Canonical: https://knowlpedia.com/algebra-fields-galois/separable-normal-galois/

# Finite Galois extensions are separable and normal

A finite extension is Galois iff it is both separable and normal.

Let \(K\subseteq L\) be a finite [field extension](https://knowlpedia.com/algebra-fields-galois/field-extension/). Then:

**Theorem.** The following are equivalent:
1. \(L/K\) is a [Galois extension](https://knowlpedia.com/algebra-fields-galois/galois-extension/).
2. \(L/K\) is both [separable](https://knowlpedia.com/algebra-fields-galois/separable-extension/) and [normal](https://knowlpedia.com/algebra-fields-galois/normal-extension/).

Under these conditions, the extension degree equals the order of the [Galois group](https://knowlpedia.com/algebra-fields-galois/galois-group/):
\[
[L:K] = |\mathrm{Gal}(L/K)|
\]
(cf. [degree equals group order](https://knowlpedia.com/algebra-fields-galois/galois-degree-equals-group-order/)).

## Remarks

Over a [perfect field](https://knowlpedia.com/algebra-fields-galois/perfect-field/), separability is automatic for finite extensions (see [perfect implies separable](https://knowlpedia.com/algebra-fields-galois/finite-extension-perfect-separable/)), so “finite Galois” is equivalent to “finite normal.”

### Examples

1. **Quadratic extensions over \(\mathbb{Q}\).**
   \(L=\mathbb{Q}(\sqrt2)\) is the splitting field of \(x^2-2\), hence normal (see [normality via splitting fields](https://knowlpedia.com/algebra-fields-galois/normality-splitting-field/)); characteristic \(0\) gives separability. Thus \(L/\mathbb{Q}\) is Galois with \(|\mathrm{Gal}(L/\mathbb{Q})|=2\).

2. **A separable but non-normal extension.**
   \(L=\mathbb{Q}(\sqrt[3]{2})\) is separable (char \(0\)) but not normal, so it is not Galois. Its normal closure is the splitting field \(\mathbb{Q}(\sqrt[3]{2},\zeta_3)\).

3. **A normal but inseparable extension (characteristic \(p\)).**
   Let \(K=\mathbb{F}_p(t)\) and \(L=K(t^{1/p})\). Then \(L/K\) is inseparable (its defining polynomial has zero derivative), hence not Galois even though purely inseparable extensions satisfy a strong form of “no new embeddings.” This illustrates why the separability hypothesis is essential.
