Let\[(a \ b … c \ d) \text{ and } (e \ f)\] be any \(p\)-cycle and any transposition. We can rename \(e=1\) such that the permutations are now: \[(a' \ b' … 1 … c' \ d')=(1 … c' \ d' \ a' \ b' …) \text{ and } (1 \ f')\] Now, for some \(1 ≤ k ≤ p-1\): \[(1 … c' \ d' \ a' \ b' …)^k=(1 \ f' …)\] and as \(p\) is supposed to be prime, \((1 \ f' …)\) is a \(p\)-cycle. We can again rename the elements of \(\{1, …, p\}\) such that \(f' = 2\) and the rest accordingly to get: \[(1 \ 2 … p-1 \ p) \text{ and } (1 \ 2)\]According to theorem 3 they generate \(S_n\).
Counterexample when $n$ is composite: $n=4,σ=(1234),τ=(13)$. Consider the action on 4 vertices of a square where $σ$ is rotation and $τ$ is a reflection, then $D_4=⟨σ,τ⟩$ is a proper subgroup of $S_4$, for example $(12)∈S_4$ cannot be generated by $σ,τ$.
Any 5-cycle and any double transposition are even permutations so they cannot generate whole $S_5$.
For example $σ=(12345)$ and $τ=(14)(23)$. Consider the action on 5 vertices of a pentagon where $σ$ is rotation and $τ$ is a reflection, then $D_5=⟨σ,τ⟩$ is a proper subgroup of $S_5$, for example $(12)∈S_5$ cannot be generated by $σ,τ$.
(⇒) Given a finite Galois extension $K/F$. By Theorem 6.6, $K/F$ is normal.
By Corollary 6.7a, $K/F$ is separable.
(⇐) Given a finite, normal, separable extension $K/F$. Let $α_1,α_2,…,α_n$ be a $F$-basis for $K$.
Since $m_{F,α_i}$ has a root $α_i$ in $K$, by normality, $m_{F,α_i}$ splits over $K$.
Irreducible factors of $f≔\prod_{i=1}^nm_{F,α_i}$ are $m_{F,α_i}$ which are separable since $K/F$ is separable. Thus, $f$ is separable.
Since $α_1,α_2,…,α_n$ generate $K$, $K$ is the splitting field of $f$, then $K/F$ is Galois.