. Then r is algebraic over F
if and only if [F(r) : F] is finite.

6.2.2 Linear transformations

Let’s briefly explore linear transformations in the context of field extensions.

Definition 6.54. Let V and W be vector spaces over a field F. A map : V ! W is called
an F-linear transformation (or homomorphism of F-vector spaces) if the following are
true for all u, v 2 V and all c 2 F:

(1) (u + v) = (u) + (v);

(2) (cu) = c (u).

61
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

p p p p
Problem 6.55. Show that : Q( 2) ! Q( 2) defined via (a+b 2) = a b 2 is a Q-linear
transformation.

Problem 6.56. Show that : C ! C defined via (z) = z is an R-linear transformation.

Problem 6.57. The results of Problem 6.24 show that

p n p p o
Q( 2, i) = a0 + a1 2 + a2 i + a3 i 2 | a0 , a1 , a2 , a3 2 Q .
p p p p
Consider
p : Q( 2, i) ! Q( 2, i) defined
p by (z) = z. As Q(
p 2, i) is an extension of Q( 2),
Q( 2, i) is a vector space over Q( 2). Show that is a Q( 2)-linear transformation.

6.3 Isomorphisms of fields

We are closing in on our goal of understanding when a polynomial is solvable by radicals
or not. Recall that p(x) 2 F[x] is solvable by radicals if all of the roots of p(x) are contained
in some radical extension of F. This implies that if r1 , . . . , rn are the roots of p(x), then
F(r1 , . . . , rn ) must also be contained in a radical extension. We’ve been working hard to
understand fields like F(r1 , . . . , rn ), and with Theorem 6.44, we are now able to explicitly
describe the elements of F(r1 , . . . , rn ) as linear combinations of a particular basis, which
involves certain powers of r1 , . . . , rn . But, we still need tools for analyzing F(r1 , . . . , rn ) in
order to understand if it could be contained a radical extension or not.
It turns out that the key idea is to study certain functions from F(r1 , . . . , rn ) to itself.
Remembering that F(r1 , . . . , rn ) is a field (hence a ring) and also a vector space over F, we
look at functions that preserve both structures, namely ring homomorphism that are also
F-linear transformations. The next theorem provides a convenient characterization of
F-linear transformation for maps that are already known to be ring homomorphisms.

Theorem 6.58. Let K and L be extension fields of F, and let : K ! L be a surjective ring
homomorphism. Then is an F-linear transformation if and only if (c) = c for all c 2 F.

In Theorem 6.58, the property that “ (c) = c for all c 2 F” will be written as fixes F
or leaves F fixed.

Definition 6.59. Let : X ! Y be a function.

• Let A ✓ X. We say that fixes A if (a) = a for all a 2 A.

• Define Fix( ) to be the set of all x 2 X such that (x) = x.

In fact, we’ve already seen many examples of morphisms that fix a field; the next
problem highlights two of them.

Problem 6.60. Let’s explore a couple familiar maps and show that they fix certain fields.
p p p p
(1) Define : Q( 2) ! Q( 2) via (a + b 2) = a b 2. Show that fixes Q.

(2) Define : C ! C via (z) = z. Show that fixes Q. What is Fix( )?

62
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

We now start to expose the implications of a morphism fixing a field. The next theo-
rem is quite important.

Theorem 6.61. Let K and L be extension fields of F, and let p(x) 2 F[x]. Suppose that
: K ! L is a ring homomorphism that fixes F. If ↵ 2 K is a root of p(x), then (↵) is also
a root of p(x).
p
Problem p 6.62. Suppose that : Q( 3 7) ! C is a ring homomorphism that fixes Q. Use the
fact that 3 7pis a root of x3 7 together with Theorems 3.26 and 6.61 to list the possible
values of ( 3 7).

Problem 6.63. Suppose that : C ! C is an isomorphism thatpfixes Q. Use Theorem 6.61

(and the idea of Problem 6.62) to list the possible values of ( 5) and (⇣5 ). Try to make
each list as short as possible, and explain your reasoning.

Problem 6.64. Show that each map below fixes Q, and then use Theorem 6.61 to explain
why neither map is a homomorphism.
p p
(1) : Q( 5) ! Q(i) defined by (a + b 5) = a + bi
p p p p p p
(2) : Q( 3 7) ! Q( 3 7) defined by (a + b 3 7 + c( 3 7)2 = a b 3 7 + c( 3 7)2

Let’s use
p what we’ve learned so far to explorep which subfields L of C could be isomor-
phic to Q( 3 11) (just as an example). Let : Q( 3 11) ! L be an isomorphism—we’ll try to
determine the possibilities for L. Notice that Theorem 6.61 may
p be helpful, but only if we
3
know that fixes the coefficients of a polynomial that has 11 as a root. Let’s first show
that must fix all of Q.

Problem 6.65. Let K and L be extension fields of Q. Suppose that : K ! L is a surjective

ring homomorphism. We’ll show that fixes Q.

(1) Use Theorem 5.111 (and properties of homomorphisms) to show that fixes every
positive integer n. Remember that 2 = 1 + 1, 3 = 1 + 1 + 1, etc.

(2) Use Theorem 5.110 to conclude that fixes all integers.

(3) Let ba 2 Q be any rational number (with a, b 2 Z). Use Theorem 5.111, to show that
fixes ba .
p
3
Problem 6.66. Assume that : Q( p 11)p! L is an isomorphism p for some subfield L of C.
3 3 2 3
Recall from Problem 6.41 that 1, 11, ( 11) is a basis for Q( 11), so
p p p
Q( 11) = {a + b 11 + c( 11)2 | a, b, c 2 Q}.
3 3 3

p
(1) Use Theorem 6.61 and Problem 6.65 to list the three possible values of ( 3 11).
p p
(2) Use Problem 6.65 to describe the possible values of (a + b 3 11 + c( 3 11)2 ).

(3) What does this imply are the possibilities for L?

63
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

Problem
p 6.66 raises an important question: can we always create an isomorphism
p
taking 11 to whenever is a root of the same minimal polynomial as 3 11? The
3

next fact answers the quesion—it will be extremely important for us. The proof is not too
difficult, but we will take it as fact. The main ingredients are some of the linear algebra
that we’ve developed and (perhaps not surprisingly) the division algorithm.

Fact 6.67. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F. Let m(x) be the
minimal polynomial of ↵ over F, and let n be the degree of ↵ over F. Suppose that is
also a root of m(x), and define : F(↵) ! F( ) by

(a0 + a1 ↵ + · · · + an 1 ↵ n 1 ) = a0 + a1 + · · · + an 1
n 1

for all a0 , a1 , . . . , an 1 2 F. Then is an isomorphism, and fixes F.

Problem 6.68. Use Theorem 6.61 (together with Problem 6.65) and Fact 6.67 to determine
if each pair of fields are isomorphic or not. If they are, write down a formula for an
isomorphism; if they are not, explain why not.
p
(1) Q(i) and Q( 2)
p p
(2) Q( 4 11) and Q(i 4 11)

6.3.1 Automorphisms
As mentioned earlier, we really want to study maps from a field to itself. We now define
one of the keys ingredients in our eventual solution to the insolvability of the quintic.

Definition 6.69. Let K be a field. An isomorphism from K to K is called an automor-

phism of K. We define Aut(K) to be the set of all automorphisms of K. If F is a subfield
of K, we define Aut(K/F) to be the set of all automorphisms of K that fix F.

Unpacking the definition, we see that automorphisms of K are bijections from K to

K (i.e permutations of K) that are also homomorphisms. Since we know that the compo-
sition of two bijections is a bijections and the composition of two homomorphisms is a
homomorphism (see Theorem 5.112), we see that Aut(K) is closed under function com-
position. Moreover, each element of Aut(K) is a bijection, hence has an inverse, and it is
not too difficult to show that the inverse is also a homomorphism. Thus, Aut(K) is closed
under taking inverses. Of course, Aut(K) contains the identity function from K to K, so
we conclude that Aut(K) is a group with respect to function composition.

Theorem 6.70. Let K be a field. Then Aut(K) is a group with respect to function compo-
sition. The identity is the identity function, which will be denoted id.

Let’s show that Aut(K/F) is also a group—to do that we need to show that if two
automorphisms of K fix F, then their composition does too.

Theorem 6.71. If F is a subfield of K, then Aut(K/F) is a subgroup of Aut(K).

64
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

In light of Theorems 6.70 and 6.71, Aut(K) will be referred to as the automorphism
group of K and Aut(K/F) will be called the automorphism group of K over F. While
we’re at it, let’s extend Theorem 6.71 a bit.
Theorem 6.72. If F ✓ K ✓ L is a chain of fields, then Aut(L/K) is a subgroup of Aut(L/F).
Theorem 6.72 is starting to build a correspondence from subfields of L to subgroups
of Aut(L/F). However, notice that the correspondence is inclusion reversing.

L Aut(L/L)

✓
K Aut(L/K)
✓

✓
F Aut(L/F)

In the picture above we listed the group Aut(L/L). These are the automorphism of L
that fix all of L—there is only one such automorphism: the identity. Thus, Aut(L/L) = {id}.
Let’s further explore these automorphism groups with some examples.
Example 6.73. Let’s try to compute Aut(Q(⇣5 /Q). First, we know that x4 + x3 + x2 + x + 1
is the minimum polynomial for ⇣5 over Q, so by Theorem 6.40
• {1, ⇣5 , ⇣52 , ⇣53 } is a basis for Q(⇣5 ) over Q, and
• Q(⇣5 ) = {a + b⇣5 + c⇣52 + d⇣53 | a, b, c, d 2 Q}.
Thus, every function 2 Aut(Q(⇣5 )/Q) can be expressed by a formula of the form
(a + b⇣5 + c⇣52 + d⇣53 ) = ???
Now, if 2 Aut(Q(⇣5 )/Q), then fixes Q. By Theorem 6.58, is a Q-linear transformation
from Q(⇣5 ) to itself, and by a result from linear algebra, is completely determined by its
values on a basis. That is, once we determine the values of (1), (⇣5 ), (⇣52 ), and (⇣53 ),
we will know a formula for . In fact, this is easy to see directly:
(a + b⇣5 + c⇣52 + d⇣53 ) = (a) + (b⇣5 ) + (c⇣52 ) + (d⇣53 )
= (a) + (b) (⇣5 ) + (c) (⇣52 ) + (d) (⇣53 )
= a + b (⇣5 ) + c (⇣52 ) + d (⇣53 ).
In fact, we can take this further:
(a + b⇣5 + c⇣52 + d⇣53 ) = a + b (⇣5 ) + c (⇣52 ) + d (⇣53 )
= a + b (⇣5 ) + c (⇣5 )2 + d (⇣5 )3 .
So, to find a formula for we just need to determine the value for (⇣5 ); it can then be
plugged into the above formula to find the value of on an arbitrary element of Q(⇣5 ).
Now, since ⇣5 is a root of x4 + x3 + x2 + x + 1, Theorem 6.61 says that (⇣5 ) must be one
of the roots of x4 + x3 + x2 + x + 1, which are ⇣5 , ⇣52 , ⇣53 , and ⇣54 . The possibilities are named
below. We will use repeatedly that ⇣55 = 1; if desired, we could also use ⇣54 = ⇣53 ⇣52 ⇣5 1.

65
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

2 3 2 3
• 1 sends ⇣5 7! ⇣5 , which implies 1 (a + b⇣5 + c⇣5 + d⇣5 ) = a + b⇣5 + c⇣5 + d⇣5 ;

• 2 sends ⇣5 7! ⇣52 , which implies 2 3 2 4

2 (a + b⇣5 + c⇣5 + d⇣5 ) = a + b⇣5 + c⇣5 + d⇣5 ;

• 3 sends ⇣5 7! ⇣53 , which implies 2 3 3 4

3 (a + b⇣5 + c⇣5 + d⇣5 ) = a + b⇣5 + c⇣5 + d⇣5 ;

• 4 sends ⇣5 7! ⇣54 , which implies 2 3 4 3 2

4 (a + b⇣5 + c⇣5 + d⇣5 ) = a + b⇣5 + c⇣5 + d⇣5 .

Here’s another way to organize the possibilities.

1 2 3 4

⇣5 7! ⇣5 ⇣52 ⇣53 ⇣54

We now have to determine if each i is in Aut(Q(⇣5 )/Q) or not. As 1 is just the identity,
1 2 Aut(Q(⇣5 )/Q). To check the other possibilities, we can use Fact 6.67. For exam-
ple, Fact 6.67 says that 2 is an isomorphism from Q(⇣5 ) to Q(⇣52 ). Since Q(⇣52 ) = Q(⇣5 )
(because ⇣52 2 Q(⇣5 ) and ⇣5 2 Q(⇣52 )), 2 is indeed an automorphism of Q(⇣5 ), so 2 2
Aut(Q(⇣5 )/Q). Similarly, 3 , 4 2 Aut(Q(⇣5 )/Q). These are all possibilities, so

Aut(Q(⇣5 )/Q) = { 1, 2, 3, 4 } = {id, 2, 3, 4 }.

We now know that Aut(Q(⇣5 )/Q) is a group of order 4. What group is it?

Problem 6.74. Let’s determine what group Aut(Q(⇣5 )/Q) is isomorphic to.

(1) Which element of Aut(Q(⇣5 )/Q) is ( 2 2

2) equal to? (Remember ( 2) means 2 2 .)

(2) Do the same for ( 2 )3 and ( 2 )4 . Are either ( 2 )3 or ( 2 )4 equal to id? What is the
order of 2 ? (The order of 2 will be the smallest positive k such that ( 2 )k = id.)

(3) What are the orders of 3 and 4?

(4) Is Aut(Q(⇣5 )/Q) cyclic or not? Is Aut(Q(⇣5 )/Q) isomorphic to Z4 or V4 ?

An important observation in Example 6.73 was that the possibilities for are deter-
mined simply by the possible values of (⇣5 ). This is true in general.

Fact 6.75. Let F be a subfield of E. Suppose that ↵ 2 E is algebraic over F, and let n be
the degree of ↵ over F. Then each 2 Aut(F(↵)/F) is completely determined by the value
(↵). Consequently, | Aut(F(↵)/F|  n = [F(↵) : F].
p
Problem 6.76. Follow Example 6.73 to determine Aut(Q( 2)/Q). What familiar group is
it isomorphic to?
p
Problem 6.77. Let’s determine Aut(Q( 3 2)/Q).
p p p p
(1) First explain why Q( 3 2) , Q( 3 2⇣3 ) and why Q( 3 2) , Q( 3 2⇣32 ).
p
(2) Follow Example 6.73 to show that Aut(Q( 3 2)/Q) = {id}.

66
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

What happens if we adjoin more than one element? Can we compute Aut(F(↵1 , ↵2 )/F)
in a similar way as to how we computed Aut(F(↵)/F)? The answer is yes, and the starting
point is the following analog of Fact 6.75.

Fact 6.78. Let F be a subfield of E. Suppose that ↵1 , . . . , ↵k 2 E are algebraic over F. Then
each 2 Aut(F(↵1 , . . . , ↵k )/F) is completely determined by the values of (↵1 ), . . . , (↵k ).
Consequently, | Aut(F(↵1 , . . . , ↵k )/F|  [F(↵1 , . . . , ↵k ) : F].
p
Problem 6.79. Let’s determine Aut(Q( 2, i)/Q).
p p
(1) Let 2 Aut(Q( 2, i)/Q). Explain why there are only two choice for ( 2) and only
two choice for (i). What are they?
p
(2) Combine the di↵erent possibilities for ( 2) and (i) to complete the table below.

1 2 3 4
p p
2 7! 2
i 7! i
p
(3) Follow Example 6.73 to determine Aut(Q( 2, i)/Q).
p
(4) What familiar group is Aut(Q( 2, i)/Q) isomorphic to?
p
Problem 6.80. Set L = Q( 2, i). In Problem 6.79 we determined Aut(L/Q). Let’s connect
the subfields of L with subgroups of Aut(L/Q) using Theorem 6.72.
p
(1) Let K1 = Q( 2). Find Aut(L/K1 ) by determining which of 1 , 2 , 3 , 4 are in Aut(L/K1 ).

(2) Repeat for K2 = Q(i). Find Aut(L/K2 ).

p
(3) Repeat for K3 = Q(i 2). Find Aut(L/K3 ).

(4) Use Theorem 6.72 to organize your findings by writing the appropriate elements in
the boxes in the subgroup lattice of Aut(L/Q).

p
Q( 2, i) {id}

p p
Q( 2) Q(i) Q(i 2) Aut(L/K1 ) Aut(L/K2 ) Aut(L/K3 )

Q { 1, 2, 3, 4}

67
CHAPTER 6. ALGEBRAIC EXTENSION FIELDS

Problem 6.80 highlights quite well the tight connection between subfields of an ex-
tension field L of F and subgroups of Aut(L/F). So far, we’ve seen how each subfield K
gives rise to a subgroup Aut(L/K). The next theorem indicates how we might reverse this.

Theorem 6.81. Let F be a subfield of L and H a subgroup of Aut(L/F). Define

FixL (H) = {k 2 L | k is fixed by every 2 H}.

Then FixL (H) is a subfield of L, and F ✓ FixL (H) ✓ L.

The picture is as follows.

L {id}
✓

✓
FixL (H) H
✓

✓
F Aut(L/F)

Taking a closer look at Problem 6.80, we can see that the maps K 7! Aut(L/K) and H 7!
FixL (H) are actually inverses of each other. For example, FixL (Aut(L/K1 )) = K1 , so the
composition of the maps looks like K1 7! Aut(L/K1 ) 7! FixL (Aut(L/K1 )) = K1 . However,
this is not true for all fields, and Problem 6.77 gives an example. In the next chapter
we’ll study an important collection of fields (in fact, the collection of fields) for which
K 7! Aut(L/K) and H 7! FixL (H) are always inverses.

68