MATH 416, Modern Algebra II
Volodymyr Nekrashevych
2020, April 23
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 1 / 17
Galois Theory
Solving equations of degree n ≤ 4
We all know 2
√ how to solve quadratic equations: the roots of x + px + q
−p± p 2 −4q
are 2 . One of the ways to deduce it is by looking at
(x − x1 )(x − x2 ) = x 2 − (x1 + x2 )x + x1 x2
and noting that (x1 − x2 )2 is a symmetric polynomial, so can be expressed
as a function of s1 = x1 + x2 and s2 = x1 x2 , namely
(x1 − x2 )2 = x12 − 2x1 x2 + x22 = (x1 + x2 )2 − 4x1 x2 = p 2 − 4q.
1 −x2 )
Then x1 − x2 = ± p 2 − 4q, and then x1,2 = (x1 +x2 )±(x
p
2 .
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 2 / 17
Galois Theory
Cubic equations
Let us try to do something similar for cubic equations. First of all, we can
simplify x 3 + ax 2 + bx + c by substitution x = y − 3a :
(y − a/3)3 + a(y − a/3)2 + b(y − a/3) + c has coefficient at y 2 equal to
−3 3a + a = 0, so we can consider polynomials of the form x 3 + px + q. If
x1 , x2 , x3 are its roots, then we have
x1 + x2 + x3 = 0
x1 x2 + x1 x3 + x2 x3 = p
x1 x2 x3 = −q
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 3 / 17
Galois Theory
x1 + x2 + x3 = 0
x1 x2 + x1 x3 + x2 x3 = p
x1 x2 x3 = −q
It follows that 0 = (x1 + x2 + x3 )3 = x13 + x23 + x33 + 3x1 x2 (x1 + x2 ) +
3x1 x3 (x1 + x3 ) + 3x2 x3 (x2 + x3 ) + 6x1 x2 x3 = x13 + x23 + x33 − 3x1 x2 x3 , so
that
x13 + x23 + x33 = −3q.
We have
(x1 x2 + x1 x3 + x2 x3 )3 = x13 x23 + x13 x33 + x23 x33 + 3 xi xj2 xk3 + 6x12 x22 x32 =
P
x13 x23 + x13 x33 + x23 x33 − 3x1 x2 x3 (x1 x2 (x1 + x2 ) + x1 x3 (x1 + x3 ) + x2 x3 (x2 +
x3 ) + 6x12 x22 x32 = x13 x23 + x13 x33 + x23 x33 − 3x12 x22 x32 , so
x13 x23 + x13 x33 + x23 x33 = p 3 + 3q 2 .
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 4 / 17
Galois Theory
Cubic equations
x13 + x23 + x33 = −3q, x13 x23 + x13 x33 + x23 x33 = p 3 + 3q 2 .
Let us look again at the discriminant
(x1 − x2 )2 (x2 − x3 )2 (x1 − x3 )2 = (x32 − 4x1 x2 )(x12 − 4x2 x3 )(x22 − 4x1 x3 ) =
−63x12 x22 x32 − 4(x23 x33 + x13 x23 + x13 x33 ) + 16x1 x2 x3 (x13 + x23 + x33 ) =
−63q 2 − 4(p 3 + 3q 2 ) + 48q 2 = −27q 2 − 4p 3 .
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 5 / 17
Galois Theory
Cubic equations
Therefore,
√ p
D = (x1 − x2 )(x1 − x3 )(x2 − x3 ) = −27q 2 − 4p 3 .
This expression is invariant under A3 ∼ = Z3 . Recall that
Q(p, q) ⊂ Q(x1 , x2 , x3 ) has S3 as the Galois group. A3 corresponds to an
intermediate field Q(p, q) ⊂ F ⊂ Q(x1 , x2 , x3 ). We have √
[F : Q(p, q)] = [S3 : A3 ] = 2, therefore F = Q(p, q)( D). We also have
[Q(x1 , x2 , x3 ) : F ] = |A3 | = 3. If u ∈ Q(x1 , x2 , x3 ) does not belong to F ,
then its irreducible polynomial over F has degree 3, so that
Q(x1 , x2 , x3 ) = F (u). We can simplify formulas by taking more than one
element to generate Q(x1 , x2 , x3 )
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 6 / 17
Galois Theory
Cubic equations
√
Let ζ = − 12 + i 23 . Then u = (x1 + ζx2 + ζ 2 x3 )/3 and
v = (x1 + ζ 2 x2 + ζx3 )/3 are multiplied by ζ and ζ 2 if we permute
x1 7→ x2 7→ x3 . Consequently,√u 3 and v 3 are invariant under A3 , hence
they belong to F = Q(p, q)( D). But u, v ∈ / F . Note that the
permutation x2 ↔ x3 interchanges u and v . The system
x1 + ζx2 + ζ 2 x3 = 3u
x1 + ζ 2 x2 + ζx3 = 3v
x1 + x2 + x3 = 0
has unique solution:
x1 = u + v , x2 = ζ 2 u + ζv , x3 = ζu + ζ 2 v
(use 1 + ζ + ζ 2 = 0).
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 7 / 17
Galois Theory
Cubic equations
In fact, a direct check shows that u 3 and v 3 satisfy the equation
p 3
y 2 + qy − = 0,
3
q
q 2 p 3
hence they are equal to − q2 ± 2 + 3 , which gives the formulas
s r s r
3 q q 2 p 3 3 q q 2 p 3
x1 = − + + + − − + ,
2 2 3 2 2 3
s r s r
3 q q 2 p 3 3 q q 2 p 3
x2 = ζ2 − + + +ζ − − + ,
2 2 3 2 2 3
s r s r
3 q q 2 p 3 2 3 q q 2 p 3
x3 = ζ − + + +ζ − − +
2 2 3 2 2 3
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 8 / 17
Galois Theory
Cubic equation: overview
The goal was to understand the splitting field Q(x1 , x2 , x3 ) of the
polynomial x 3 + px + q over Q(p, q). The Galois group is the symmetric
group S3 permuting the roots x1 , x2 , x3 . We have a chain of subgroups
{1} < A3 < S3 . Therefore, we will have a chain of subfields
Q(p, q) ⊂ F ⊂ Q(x1 , x2 , x3 ). Since the indices are [S3 : A3 ] = 2,
[A3 : {1}] = 3, the degrees are [F : Q(p, q)] = 2 and
[Q(x1 , x2 , x3 ) : F ] = 3. We check that u 3 = (x1 + ζx2 + ζ 2 x3 )3 is fixed
under A3 , but u is not. It follows that u 3 ∈ F but u is not in F , so
Q(x1 , x2 , x3 ) = F (u). We also check that D = (x1 − x2 )(x2 − x3 )(x3 − x1 )
is fixed by A3 and not by S3 , hence D ∈ F but not in Q(p, q). We also see
that D 2 is fixed by S3 , so D 2 ∈ Q(p, q). It follows that D is a square root
of a function in p and q. u 3 ∈ F , so u 3 can be expressed using D and p, q.
Consequently, u is a cube root of an expression involving p, q, D. Since
x1 , x2 , x3 ∈ F (u), all roots can be expressed using p, q, D, u.
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 9 / 17
Galois Theory
We see that the main idea was to find a tower of fields
Q(p, q) ⊂ F ⊂ Q(x1 , x2 , x3 ) such that each extension F1 ⊂ F2 in the tower
can be written as F2 = F1 (α) for some α such that αn ∈ F1 for some n,
i.e., α is a root of x n − a for some a ∈ F1 . Such extensions are called
radical. An equation can be solved in radicals if its splitting field can be
constructed using a tower of consecutive radical extensions.
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 10 / 17
Galois Theory
Degree 4 equations
A degree 4 equation can be also reduced to x 4 + px 2 + qx + r = 0 by a
change of variable x 7→ y − a/4. We can look at x 4 + px 2 + qx + r as at a
polynomial over Q(p, q, r ). Let x1 , x2 , x3 , x4 be its roots, so that the
splitting field is Q(x1 , x2 , x3 , x4 ). The Galois group of the polynomial is S4 .
We have a composition series
{1} ≤ Z2 ≤ Z2 ⊕ Z2 ≤ A4 ≤ S4
with factors Z2 , Z2 , Z3 , Z2 . This will correspond to a tower of field
extensions with degrees 2, 3, 2, 2.
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 11 / 17
Galois Theory
Degree 4 equations
The Klein’s four-group V ∼ = Z2 ⊕ Z2 plays an important role here. Recall
that it consists of the permutations
(x1 , x2 )(x3 , x4 ), (x1 , x3 )(x2 , x4 ), (x1 , x4 )(x3 , x2 ).
Let F be the corresponding fixed field. It is easy to see that the expressions
z1 = 12 (x1 x2 + x3 x4 ), z2 = 21 (x1 x3 + x2 x4 ) and z3 = 21 (x1 x4 + x2 x3 ) are
fixed by V , i.e., belong to F . The symmetric group S4 permutes z1 , z2 , z3 ,
and elements of V are the only elements fixing each zi . (BTW, this
explicitly gives an epimorphism S4 −→ S3 with kernel V .)
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 12 / 17
Galois Theory
Degree 4 equations
Since S4 permutes z1 , z2 , z3 , the elements
z1 + z2 + z3 , z1 z2 + z1 z3 + z2 z3 , z1 z2 z3 are fixed by S4 , hence belong to
Q(p, q, r ). It follows that z1 , z2 , z3 are roots of a cubic polynomial with
coefficients in Q(p, q, r ). In fact, they are roots of the cubic resolvent
q2
3 p 2 pr
z − z − rz + − = 0.
2 2 8
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 13 / 17
Galois Theory
Degree 4 equations
We know how to solve it, so we will get expressions for z1 , z2 , z3 . We have
2z1 = x1 x2 + x3 x4 and r = x1 x2 · x3 x4 . It follows that x1 x2 and x3 x4 are
roots of the polynomial x 2 − 2z1 x + r . We also have,
2(z2 + z3 ) = x1 x3 + x2 x4 + x1 x4 + x2 x3 = (x1 + x2 )(x3 + x4 ) and
(x1 + x2 ) + (x3 + x4 ) = 0. Consequently, (x1 + x2 ) and (x3 + x4 ) are roots
of x 2 + 2(z2 + z3 ). Solving these quadratic equations, we will find
x1 x2 , x1 + x2 , x3 x4 , x3 + x4 . Then, solving the quadratic equations
x 2 − (x1 + x2 )x + x1 x2 = 0 and x 2 − (x3 + x4 )x + x3 x4 = 0 we will find
x1 , x2 , x3 , x4 .
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 14 / 17
Galois Theory
General discussion
An extension F ⊂ E is an extension of F by radicals if there is a sequences
of extensions
F = F0 ⊂ F 1 ⊂ F2 ⊂ . . . ⊂ Fm = E
such that for each Fi ⊂ Fi+1 there exist αi and ni such that Fi+1 = Fi (αi )
and αini ∈ Fi .
We say that a polynomial f (x) ∈ F [x] is solvable by radicals if its splitting
field is contained in a radical extension of F . Solving a general equation of
degree n in radicals corresponds to solvability of a general polynomial
x n + an−1 x n−1 + · · · + a1 x + a0 ∈ Q(a1 , a2 , . . . , an−1 )[x] in radicals. We
have seen that general polynomials are solvable in radicals for n = 1, 2, 3, 4.
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 15 / 17
Galois Theory
Some particular (non-general) equations of higher degree may be solvable
in radicals. For example, x n − 1 or x n − a are solvable for every n and a.
Recall that a group G is called solvable if there exists a series
{1} = G0 C G1 C G2 C · · · C Gn = G
such that all factor groups Gi+1 /Gi are abelian.
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 16 / 17
Galois Theory
Theorem 1
A polynomial f (x) ∈ F [x] is solvable in radicals (if and) only if its Galois
group is solvable.
As a corollary, we get
Theorem 2
The general polynomial equation of degree n is solvable in radicals if and
only if n ≥ 4.
Namely, for n ≥ 5 the subgroup An < Sn is simple. (It is easier to show
that the subgroup of An generated by the commutators g −1 h−1 gh is the
whole group An , so that any homomorphism to an abelian group from An
has An as the kernel, so there are not subgroups H C An such that An /H
is abelian.)
V. Nekrashevych (Texas A&M) MATH 416, Modern Algebra II 2020, April 23 17 / 17