Math 414 Answers for Homework 10

1. For each of the following quartic polynomials f compute Gal(L/Q), where L is

the splitting field of f . The resultant cubic p(t) and discriminant ∆(f ) for each f are
included in the table. All of the polynomials are irreducible over Q, a fact you may
assume without having to prove.

f (x) p(t) ∆(f )

(a) x4 − 2x3 + 2x2 + 2 t3 − 2t2 − 8t + 8 3136

(b) x4 + x3 + 2x2 + 2x + 1 t3 − 2t2 − 2t + 3 117
(c) x4 + 2x3 + 2x2 − 2x + 1 t3 − 2t2 − 8t 2304
4 3 2
(d) x +x +x +x+1 t − t2 − 3t + 2
3
125
(e) x4 + 2x3 + x2 − 3x + 1 t3 − t2 − 10t − 9 257
Solution. The algorithm for determining the Galois group of an irreducible quartic
f (x) = x4 + bx3 + cx2 + dx + e ∈ K[x] is explained in the chart below.
p p
p(t) δ(f ) (b2 − 4c + 4β)∆(f ) and (β 2 − 4e)∆(f ) Group
irred. over K 6∈ K S4
irred. over K ∈K A4
splits completely in K (∈ K) V
one root β ∈ K (6∈ K) one or both 6∈ K D4
one root β ∈ K (6∈ K) both ∈ K C4
Here p(t) = t3 − ct2 + (bd − 4e)t + (4ce − d2 − b2 e) is the resolvent cubic, and the
discriminant of f (x) may be computed by computing the discriminant of p(t). Applying
the algorithm we see the following.

(a) The Galois group is A4 . We have ∆(f ) = 26 · 72 is a square, so δ(f ) = 23 · 7 ∈ Q.

On the other hand, p(t) = t3 − 2t2 − 8t + 8 is irreducible over Q. One way to see
that p(t) is irreducible is to note that p(t) ≡ t3 + t2 + t + 2 (mod 3), and that
t3 + t2 + t + 2 has no root in F3 . Therefore, by the chart, the Galois group is A4 .

(b) The Galois group is D4 . The resolvent cubic factors as t3 − 2t2 − 2t + 3 = (t −

1) · (t2 − t − 3), and the quadratic factor is irreducible since its discriminant
(−1)2 − 4 · (−3)(1) = 13 is not a square. Therefore the Galois group must be
either V or D4 . The root of p(t) in Q is β = 1, and ∆(f ) = 117 = 32 · 13. Since
(b2 − 4c + 4β) = (12 − 4 · 2 + 4 · 1) = −3, and (b2 − 4c + 4β)∆(f ) = −3 · 117 = −351
is not a square in Q, we see that the Galois group must be D4 . (It’s also true that
(β 2 − 4e)∆(f ) = (12 − 4 · 1) · 117, which is again −351, is not a square in Q.)

1
 (c) The Galois group is V . The resolvent cubic factors completely over Q as t3 −
2t2 − 8t = t(t + 2)(t − 4), and V is the only possibility where the cubic factors
completely.

(d) The Galois group is C4 . There are two ways to see this. The polynomial f (x) =
x4 + x3 + x2 + x + 1 is the minimal polynomial of the primitive 5-th root of unity
ζ = e2πi/5 (for example by H3 Q1(a)). The Galois group after adding any p-th
root of unity to Q is the same as the multiplicative group F∗p , which we know is
cyclic of order p − 1. (We did a computation like this for p = 7 in H7 Q2.) Hence
the Galois group is cyclic of order 5 − 1 = 4.

Alternatively, we can use algorithm. The resolvent cubic factors as p(t) = t3 −

t2 − 3t + 2 = (t − 2)(t2 + t − 1). The quadratic factor is irreducible over Q since its
discriminant is 12 −4(1)(−1) = 5, which is not a square in Q. Therefore the Galois
group must be C4 or D4 . The root of p(t) in Q is β = 2, and ∆(f ) = 125 = 53 .
Since (b2 − 4c + 4β) · ∆(f ) = (12 − 4(1) + 4(2)) · 125 = 5 · 125 = 54 is a square in
Q, and since (β 2 − 4e)∆(f ) = (22 − 4(1))∆(f ) = 0 is a square in Q, we see that
the Galois group must be C4 .

(e) The Galois group is S4 . We have ∆(f ) = 257 is prime, so ∆(f ) is not a square
in Q. The resolvent cubic p(t) = t3 − t2 − 10t − 9 and is irreducible over Q. One
way to see that p(t) is irreducible over Q is to note that p(t) ≡ t3 + t2 + 1 (mod
2), and that t3 + t2 + 1 has no root over F2 . Therefore (by the chart), the Galois
group is S4 .

2. In class we skipped over almost all of the details of the algorithm for detecting the
difference between C4 and D4 when computing the Galois group of an irreducible quartic
polynomial. In this problem we will check some of the claims for the polynomial g1 (t).
Let K be a field of characteristic zero, and let f = x4 + bx3 + cx2 + dx + e ∈ K[x] be
an irreducible quartic with splitting field L. We suppose that the resultant cubic p(t)
has a single root β ∈ K, β = γ13|24 . Recall that this means that the Galois group G is
contained in h(1 2 3 4), (1 3)i = D4 , and is either D4 or C4 = h(1 2 3 4)i. We set

g1 (t) = (t−(α1 +α3 ))(t−(α2 +α4 )) = t2 −(α1 +α2 +α3 +α4 )t+(α1 +α3 )(α2 +α4 ) = t2 +bt+(c−β).

The importance of the last equality is that it shows that g1 (t) ∈ K[t].

(a) Explain what σ = (1 2 3 4) does to each of the roots α1 , α2 , α3 , and α4 . (This

question is to asking if you understand the isomorphism Perm(S) ∼ = S4 we have
been using.)
(b) Show that {α1 + α3 , α2 + α4 } is a single orbit under C4 and D4 .

2
 (c) There is nothing to say that α1 + α3 and α2 + α4 couldn’t be equal. (The set in (b)
could consist of a single element, e.g. {z, z} = {z}.) Show that α1 + α3 = α2 + α4
if and only if α1 + α3 ∈ K. (Hint : Your calculation in (b) is relevant.)
(d) Conclude that either g1 (t) is irreducible over K or that g1 (t) has a double root.
(e) Explain why either δ(g1 ) 6∈ K or δ(g1 ) = 0.

We now want to show that if G = C4 then δ(g1 )δ(f ) ∈ K. This is clear when δ(g1 ) = 0,
so for the rest of the problem we assume that δ(g1 ) 6= 0. We also assume that G = C4 .
(f) Explain why δ(f ) 6∈ K. (Reminders : What does δ(f ) detect? What is G?)
(g) Explain why there is only one intermediate field M ⊂ L of degree 2 over K.
(h) Explain why K(δ(g1 )) and K(δ(f )) are degree 2 extensions of K.
(i) By (g) and (h) there are a0 , a1 ∈ K such that δ(g1 ) = a0 +a1 δ(f ). Since δ(g1 ) 6∈ K,
δ(g1 ) is not fixed by G, and hence there is some τ ∈ G so that τ · δ(g1 ) = −δ(g1 ).
Applying τ to the equation above, explain why this means that a0 = 0.
(j) Explain why δ(g1 )δ(f ) ∈ K.
Remarks. () The same argument, with α1 α3 and α2 α4 replacing α1 + α3 and α2 + α4 ,
shows that if G = C4 then δ(g2 )δ(f ) ∈ K. () A separate (shorter) computation shows
that both of these cannot happen if G = D4 , which leads to the criterion for the test.
Solution.
(a) The isomorphism of Perm(S) and S4 is obtained by matching the root αi with the
integer i. Since σ sends 1 to 2, 2 to 3, 3 to 4, and 4 to 1, we have σ(α1 ) = α2 ,
σ(α2 ) = α3 , σ(α3 ) = α4 , and σ(α4 ) = α1 .

(b) From the formulae above we have σ(α1 + α3 ) = σ(α1 ) + σ(α3 ) = α2 + α4 , and
σ(α2 + α4 ) = σ(α2 ) + σ(α4 ) = α1 + α3 . Since σ swaps α1 + α3 and α2 + α4 , the
set {α1 + α3 , α2 + α4 } is a single orbit of C4 = hσi. The group D4 , which contains
C4 , might have a larger orbit since other elements of D4 might send α1 + α2 and
α2 + α4 to different elements of L. However τ = (1 3) fixes each of α1 + α3 and
α2 + α4 (i.e, τ (α1 + α3 ) = α1 + α3 , and τ (α2 + α4 ) = α2 + α4 ). Therefore the orbit
of D4 = hσ, τ i is again the set {α1 + α3 , α2 + α4 }.

(c) If α1 + α3 is in K then it is fixed by all elements of the Galois group. In particular

σ(α1 + α3 ) = α1 + α3 . Above we have calculated that σ(α1 + α3 ) = α2 + α4 , and
combining these we conclude that α1 +α3 = α2 +α4 . Conversely, if α1 +α3 = α2 +α4
then by (b) the set {α1 + α3 , α2 + α4 } = {α1 + α3 } is fixed by G, and so α1 + α3
is in LG = K.

3
 (d) If g1 (t) is reducible then both of its roots, namely α1 + α3 and α2 + α4 are in K.
Then by (c) α1 + α3 = α2 + α4 , so that g1 (t) has a double root. Conversely, if g1 (t)
has a double root then α1 + α3 = α2 + α4 so again by (c) both are in K and g1 (t)
factors over K.

(e) The quadratic formula tells us that a degree two polynomial g(t) ∈ K[t] factors
over K if and only if its discriminant ∆(g) is a square in K. By (d) we either have
g1 (t) is irreducible (and so δ(g1 ) 6∈ K) or g1 (t) has a double root (so δ(g1 ) = 0).

(f) First note that δ(f ) 6= 0 : since f is an irreducible polynomial over a field of
characteristic zero it has no repeated roots. The action of the Galois group on
δ(f ) detects whether or not G is contained in the alternating group. The element
σ = (1 2 3 4) ∈ C4 is not in A4 and so σ(δ(f )) = −δ(f ), showing that δ(f ) 6∈ K.

(g) By the Galois correspondence, an intermediate field K ⊂ M ⊂ L of degree 2 over

K corresponds to a subgroup H of index 2 in G = C4 . Since C4 is a cyclic group,
it has only one subgroup of degree e for any e dividing 4. In particular, it has only
one subgroup of index 2, and so there is only one intermediate field of degree 2
over K.

(h) We know that ∆(f ) = δ(f )2 and ∆(g1 ) = δ(g1 )2 are in K. Therefore the generators
of K(δ(g1 )) and K(δ(f )) satisfy degree 2 equations over K. The generators δ(g1 )
and δ(f ) are also not in K themselves by our assumption and (f). Therefore the
extensions are of degree 2 over K.

(i) Applying τ to δ(g1 ) = a0 + a1 δ(f ) we get

−δ(g1 ) = τ (δ(g1 )) = τ (a0 + a1 δ(f )) = a0 + a1 τ (δ(f )) = a0 ± a1 δ(f ).
(At the moment we don’t’ know whether τ (δ(f )) = δ(f ) or −δ(f ).) But we also
know that −δ(g1 ) = −(a0 + a1 δ(f )) = −a0 − a1 δ(f ), and comparing these gives
−a0 − a1 δ(f ) = a0 ± a1 δ(f ). Since 1 and δ(f ) are a basis for K(δ(f )) over K, this
means that we have −a0 = a0 (and so a0 = 0) and τ (δ(f )) = −δ(f ) (which we
don’t need at the moment).

(j) Since a0 = 0 this means that δ(g1 ) = a1 δ(f ) with a1 ∈ K, and so δ(g1 )δ(f ) =
(a1 δ(f ))δ(f ) = a1 δ(f )2 = a1 ∆(f ) ∈ K.

3. Let L/K be a Galois extension with Galois group G, and β an element of L. Let
S = OrbQGs(β) be the orbit of β under G, say S = {β = β1 , β2 , . . . , βs }, and finally set
q(x) = j=1 (x − βj ).
In this problem we will show that q(x) is the minimal polynomial of β over K.

4
 (a) Explain why all the coefficients of q(x) are in K, so that q(x) ∈ K[x]. (Sugges-
tion : what does acting by G do to the elements of S?)

It is clear that q(x) is a monic polynomial with β as a root. Therefore to show that q(x)
is the minimal polynomial of β, it is sufficient to show that q(x) is irreducible over K.
(b) Suppose that q(x) factors as q(x) = q1 (x)q2 (x), with each of q1 , q2 ∈ K[x], and
of degree at least one. By relabelling q1 and q2 if necessary, we may assume that
q1 (β) = 0. Explain why, for every σ ∈ G, σ(β) is a root of q1 (x).
(c) Show that none of the roots of q2 (x) are in the orbit of β.
(d) Explain why the result in (c) is a contradiction, and hence that q(x) must be
irreducible.

Solution.
(a) Since the set S is a single orbit of G, G acts on S by permuting its elements.
The coefficients of q(x) are, up to sign, the elementary symmetric polynomials in
β1 ,. . . , βs . Hence any permutation leaves them unchanged, in particular, action
by G leaves them unchanged. Thus the coefficients are in LG = K.

(b) We have seen this argument many times, first in the class “Automorphisms fixing
a subfield” from January 20th. If q1 (x) = xn + an−1 xn−1 + · · · + a1 x + a0 with the
ai ∈ K, then
0 = σ(0) = σ(q1 (β)) = σ(β n + an−1 β n−1 + · · · + a1 β + a0 )
= σ(β)n + an−1 σ(β)n−1 + · · · + a1 σ(β) + a0 = q1 (σ(β)),
so σ(β) is a root of q1 (x).

(c) The elements of the set S are distinct, so q(x) has no repeated roots. This means
that q1 (x) and q2 (x) have no roots in common. Part (b) shows that the orbit of β
is contained in the subset of roots of q1 (x), therefore no roots of q2 (x) are in the
orbit of β.

(d) Since deg q2 (x) > 1, q2 (x) must have at least one root. By construction the roots
of q(x) are the elements in the orbit of β, so all roots of q2 (x) are a subset of the
orbit. This contradicts (c), and shows that we could not have had the factorization
q(x) = q1 (x)q2 (x) over K with both q1 (x) and q2 (x) of degree > 1. Therefore q(x)
is irreducible over K.

Remark. We have used the argument in part (a) before : see the proof of () =⇒ ()
on February 4th (“Galois Extensions”).