Lang Chapter 6
Lang Chapter 6
KELLER
Problem 1
τ : i 7→ −i
We see that
τ σ(α) = i3 α
= σ 3 (α)
Thus the Galois group is nonabelian and has an element of order 4.
The only possibility left is that Gal(Q(α, i)/Q) = D8 . By the Galois
correspondence,
√
Gal(Q(α, i)/Q( 5)) = Z/2Z
√
Gal(Q(α, i)/Q( −5)) = Z/2Z
And as the above individual fields intersect trivially, we know that the
Galois group is the product of the Galois groups of each extension in
the compositum. Each of the above are quadratic extensions, so we see
that
√ √ √ √
Gal(Q( 2, 3, 5, 7)/Q) = (Z/2Z)4
(Z/2Z)n
(`). Since the splitting fields intersect trivially, we find may find each
Galois group separately and then take the product. We see that x3 − 2
has Galois group S3 . Similarly, x3 − 3 has Galois group A3 since the
√
discriminant is a perfect square over Q( −3). Similarly, x2 − 2 has
Galois group Z/2Z.
Thus the compositum has Galois group
S3 × A3 × Z/2Z ∼
= S3 × Z/6Z
[C(α) : C(t)] = n
Gal(C(α)/C(t)) = Z/nZ
4 KELLER
σ : α 7→ iα
τ : i 7→ −i
Problem 2
(a). Observe that x3 +x+1 is irreducible since the only possible rational
roots are ±1 by the Rational Root Theorem. We see as well that the
discriminant is −31; as this is not a perfect square, the Galois group is
isomorphic to S3 .
Problem 3
(a). Note that irreducibility over C(t) is the same as irreducibility over
C[t] by Gauss’ Lemma. As x3 + x + t is linear in t, this is certainly
irreducible. We also see that the discriminant is −27t2 − 4, which is
not a perfect square in C(t). Thus the Galois group is S3 .
nontrivially:
x3 + t2 x − t3 = (x − f (t))(x2 + g(t)x + h(t))
=⇒ f (t)h(t) = t3
h(t) − g(t)g(t) = t2
This forces f (t) = t, h(t) = t2 , and g(t) = 0; this however implies
the above has root t, which is a contradiction. Thus x3 + t2 x − t3 is
√ 2
irreducible. As this has discriminant −31t6 = −31t3 which is a
perfect square, the Galois group is A3 .
Problem 4
√
(1) =⇒ (2): Suppose E/k is Galois. We ahve that α has minimal
polynomial f (x) := x4 − 2ax2 + a2 − b2 c. By definition of a Galois
extension, adjoining any other root of f induces the same extension.
Thus,
√
F ( α0 ) = E
√
Since α0 is another root of f .
√ √
(2) =⇒ (3): Assume F ( α) = F ( α0 ). Then, the nontrivial
σ ∈ Gal(E/F ) acts by
√ √
α 7→ − α
√ √
α0 7→ − α0
√ √ √ √
Implying α · α0 is fixed by σ; that is, α · α0 ∈ F .
Similarly, the only nontrivial σ ∈ Gal(F/k) is such that
√ √
c 7→ − c
√ √ √ √ √ √
and either α · α0 remains fixed or α · α0 7→ − α · α0 . If σ fixes
√ √ 0
α · α , then by definition α · α0 ∈ k 2 . If not, we see:
√ √ √ √
σ( c · αα0 ) = c · αα0
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 7
whence cαα0 ∈ k 2 .
(3) =⇒ (2): Suppose αα0 ∈ k 2 or cαα0 ∈ k 2 . In either case, this
√ √ √
tells us that αα0 ∈ F , implying α0 ∈ E. However, [F ( α0 ) : F ] = 2,
√ √ √ √
and F ( α) ⊇ F ( α0 ), so that F ( α) = F ( α0 ).
(2) =⇒ (1): Let σ ∈ Embk (E, E) where E denote an algebraic
√ √
closure of E. We have that σ( c) = ± c since it must be a root of
x2 − c.
√ √ √ √
If σ( c) = c; this implies σ(α) = α, so σ( α) = ± α.
√ √ √
Similarly, if σ( c) = − c, we have that σ(α) = α0 so that σ( α) =
√
± α. The above also tells us that for any σ ∈∈ Embk (E, E),
√ √ √ √
σ F ( α) = F ( α) or σ F ( α) = F ( α0 )
√ √ √
As we have assumed F ( α) = F ( α0 ), this tells us that F ( α) is
invariant under any embedding into E; that is, E is Galois over k.
√
Assume that Gal(E/k) is cyclic; note that it is order 4 since α has
a degree 4 minimal polynomial. Let σ be a generator of the Galois
√ √
group; σ is completely determined by its action on c and α. We
√ √
must have σ( c) = − c, since else σ 2 = id. This also implies that
√ √ √ √
σ( α) = ± α0 ; by condition (3), either αα0 ∈ k or cαα0 ∈ k.
√ √ √ √
If αα0 ∈ k, then we must have that σ 2 ( α) = α and σ 2 ( α0 ) =
√
α0 . This would again imply σ 2 = id, which is a contradiction.
√
This leaves the only possibility cαα0 ∈ k as contended. Conversely,
consider the automorphism
√ √
σ : c 7→ − c
√ √
α 7→ α0
8 KELLER
√ √
Then σ 2 ( α) = − α, in which case we deduce that σ must be an
order 4 automorphism of the Galois group Gal(E/k). Thus, Gal(E/k)
is cyclic and generated by σ as above.
Problem 5
[k(γ); k] = 12/2 = 6
as contended.
Problem 6
σ(z)z = −1
as desired.
(b). Suppose that zτ (z) = −1. Squaring both sides of this equality,
we find that z 2 τ (z 2 ) = 1. Observe that z 2 τ (z 2 ) = NE/k (z 2 ); employing
Hilbert’s Theorem 90, we deduce that there exists some γ ∈ E such
τ (γ)
that z 2 = γ
.
Let α denote a root of x2 − γ in some algebraic closure and set
K = k(α). Denote by mα,k (x) the minimal polynomial of α over k;
we must have that mα,k (x) is separable since else mα,k (x) = (x2 − γ)2 ,
implying γ ∈ k. This is a contradiction, and we deduce K is Galois as
it is the splitting field of a separable polynomial, namely mα,k (x).
To see cyclicity, let σ be any extension of τ to k(α). Since τ 2 ≡ id, we
have that E is the fixed field of σ 2 . Thus, σ 2 (α) must be sent to another
root of x2 − γ, implying σ 2 (α) = −α. Since σ is not order 2, it must be
10 KELLER
σ(α)2
z2 =
α2
σ(α) σ(α)
=⇒ z − z+ =0
α α
=⇒ σ(α) = ±zα
Problem 7
√
Q → Q( α) ,→ F
Gal(F/Q) = Z/nZ
√
Since α < 0, we know that α is pure imaginary. Consider the au-
tomorphism τ : F → F corresponding to complex conugation. This
has order 2 and hence by cyclicity is the unique such element. By the
Galois correspondence, we have the following diagram:
F 1
√
Q( α) hσ 2 i
Q hσi
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 11
√
So that [F : Q( α)] = n/2. Since n/2 remains even, the fixing auto-
√
morphisms of Q( α) must contain complex conjugation. But this is a
contradiction, whence we conclude that no such embedding can exist.
= 15 ∈ Q
By Problem 4, this immediately tells us that the Galois group is cyclic
of order 4.
=2∈Q
By Problem 4, we deduce that the Galois group is cyclic of order 4.
Problem 8
(a). Let α and β denote two roots of our polynomial. Consider the
following diagram:
α β
−β −α
Then every automoprhism of Gal(K/Q) induces a symmetry of the
above square. By definition of the dihedral groups, this implies that
Gal(K/Q) is a subgroup of D8 , whence
Gal(K/Q) ∼
= D8 , Z/4Z, or Z/2Z/2Z × Z/2Z
12 KELLER
(b). Let us first consider the case for Gal(K/Q) = Z/4Z. We have
that Q(α) = Q(β) by the Tower law. If D := a2 − 4b2 , we know
√
that |bbq( D) is contained in both of the aforementioned fields. Let σ
generate Gal(K/Q); then σ(α) = β and
√
−a + σ( D)
2
σ(α ) =
2
2
=β
√
−a − D
=
2
√ √
=⇒ σ( D) = − D
Thus σ fixes α
β
− αβ ; since our fixed field is Q, we have that α
β
− αβ ∈ Q.
Assume now that Gal(K/Q) = Z/2Z × Z/2Z. We have two cases:
in the first case we have generators φ, ψ such that:
φ : α 7→ −α
β 7→ β
ψ : α 7→ α
β 7→ −β
φ : α 7→ β
β 7→ α
ψ : α 7→ −β
β 7→ −α
√ √ √ √
q q q q
2 − 5, − 2 − 5, 2 + 5, − 2 + 5
Set
√ √
q q
α := 2− 5, β := 2+ 5
As α ∈
/ R, this is a degree 8 extension with splitting field Q(i, β) and
Gal(Q(i, β)/Q) = D8 . With the standard notation for the dihedral
group,
D8 = hr, s | r4 = s2 = 1, rsrs = 1i
We have that
r(α) = β, r(β) = −α
and
s(α) = β, s(β) = α
14 KELLER
D8
2
4 4
hri 4
Let us first examine the fixed field of hri. Note that r should always
√ √
fix D · αβ, in which case we deduce that the fixed field is Q( D · αβ)
since it will be a quadratic extension.
Next, consider hsi. We know that s(α) = β, s(β) = α, and we see
that the fixed field is
Q(α + β)
sr(α) = α, sr(β) = −β
Q(α, β)
2
2
2
4
Q(α + β) Q(α) Q(β)
√
Q( Dαβ) 4
4 4
2
Q
And this completes the problem.
Problem 9
Set Ei := k(θi ). We may embed the splitting field of θi into its Galois
closure F . Consider the equivalence relation
θi ∼ θj ⇐⇒ Ei = Ej
As the set of roots has order p, with p prime, we know that Gal(F/k)
either acts transitively or trivially. It cannot be a trivial action since
else all θi ∈ k; thus Gal(F/k) acts transitively.
As we have assumed θ1 ∼ θ2 , we use transitivity to deduce that all
roots are related. That is, k(θ1 ) = k(θi ) for all i > 2. This tells us that
k(θ) is normal and separable, hence Galois. Moreover, as [k(θ) : k] = p,
the Galois group is cyclic of order p, so this is a cyclic extension.
16 KELLER
Problem 10
(a). Note that Gal(K/Q) acts transitively on the set of roots of f . This
is equivalent to equivalent to irreducibility of f , so f is irreducible as
contended.
[Q(α, ζn ) : Q] = nφ(n)
6 n(n − 1)
Problem 11
1 = NF/k (1)
= NF/k (αα)
= NF/k (α)2
Problem 12
Problem 13
For each part of this problem, recall that there exists an embedding
Gal(F/Fp ) ,→ Gal(F/Q) as this will not be mentioned explicitly.
x4 + 2x2 + x + 3 ≡ x(x3 − x + 1)
x4 + 2x2 + x + 3 ≡ x4 + x + 1
(x2 + ax + b)(x2 + cx + d) = x4 + x + 1
Gal(F/Q) ∼
= S4
x4 + 3x3 − 3x − 2 ≡ x(x3 + x2 + 1)
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 19
Gal(F/Q) = S4
x6 +22x5 −9x4 +12x3 −37x2 −29x−15 ≡ x6 +x5 −x2 +x = x(x5 +x4 +2x+1)
Gal(F/Q) = S6
20 KELLER
Problem 14
g := φ1 · f3
g := (x − 1)φ1 f3
Gal(F/Q) = Sn
as desired.
Problem 15
Define
H := {σ ∈ Gal(K/k) | σ(F ) = F }
=α
Whence σ ∈ NG (Gal(F/k)). Conversely, suppose σ(F ) 6= F for some
σ ∈ Gal(K/k). Thus for all τ ∈ Gal(K/F ), there exists α ∈ F such
that σ(α) ∈
/ F , implying τ (σ(α)) 6= σ(α).
Inverting, σ −1 τ σ(α) 6= α so that σ ∈
/ NG Gal(K/F ) . Thus we
conclude that H = NG Gal(K/F ) as contended.
Problem 16
Problem 17
(b). Note that if [F : Q] = n, then the only roots of unity that could
be contained in F are the kth roots of unity for which φ(k) 6 n. The
set of integers k for which φ(k) 6 n is finite, so that F can only contain
finitely many roots of unity.
Problem 18
(a). Observe that i is a primitive 4th root of unity, so that Q(i) contains
a 4th root of unity.
√
For Q( −2), there are no roots of unity other than −1. The same
√
holds for Q( 2).
√ √ √ √
In Q( −3), we see that −1
2
+ i 2
3
∈ Q( −3), so that Q( −3) has a
primitive 3rd root of unity and hence a primitive 6th root of unity by
Problem 17 part (a).
√ √
In Q( 3) there are no nontrivial roots of unity. Also, in Q( −5)
there are no nontrivial roots of unity.
(b). The only integers for which an mth root of unity can be of degree
2 are such that φ(m) = 2, so that m = 3, 4, or 6.
Problem 19
[Q(ζ) : Q] = pr − pr−1
Note that
is cyclic. Then,
Y
NQ(ζ)/Q (1 − ζ) = 1 − σi (ζ)
i
= Φpr (1)
= Φp (1) = p
(b). By part (a), NQ(ζ)/Q (1 − ζ) = Φn (1). Recall that by Möbius in-
version,
Y
Φn (x) = (xd − 1)µ(n/d)
d|n
We want to show that Φn (1) = 1 for all n such that n has multiple
distinct prime factors, with base case for n = 6. Since Φ6 (x) = x2 −
x + 1, we see that Φ6 (1) = 1.
In the general case, write n = pm with m > 1 and p prime. We see:
Φm (xp )
Φn (x) =
Φm (x)
So that Φn (x) = Φm (1)/Φm (1) = 1, and we conclude that NQ(ζ)/Q (1 −
ζ) = 1 as desired.
Problem 20
Pn
Suppose f (x) = i=0 ai xi . Without loss of generality we may as-
sume a0 is nonzero since else f (p) ≡ 0 mod p for all p. Now,
n−1
X
i+1 i
f (a0 x) = a0 1 + x · a0 x ai+1
i=0
Now set
n−1
X
g(x) := ai+1 ai+1
0 x
i
i=0
Then there are only finitely many solutions to 1 + xg(x) = 1 and
1 + xg(x) = −1. Let M > 0 and choose N ∈ N such that N > M and
N ! is greater than any solution to the above system. Then there must
exist a prime p such that p | 1 + N !g(N !).
24 KELLER
Problem 21
Φn (M np1 · · · pk ) > 1
Φn (M np1 · · · pk ) ≡ 1 mod pi
We then see that we have found a new distinct prime divisor, implying
that the set of primes dividing Φn (a) as a ranges through all integers
is infinite.
Combining this with part (a), we see that there are infinitely many
primes such that p ≡ 1 mod n, which yields a special case of Dirichlet’s
theorem on primes in arithmetic progression.
Problem 22
Problem 23
G∼
Y
= Z/di Z
i
∼
Y Z
=
i
(pi − 1)Z
By construction,
Y Z
G6
i
(pi − 1)Z
Moreover we may find H 6 Gal(Q(ζn )/Q) such that
Gal(Q(ζn )/Q)/H ∼
=G
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 27
(b). By Problem 21, there are infinitely many primes satisfying the
conditions imposed in part (a). Thus, there are infinitely many abelian
Galois extensions of Q with Galois group G.
Problem 24
mα,Q (x + a3 /3) = x3 + ax + b
Problem 25
(E, σE ) 6 (E 0 , σE 0 ) ⇐⇒ E ⊂ E 0 and σE 0 |E = σE
Observe that every chain under the above partial relation has upper
bound
[
( Eλ , σ)
λ
with σ ∈ Aut(E) such that σ|Eλ := σEλ . By Zorn’s Lemma, S has a
maximal elements (M, σ). It remains only to show that M = k a .
Toward this end, suppose not. We may choose α ∈ k a \M with
[M (α) : M ] minimal. We may extend σ to an automorphism σ 0 of the
algebraic closure k a . Then there exists β ∈ k a such that β ∈
/ k and
σ 0 (β) = β by maximality of M . Consider the minimal polynomial of
β over M , denoted f . The coefficients of f are fixed by σ 0 ; moreover,
f (x) ∈ M [x] so that σ 0 (f ) = σ(f ). This tells us that f (x) ∈ k[x] since
the fixed field of σ is k.
By the assumption that all finite extensions are cyclic, we deduce
that k(β) is the unique extensions of k with degree [k(β) : k]. Choose
any automorphism mapping β to another root of f , denoted τ .
Like before, τ (γ) = γ for some γ ∈
/ M . We also have that mγ,M (x) ∈
k[x] and hence [k(γ) : k] = [k(β) : k], implying k(γ) = k(β) by cyclicity.
However, by definition of τ , this automorphism fixes β, implying τ
fixes γ. This contradiction implies that M = k a , and hence σ is an
automorphism of k a with fixed field k, as desired.
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 29
Problem 26
√
Set α := 2 and note that α can in fact be any algebraic number in
this proof.
Claim: E(α) = Qa . To see this, suppose not. Then, there exists
β ∈ Qa \E(α) such that E(β) ⊃ E properly and α ∈
/ E(β). However,
this containment contradicts maximality of E, whence we must have
E(α) = Qa .
As α is algebraic, [k(α) : k] < ∞ so that [E(α) : E] 6 [k(α) : k] < ∞.
Thus, the algebraic closure Qa is a finite extension of E. By Corollary
9.3 of the book, this forces
Qa = E(i)
whence
[Qa : E] = 2
Problem 27
Gal(E/F ) = G/H
and as the quotient of any cyclic group remains cyclic, we see that E/F
is a cyclic extension as desired.
Problem 28
As suggested by the hint, we split the proof into the cases where E
is a separable extension and E is purely inseparable. Suppose first that
E is a separable extension.
Let f (x) ∈ k[X] monic and irreducible. By assumption, some root
of f denoted α is contained in E. Observe that by definition, f (x) is
the minimal polynomial of α over k. Let F denote the splitting field of
f.
Since f is separable, the primitive element theorem implies that F =
k(β) for some β ∈ F . Consider the minimal polynomial mβ,k (x) of β
over k. By assumption, E contains some root γ of mβ,k (x); as [k(β) :
k] = [k(γ) : k], k(β) and k(β) are isomorphic as k-vector spaces.
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 31
k⊂F ⊂E
=⇒ α = β p
n+1
Since β p ∈ k, we see that β ∈ F . Thus the Frobenius endomorphism
is surjective, hence an automoprhism of F . By definition, F is perfect
so that E is a separable extension of F .
It remains to show that every g ∈ F [x] has a root in E. By definition
m
of F , there exists m > 0 such that g(x)p ∈ k[x]. Thus there exists
n
α ∈ E with g(α)p = 0, implying g(α) = 0. We then see that E/F is
a separable extension with the property that every polynomial in F [x]
has a root in E. By the first part of the proof, E is an algebraic closure
of F . That is, E is algebraically closed as desired.
32 KELLER
Problem 29
(a). Note that since we are working over a field of characteristic p > 0,
X
Tr(β p ) = σ(β p )
σ∈G
X
= σ(β)p
σ∈G
= Tr(β)p
Whence
Tr(β p − β) = 1p − 1 = 0
=⇒ σ(θ) − θ − β ∈ F
Note that since [K : F ] = pm−1 , Tr(γ) = pm−1 γ = 0 for all γ ∈ F .
Thus
Tr(σ9θ) − θ − β) = 0
=⇒ Tr(σ(θ) − θ) = Tr(β)
Gal(F (θ)/F ) = hσ ∗ i
pm−1
We see that σ ∗ (θ) = θ + β + 1 6= 0, implying σ ∗ has order pm−1
and hence generates our Galois group.
Problem 30
f (σ k ) = σf (σ n−1 ) + f (σ)
φ
ATr / H 1 (G, A)
a / f
a = σ(β) − β ∈ (1 − σ)A
f (σ k ) = σf (σ k−1 ) + f (σ)
σ k (β) − β
ATr
H 1 (G, A) =
(1 − σ)A
Problem 31
q n −1
So that N (α) = α q−1 ; this element has order q − 1 in F × and hence
generates the entire unit group. Noting that N (0) = 0, we deduce that
N : F → K is a surjective map.
Next we look at the trace Tr; recall that this is an F -linear map
of vector spaces. Thus the image of Tr has dimension either 0 or 1.
Observe that
n−1
Tr(α) = α + αq + · · · + αq
Problem 32
P
(a). Suppose that wi = jcij vj . We then have that
X
σk (wi ) = cij σk (vj )
j
(b). Consider
σ(DE/k (W ))
for any σ ∈ Embk (E, k a ). This has the action of interchanging the
columns of the matrix σi (wj ) i,j , implying
σ σi (wj ) i,j = ± σi (wj ) i,j
n
Since there are 2
pairs of indices (i, j) we have a factor of (−1)n(n−1)/2
pulling out. We compute:
Y Y
(−1)n(n−1)/2 · αi − αj · αj − αi ) = (−1)n(n−1)/2 (αi − αj )
i<j i6=j
Y n Y
= (−1)n(n−1)/2 (αi − αj )
i=1 j6=i
n
Y
= (−1)n(n−1)/2 f 0 (αi )
i=1
n Y
Y n
σi f 0 (α)
=
i=1 i=1
n
Y
NkE f 0 (α)
=
i=1
(d). Observe:
X
Tr(wi wj ) = σ(wi wj )
σ∈G
X
= σ(wi )σ(wj )
σ∈G
But this is precisely the (i, j) entry of the product t AA, so that
det Tr(wi wj ) = det(t AA)
i,j
= det(A)2
= DE/k (W )
As desired.
38 KELLER
Problem 33
We see that
σ 3 (x) = ζ 3 x
=x
=x
= ζ −1 x−1
= ζ −1 τ (x)
= σ −1 τ (x)
= x−3 + x3
=y
= ζ 3 x3 + ζ −3 x−3
= x3 + x−3
=y
Problem 34
Gal(K/E), Gal(K/F )
Problem 35
As contended.
Problem 36
Aut k(X) ∼
= PGL2 (k)
GL2 (k) = (q 2 − 1)(q 2 − q)
PGL2 (k) = |GL2 (k)|
|Z(GL2 (k))|
(q 2 − 1)(q 2 − q)
=
q−1
3
=q −q
As contended.
2 2
(X −q − X −1 )q+1 X q+1 (1 − X q −1 )q+1
2 =
(X −q + X −1 )q +1 X q2 +1 (1 − X q−1 )q2 +1
2
(X − X q )q+1
=
(X − X q2 )q+1
=Y
42 KELLER
We also see:
(X q2 − X)q+1 2 q+1
(X + b)q − (X + b)
τb =
(X q − X)q2 +1
q2 +1
(X + b)q − (X + b)
2 q+1
Xq + b − X − b
= q2 +1
Xq + b − X − b
=Y
(X q2 − X)q+1 2 q+1
(aX)q − aX
σa =
(X q − X)q2 +1
q2 +1
(aX)q − (aX)
2
aq+1 (X q − X)q+1
= q2 +1 q
a (X − X)q2 +1
a2
= 2Y = Y
a
Whence Y is fixed by all generators of Aut(k(X)). As [k(X) : k(Y )] =
q 3 + q and Y is fixed, we know that k(Y ) is the fixed field.
=Y
q2 −1
σa (T ) = (aX)q − aX
2 −1 2 −1
= aq (X q − X)q
=T
Thus T is fixed by H1 and
[k(X) : k(T )] = q 2 − q
SERGE LANG’S ALGEBRA CHAPTER 6 SOLUTIONS 43
= Xq + b − X − b
=Z
Thus precisely as in parts (a) and (b), we see that H2 has fixed field
k(Z), so we are done.
Problem 37
Problem 38
Let ζp denote a primitive pth root of unity, and a1/p the real number
with a = (a1/p )p . We have that Q(a1/p , ζp ) is the splitting field of the
polynomial X p − a, and this is a degree p(p − 1) extension.
If km = p|m kp , we know that kp = Q(a1/p , ζp ) and for coprime
Q
integers m and n,
= nφ(n)
= dn
So that d2n = dn .
Problem 39