221 Notes 10-30

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Math 221 : Algebra notes for Oct.

30

Alison Miller

1 More examples of Galois extensions


n
Example. L/K = Fpn /Fp - splitting field of xp x, so Galois, and Gal(Fpn //Fp ) = n.
Can write down an automorphism Frob Gal(Fpn /F) (the Frobenius automorphism)
given by Frob(a) = ap . (Since (a + b)p = ap + bp in characteristic p.)
We claim that Frob has order n, so that it generates Gal(Fpn /Fp ). Indeed, for a
n k
Fpn , Frobn (a) = ap = a. On the other hand, for k < n, the polynomial xp x has at
most pk < |Fpn | roots in Fpn , so there exists a Fpn such that Frobk (a) 6= a, so Frobp (a)
is not the identity. Hence Gal(Fpn /Fp ) is a cyclic group generated by Frob.
Any subgroup of H Gal(Fpn ) will then also be cyclic, generated by Frobr for some r
r
dividing n. In this case, the fixed field F = (Fpn )H consists of the roots of xp x in Fpn .
r n n
We can show that xp x divides xp x, so xp x splits into pr distinct linear factors
of in Fpn , and so F is a copy of Fpr inside Fpn .
(Alternatively, as suggested in class: use the Galois correspondence to show that
[L : Fp ] = r, so L must be isomorphic to Frp .)
Example. Now we do an example that we can construct by letting K be LG for some
subgroup G of Aut(L). Let L = C(x1 , . . . , xn ). We saw before that Sn acts on C[x1 , . . . , xn ]
with subring of invariants C[x1 , . . . , xn ]G = C[s1 , . . . , sn ]. By HW, this action extends to
C(x1 , . . . , xn ) with fixed field C(x1 , . . . , xn )Sn = C(s1 , . . . , sn ). (And we know that all the
si are algebraically independent, so this field is the isomorphic to the rational function
field in the variables si over C.)
We could also have constructed this as a splitting field, taking K = C(s1 , . . . , sn ) as
a field of rational functions in the variables si , and letting L be the splitting field of the
polynomial

(x x1 )(x x2 ) (x xn ) = xn s1 xn1 + s2 xn2 + + (1)n sn K[x].

For an example of the Galois correspondence: let H = An Sn . Then

[LAn : K] = [L : K]/[L : LAn ] = |Sn |/|An | = 2,

1
Q
so LAn is a quadratic extension of K. One can check that the element = i<j (xi xj )
is in LAn but not in K, so it generates LAn over K, and C(x1 , . . . , xn )An = C(s1 , . . . , sn , ).
One can also check that 2 C(s1 , . . . , sn ), confirming that we do have a quadratic
extension.
Example. Let K = Q, and let L = Q(n ), which we define first as the subfield of C
generated by Q and by n = e2i/n . The L is the splitting field of xn 1 over Q, since

xn 1 = (x 0n )(x n )(x 2n ) (x n1
n ).

Now, for any g Gal(Q(n )/Q), g(n ) must satisfy g(n )n = 1 but g(n )k 6= 1 for
any k < n. The first condition means that g(n ) = rn for some r > 0, and the second
implies that gcd(n, r) = 1
This gives us a map Gal(Q(n )/Q) (Z/nZ) , sending n to the r such that g(n ) =
rn . Easy to check its a group homomorphism; also, its injective because n generates
Q(n ). Whats hard is to show that its always surjective; well show this either Monday
or on the next problem set. (The statement of surjectivity is equivalent to the statement
Q
that the cyclotomic polynomial gcd(r,n)=1 (x rn ) Q[x] is irreducible.
Now we move on to do some corollaries to Galois theory.

Theorem 1.1. If L/K is a finite separable field extension, there are only finitely many intermedi-
ate fields F with K F L.

Proof. If this statement is true for some L 0 containing L, its also true for L. So, enlarge
L so that L/K is Galois; e.g. by taking generators 1 , . . . , n for L as a K-algebra and
letting L 0 be the splitting field of a separable polynomial which has 1 , . . . , n as some
of its roots.
Then this follows since Gal(L/K) has only finitely many subgroups.

We now prove a theorem which in some books is done before the Fundamental The-
orem of Galois theory (and is used in the proof); were deducing it as a corollary instead.

Theorem 1.2 (Primitive Element). Suppose that L/K is a finite separable field extension. Then
there exists L such that L = K[].

Proof. Well split into cases depending upon whether K is a finite or an infinite field. The
cases will use completely different arguments.
Case 1:|K| < . Then |L| < also. We use the following fact:
For any finite field L, L is a cyclic group. This can be proved using the classification
of finitely generated abelian groups: L = Z/d1 Z Z/d2 Z Z/dn Z for positive
integers d1 , . . . , dn such that d1 | d2 | | dn . Then the equation xdn = 1 is satisfied by
every x L , but on the other hand it can have at most dn roots in any field, so we must
have |L | dn . This is only possible if L = Z/dn Z is cyclic.

2
Now, let a L be such that a generates L as a cyclic group. Then a generates L as
a K-algebra as well, so L = K[].
Case 2: |K| = . For this, we use a lemma:
Lemma 1.3. Let V be a finite-dimensional vector space over an infinite field K. Then V cannot
be written as the union of finitely many subspaces Vi .

Proof of Lemma. Identify V = Kn . For each i, let `i K[x1 , . . . , xn ] be a linear function


Q
that vanishes on Vi . Let p K[x1 , . . . , xn ] = i `i . Then p is a nonzero polynomial
in n variables over an infinite field, so p is not identically zero. Hence there is some
(a1 , . . . , an ) Kn = V that is not in any Vi , so V is not the union.

Now, apply the lemma with V = L and the Vi being the finite set of intermediate
fields F with K F ( L. The lemma tells us that we can choose L such that /F
for any intermediate field F ( L. So K[] cant be any proper subfield of L, and must be
L itself.

(Note that I didnt remember to do in class: the non-separable extension L/K =


(x, y)/k(xp , yp ), where k is an infinite field, does not satisfy the conclusion of either
theorem above. To show that L is not of the form K[a]: [L : K] = p2 , but for any a L,
ap K so [K[a] : K] p. Also, the subfields K[a] for a = x + cy where c runs through
the elements of k can be seen to be all distinct.)

2 Coming up next: Galois theory and commutative alge-


bra
Well now combine what we know about Galois theory with some commutative algebra:
Let L/K be a Galois extension with Galois group K, let B be a subring of L such that
Frac(B) = A, let B be a subring such that gB = B for all g G, and let A = B K = BG .
Then by your current HW, K = Frac(A); and also note that we showed on a previous
HW that B is integral over A.
A special case of this construction is the following: let L/K be a Galois extension, let
A be a subring integrally closed in K, and let B be the integral closure of A in L.
Next time well talk about how the Galois group Gal(L/K) acts on the set of prime
ideals of B.

You might also like