The Galois Group
Look at the field \mathbb{C}. It has a hidden symmetry you have used for
years without naming it: complex conjugation,
a + bi \mapsto a - bi. It scrambles \mathbb{C}, yet
it leaves every real number untouched and — this is the magic — it respects both arithmetic operations:
\overline{z + w} = \bar z + \bar w and
\overline{zw} = \bar z\, \bar w. A shuffle of a field that preserves its
addition and multiplication and fixes a chosen subfield is exactly what
Galois built his entire theory on. Collect all such
shuffles and you get a group — the Galois group — and it turns out to encode, in the
language of symmetry, everything about how the roots of a polynomial relate to one another.
This is the idea that ties the whole module together. An algebraic question — can this polynomial be
solved by radicals? is this length constructible? — is translated into a question about a finite
group, where it becomes answerable. This page defines that group and shows it acting.
Automorphisms that fix the base
Fix a field extension E/F. An automorphism of
E is a bijection \sigma : E \to E that is a field
homomorphism — it
respects the arithmetic:
\sigma(a + b) = \sigma(a) + \sigma(b), \qquad \sigma(ab) = \sigma(a)\,\sigma(b).
We want the automorphisms that leave the base field alone. Say \sigma
fixes F if \sigma(a) = a for
every a \in F. The collection of all automorphisms of
E that fix F is the Galois group,
written \mathrm{Gal}(E/F), with composition of maps as the group operation.
-
\mathrm{Gal}(E/F) is the set of field automorphisms
\sigma : E \to E with \sigma(a) = a for all
a \in F, under composition.
-
It is genuinely a group:
the identity map fixes F, composites of such maps do too, and the inverse
of an automorphism is one.
-
When E/F is a Galois extension (normal and separable),
the group's order equals the degree:
\lvert \mathrm{Gal}(E/F) \rvert = [E : F].
The central insight: automorphisms permute roots
Here is why the Galois group is so powerful. Let \alpha \in E be a root of a
polynomial f with coefficients in F, say
f(\alpha) = \alpha^2 - 2 = 0. Apply
\sigma \in \mathrm{Gal}(E/F). Because \sigma
respects arithmetic and fixes the coefficients of f (they live in
F),
f(\sigma(\alpha)) = \sigma\big(f(\alpha)\big) = \sigma(0) = 0.
So \sigma(\alpha) is also a root of f! An
automorphism can only send a root to another root of the same polynomial. Since
\sigma is a bijection, it permutes the finite set of roots.
This gives a
group action
of \mathrm{Gal}(E/F) on the roots, and — because an automorphism of
F(\text{roots}) is pinned down by where it sends the roots — the action is
faithful, embedding the Galois group into a symmetric group
S_n. A subtle field extension becomes a concrete group of permutations you
can list.
Three worked Galois groups
Complex over real,
\mathrm{Gal}(\mathbb{C}/\mathbb{R}). An automorphism of
\mathbb{C} fixing \mathbb{R} must send
i (a root of x^2 + 1) to another root, so
i \mapsto i or i \mapsto -i. Two automorphisms: the
identity and conjugation. Hence
\mathrm{Gal}(\mathbb{C}/\mathbb{R}) = \{\,\mathrm{id}, \text{conj}\,\} \cong \mathbb{Z}/2\mathbb{Z},
matching [\mathbb{C} : \mathbb{R}] = 2.
A single square root,
\mathrm{Gal}(\mathbb{Q}(\sqrt 2)/\mathbb{Q}). An automorphism sends
\sqrt 2 to a root of x^2 - 2, i.e.
\pm\sqrt 2. So there are two: the identity, and
\sigma : \sqrt 2 \mapsto -\sqrt 2 (which sends
a + b\sqrt 2 \mapsto a - b\sqrt 2). The group has order
2, again equal to the degree.
Two independent square roots,
\mathrm{Gal}(\mathbb{Q}(\sqrt 2, \sqrt 3)/\mathbb{Q}). Now
\sqrt 2 \mapsto \pm\sqrt 2 and independently
\sqrt 3 \mapsto \pm\sqrt 3. Four combinations, each a genuine automorphism —
so the group has order 4, matching
[\mathbb{Q}(\sqrt 2, \sqrt 3) : \mathbb{Q}] = 4. Every non-identity element
squares to the identity (flipping a sign twice restores it), so this is the
Klein four-group:
\mathrm{Gal}(\mathbb{Q}(\sqrt 2, \sqrt 3)/\mathbb{Q}) \cong \mathbb{Z}/2 \times \mathbb{Z}/2 = V_4.
Three extensions, three groups — \mathbb{Z}/2,
\mathbb{Z}/2, and V_4 — each read off simply by
asking "where can the roots go?"
The commonest misconception is to imagine an automorphism can send \sqrt 2
anywhere it likes — to \sqrt 3, to 5, to some random
number. It cannot. Two iron constraints bind it. First, \sigma fixes
\mathbb{Q}, and second, it respects the relation
(\sqrt 2)^2 = 2, so
\sigma(\sqrt 2)^2 = \sigma(2) = 2 — meaning
\sigma(\sqrt 2) must itself square to 2. The only
such numbers are \pm\sqrt 2. An automorphism may only move a root to a
conjugate root of its own minimal polynomial, nowhere else.
A second, deeper trap: for a non-normal extension the Galois group can be smaller than
the degree. Take \mathbb{Q}(\sqrt[3]{2})/\mathbb{Q}, of degree
3. Its only real cube root of 2 is
\sqrt[3]{2} itself; the other two roots of x^3 - 2
are complex and do not lie in the field. An automorphism must send
\sqrt[3]{2} to a root inside the field — but there is only one — so
\sigma is forced to be the identity. Thus
\mathrm{Gal}(\mathbb{Q}(\sqrt[3]{2})/\mathbb{Q}) is trivial,
and here \lvert \mathrm{Gal} \rvert = 1 < 3 = [E : F]. Equality of order and
degree is a privilege of Galois (normal, separable) extensions only.
Because groups are finite and computable where fields are unwieldy. The crown jewel — the
Fundamental Theorem of Galois Theory — says the subfields sitting between
F and E correspond exactly, and order-reversingly,
to the subgroups of \mathrm{Gal}(E/F). A tangled lattice of fields becomes a
tidy lattice of subgroups you can draw on a page. This dictionary is what let Galois prove that the
general quintic has no formula in radicals: the relevant group is
S_5, which is not "solvable," and solvability of the group is precisely
solvability by radicals. A teenager's idea, still running the whole show.