Math 504, Fall 2013

HW 5

1. Let G be a group and N a normal subgroup. Show that G is solvable if and only if N
and G/N are solvable.

(⇐) Let G be a group and H C G. Suppose that H and G/H are solvable. Then we can
build a solvable series for G in the following manner.
Consider a solvable series 1 C · · · Hi C · · · C H for H, and a solvable series 1 C Ki C
· · · C G/H for G/H. Let φ : G → G/H be the quotient map. Then φ defines a bijection

{subgroups of G containing H } ⇐⇒ {subgroups of G/H }

Moreover, by the Third Isomorphism Theorem we know that φ−1 (Ki+1 )/φ−1 (Ki ) ∼ = Ki+1 /Ki .
It follows that the quotients of consecutive groups in the series are unchanged under φ−1 .
Thus
1 C · · · Hi C · · · C H C · · · C φ−1 (Ki ) · · · C G
is a solvable series for G.
(⇒) Suppose G is solvable and H C G. If 1 C · · · Gi C · · · C G is a solvable series for
G, then consider the groups Hi = H ∩ Gi . We claim that 1 C · · · Hi · · · C H is a solvable
series for G. First, we need to verify that Hi C Hi+1 .
Let q : Gi+1 → Gi+1 /Gi be the projection. Then q| N : Hi+1 → Gi+1 /Gi has kernel Hi .
Thus Hi C Hi+1 . By the subgroup correspondence used in the previous part, it follows
that Hi+1 /Hi is a subgroup of Gi+1 /Gi , which is abelian. So Hi+1 /Hi is abelian, and thus
the Hi form a solvable series for H.
Now we construct a solvable series for G/H. Use the same map φ : G → G/H from
the last part. We claim that
1 C · · · φ( Gi ) · · · C φ( G )
is a solvable series for G/H (obviously since φ is surjective, φ( G ) = G/H). Let Ki =
φ( Gi ). By the subgroup correspondence, Ki C Ki+1 .
Moreover, Ki+1 /Ki is abelian. Indeed, consider the map

ρ : Gi+1 → Ki+1 /Ki

given by applying φ followed by applying the quotient map. Then ρ( Gi ) = 1, so ρ passes

to the quotient over Gi+1 /Gi . Thus there is a surjective map ρ̃ : Gi+1 /Gi → Ki+1 /Ki . Since
Gi+1 /Gi is abelian, it follows that Ki+1 /Ki is also abelian. Hence the Ki form a solvable
series for G/H.


1
2. Compute the Galois group of x6 − 2x3 − 1 over Q.

Define f ( x ) = x6 − 2x3 − 1.
Let y = x3 . It’s clear that f ( x ) = g(y) = y2 − 2y − 1, so:
√
2± 8 √
f ( x ) = 0 ⇔ g(y) = 0 ⇔ y = = 1± 2
2
p √ p √ √
Let α = 1 + 2, β = 1 − 2, ω = 1+ 2 −3 . Then clearly, f ( x ) splits as:
3 3

f ( x ) = ( x − α)( x − ωα)( x − ω 2 α)( x − β)( x − ωβ)( x − ω 2 β)

q √ √ √ √
Also, compute that αβ = 3 (1 + 2)(1 − 2) = 3 1 − 2 = 3 −1 = −1, so α = − β−1 .
Thus, Q(α, β) = Q(α), and consequently, K = Q(α, ω ) is a splitting field for f over Q.
Note that K is the smallest such field: obviously α, ω ∈ √ K because α is a root and ω is a
quotient of roots. We know α ∈ / Q, because α − 1 = 2 ∈
3 / Q, and we know ω ∈ / Q( α )
because Q(α) is a real extension. Consequently, we need exactly those two elements to
generate K over Q. √
Next, note that Q( 2) ⊂ Q(α), so
√ √ √
[Q(α) : Q] = [Q(α) : Q( 2)][Q( 2) : Q] = 2[Q(α) : Q( 2)].
√
Note that in Q( 2)[ x ]:
√ √
f ( x ) = ( x3 − 2 − 1)( x3 + 2 − 1)

I claim that those factors are irreducible.

Let a, b ∈ Z and suppose that:
√ √ √
( a + b 2)3 = a3 + 3a2 b 2 + 6ab2 + 2 2b3
√
= ( a3 + 6ab2 ) + 2(3a2 b + 2b3 )
√
= 1± 2

Then a3 + 6ab2 = a( a2 + 6b2 ) = 1. Since a, b ∈ Z and a is a unit, a = ±1. But since

a2 + 6b2 = 1 +√6b2 is also a unit, b = 0, which forces a = 1 because a(√ a2 + 6b2 ) = a√ 3 = 1.

But 1 6= 1 ± 2; thus, there do not exist a, b√∈ Z for which ( a + b 2) = 1 ± 2, or

3 3

equivalently, f ( x ) does not√ have a root in Z[ 2]. Since f is a product √ of cubics, neither√
of which has a root,
√ in Z [ 2 ] , those cubics remain irreducible over Z [ 2 ] . Since Z [ 2]
is a√UFD and Q[ 2] is its field√of fractions, the cubic factors of f are still irreducible over
Q[ 2]. As a result, [Q(α) : Q( 2)] = 3, and [Q(α) : Q] = 6. To get the full splitting field,
we need to adjoin ω, which is quadratic over any real field. Thus, [Q(α, ω ) : Q] = 12,
which means we are looking for a group of order 12. Since [Q(α) : Q] = 6, f is irreducible,
so the group will be a transitive subgroup of S6 . There are only two such groups: D12 and
Z2 × Z6 , and since Q(α)/Q is a non normal extension, G cannot be abelian, hence G is
isomorphic to the dihedral group D12 . 

2
3. Solve the equation x6 − x5 + x4 − x3 + x2 − x + 1 = 0 by radicals. Hint: Think about
t = x + x −1 .
Start by computing that:

( x + 1) f ( x ) = x 7 + 1
πi
Thus, the roots of f ( x ) are primitive 7th roots of −1, i.e. they are equal to e 7 , where
i = 1, 3, 5, 9, 11, 13.
Let t = x + x −1 , and consider x −3 f ( x ) = x3 − x2 + x − 1 + x −1 − x −2 + x −3 . Since:

t 2 = x 2 + 2 + x −2
t3 = x3 + 3x + 3x −1 + x −3

we have that:
x −3 f ( x ) = t3 − t2 − 2t + 1 := f˜(t)
Next, compute that:
 
˜f t + 1 = t3 − 7 x + 7 := g(t)
3 3 27

I will solve g(t) and use that to solve f ( x ).

To solve a depressed cubic, we begin by solving:

= − 73

3st
3 3 7
s − t = − 27

Then, plugging s − t into the cubic will produce 0, because:

(s − t)3 + 3st(s − t) − (s3 − t3 ) = s3 − 3s2 t + 3st3 − t3 + 3s2 t − 3st3 − s3 + t3 = 0

Also, if ω = e2πi/3 , then:

g(ωs − ω 2 t) = (ωs − ω 2 t)3 + 3(ωs)(ω 2 t)(ωs − ω 2 t) − ((ωs)3 − (ω 2 t)3 )

= −3(ω 2 s2 )(ω 2 t) + 3(ωs)(ω 4 t2 ) + 3s2 tω − 3ω 2 st3
=0

g(ω 2 s − ωt) = (ω 2 s − ωt)3 + 3(ω 2 s)(ωt)(ω 2 s − ωt) − ((ω 2 s)3 − (ωt)3 )

= −3(ω 4 s2 )(ωt) + 3(ω 2 s)(ω 2 t2 ) + 3s2 tω 2 − 3ωst3
=0

Thus, once we have s, t, we have all 3 roots.

Now:
−7 −7 −73
 
7 7
3st = − =⇒ s = =⇒ − = − t3 = 6 3 − t3
3 9t 27 9t 3 t

3
Rearranging:
7 73 7 3 73
t3 − + = 0 = t 6
− t + 6
33 36 t3 33 3
Solving the quadratic:
√ √
q
7 72 2 3
33
± 36
− 2376 7 ± 7 1 − 28 7 ± 21i 3
3
t = = =
2 2 · 33 54
From that, we can directly obtain s3 :
√ √
− 7 7 ± 21i 3 7 − 7 ± 21i 3
s3 = t3 − = − =
27 54 27 54
Replace the ± by a + in both s3 , t3 from now on.
By the work I’ve already done, we know the roots of g, and from those, we get that
the roots of f˜ are:
s √ s √ 
1 3 −7 + 21i 3 3 7 + 21i 3
ρ1 =  − − 1
3 2 2
 s √ s √ 
1 −7 + 21i 3
3 3 7 + 21i 3
ρ2 = ω − ω2 − 1
3 2 2
 s √ s √ 
1 3 −7 + 21i 3 3 7 + 21i 3
ρ3 =  ω 2 −ω − 1
3 2 2

Let ρ denote any one of the ρi . To obtain the roots of f , we need to solve x + x −1 = ρ
for each of the three ρ’s; this will give us the 6 roots of f . Therefore an expression for them
by radicals is
q
ρi ± ρ2i − 4
for i = 1, 2, 3
2


4. Let k be a subfield of C and let K = k (α, β) where α2 = a ∈ k and β2 = b ∈ k and

none of a, b, or ab, is a square in k. Prove that Gal (K/k ) ∼
= Z2 × Z2 .

Let k be a subfield of C and let K = k(α, β) where α2 = a ∈ k and β2 = b ∈ k and none of

a, b, or ab is a square in k. Since K is the splitting field of the polynomials x2 − a and x2 − b,
it is a normal extension. Since each of α and β are separable, K is a separable extension.
Thus, K/k is Galois.
Firstly, k (α)/k is a Galois extension of degree 2 as the splitting field of the irreducible
polynomial x2 − a over k. Suppose that β ∈ k (α), so that β = q + rα for some q, r ∈

4
k. Squaring both sides, we get that β2 = q2 + 2qrα + r2 α2 and equivalently, b = q2 +
2qrα + r2 a. We must have that either q or r is zero; otherwise, this lets us express α as
a combination of the elements a, b, q, r in k. Since β 6∈ k, we must have q = 0 and that
b = r2 a. Then ab = r2 a2 , but this contradicts our assumption that ab is not a square in k.
Hence, β 6∈ k (α).
Now we see that K is a degree 2 extension of k(α) as the splitting field of x2 − b, so
[K : k] = [K : k(α)][k(α) : k] =. Then |Gal (K/k)| = 4, so our options for the Galois
group are Z4 or Z2 × Z2 . We have two distinct intermediate subfields k (α) and k( β)
lying between k and K (we’ve already shown that β 6∈ k(α), and the same argument
works in the reverse direction). This means that there is a distinct non-trivial subgroup
of the Galois group that fixes each of these fields, i.e., there are (at least) two distinct
subgroups in the Galois group. The only option is then that Gal (K/k ) is Z2 × Z2 , as Z4
has only one non-trivial subgroup. 

5. Let f ∈ k [ x ] be an irreducible polynomial where k is a subfield of C. Suppose f has

at least one root that is expressible in radicals. Show that all roots of f can be written in
terms of radicals.
Let K be a splitting field for f over k, and let α ∈ K be a root of f that is expressible in
radicals. This means k(α) is a radical extension of k.
But, because f is irreducible, the Galois group G := Gal (K/k ) acts transitively on
the roots of f . Thus, for any root β of f , there exists an automorphism τ ∈ G such that
τ (α) = β. But then, as τ is k-linear ring automorphism, we also have
!
n n n
τ ∑ ci αi = ∑ ci (τ (α))i = ∑ ci ( β )i ∈ k ( β )
i =1 i =1 i =1

This is well-defined since α, β have the same minimal polynomial f , as f is irreducible.

Therefore, τ maps k (α) into k( β). Let σ = τ |k(α) : k(α) → k( β) be the restriction. And,
since σ is nonzero an automorphism of a field, ker(σ) = {0}, and it is injective. Further-
more, by the above, σ is clearly surjective, since all linear combinations of powers of β are
hit by σ. Thus, σ is a field isomorphism, and we conclude that
k(α) ∼
= k ( β ).
Now, as k (α) is a radical extension, we can write
k = k0 ⊂ k1 ⊂ · · · ⊂ kr = k(α)
√
for subfields k i , where k i+1 /k i is a cyclic extension, or k i+1 = k i ( ni ai ) for some ai ∈ k i ,
ni ∈ N. Then, mapping everything by the isomorphism σ, we get a radical tower
k = k00 ⊂ k01 ⊂ · · · ⊂ k0r = k ( β)
√
for subfields k0i = σ (k i ), where again k0i+1 /k0i is a cyclic extension, or k0i+1 = k0i ( ni ai )
for some ai ∈ k0i , ni ∈ N. Thus, k ( β) is a radical extension as well. It follows that β is
expressible in terms of radicals as well. Hence, all roots of f can be written in terms of
radicals.


5
6. Give an example of an extension K/k such that [K : k ] = n and a positive integer d
dividing n such that there is no intermediate field k ⊂ L ⊂ K with [ L : k ] = d.

Let K = C[ x1 , x2 , x3 , x4 ] and let k = K A4 be the ring of invariants under the alternating

group. By Artin’s theorem, [K : k ] = | A4 | = 12 and the extension is Galois with Galois
group A4 . The integer 3 divides 12, but if there were an intermediate field L with [ L : k] =
3, then the subgroup of automorphisms in Gal (K/k ) fixing L would be a subgroup of A4
of order 6, of which there are none. 

7. Let K/k be a Galois extension of degree n with Galois group G. Define the norm and
trace maps NK/k : K → k and TK/k : K → k by

N ( a) = ∏ σ( a) and T ( a) = ∑ σ ( a ).
σ∈G σ∈G

Suppose that the minimal polynomial of a ∈ K is ∑ri=0 ci xi ∈ k [ x ]. Show that

1. the norm and trace do take values in k, and

2. N ( a) = (−1)n c0n/r and T ( a) = − nr cr−1 , and

3. N ( a) = det( T ) and T ( a) = TraceT where T : K → K is the k-linear map T (b) := ab.

1. Let τ ∈ G. Then τ ( N ( a)) = ∏σ∈G τσ( a) = N ( a), since as σ runs through all of G,
τσ does. Thus N ( a) is in the fixed field of G, which is exactly k. Similarly τ ( T ( a)) =
∑σ∈G τσ ( a) = T ( a) so that T ( a) is in the fixed field of G, i.e., in k.
2. Let the roots of f be α1 , . . . , αr (one of which is equal to a). Since the minimal poly-
nomial of f has degree r, if we let H = Gal(K/k ( a)), then H has index r in G.
Let τ1 H, . . . , τr H be the left cosets of G/H, and fix τ1 , . . . , τr as representative ele-
ments, and call T = {τ1 . . . , τr }. First we notice, because τ ∈ G, that f (τ ( a)) =
τ ( f ( a)) = 0 so that τ ( a) = αi for some i. Since G acts transitively on the roots of f ,
for any i there is some τ ∈ G so that τ ( a) = αi . Finally, if τ ( a) 6= τi ( a) then τ ∈
/ τi H,
because H fixes a so every element in the coset must take a to the same point. This
means that we can reorder the roots so that τi ( a) = αi (since each αi must be reached
by some τ, and therefore by the representative of the coset that τ is in).
Notice that f = ( x − α1 ) . . . ( x − αr ) so that ∏ri=1 (αi ) = (−1)r c0 . Therefore, because
a lies in the fixed field of H:
!n/r !n/r
r
N ( a) = ∏ σ(a) = ∏ ∏ τσ(a) = ∏ τ (a) = ∏ αi = (−1)n c0n/r .
σ∈G τ ∈T σ∈ H τ ∈T i =1

Next notice that ∑ri=1 αi = −cr−1 , so that,

n n r n
T ( a) = ∑ σ( a) = ∑ ∑ τσ( a) = ∑
r τ ∈T
τ ( a ) = ∑
r i =1
αi = − cr .
r
σ∈G τ ∈T σ∈ H

6
3. To illustrate the idea behind the proof, we first assume that r = n. Then 1, a, . . . , an−1
is a basis for K/k (order it and call it b1 , . . . , bn ). Since f ( a) = 0 then we solve and
get that an = −c0 − . . . − cn−1 an−1 . So computing the matrix for T as if r = n
then if i 6= n then T (bi ) = T ( ai−1 ) = ai = bi+1 . And T (bn ) = T ( an−1 ) = an =
−c0 − . . . − cn−1 an−1 = −c0 b1 − . . . − cn−1 bn . So we have the following matrix.

0 1 0 ··· 0
 
 0 0 1 ··· 0 
 .. .. ... ... .. 
R= . . . .
 

 0 ... 
0 0 1 
− c0 − c1 − c2 · · · − c n −1

It is clear that TraceR = −cn−1 and cofactor expanding along the first column we
see that det R = (−1)r c0r .
Now we assume that r < n. We see now that 1, . . . , ar−1 is a basis for k ( a)/k. Let
β 1 , . . . , β r/n be a basis for K/k( a). Then { β i a j |1 ≤ i ≤ r/n, 0 ≤ j ≤ r − 1} is a basis
for K/k. If we order this as β 1 , β 1 a . . . , β 1 ar−1 , β 2 , . . . , β r/n ar−1 , we can compute the
matrix for T by seeing what T does to the basis elements. But this is easy, because
T ( β i a j ) = β i a j+1 where j 6= n − 1 and T ( β i an−1 ) = − β i c0 − β i ac1 − . . . − β i an−1 cn−1 .
So the matrix for T is exactly,
 
R 0 ··· 0
0 R · · · 0
T =  .. .. . . . .
 
. . . .. 
0 0 ··· R

Then clearly Det T = Det Rn/r = (−1)r c0n/r = N ( a), and TraceT = n
r TraceR =
n
r (− cn−1 ) = T ( a ).