Let $G$ be a group and $H$ a subgroup of $G$ and let $[G:H]=n$. We consider here the question of whether there is an element in $g\in G$ such that $\left\{H,gH,\ldots,g^{n-1}H\right\}$ are all the cosets of $H$ in $G$.
(a) Show that if $n$ is not a prime this may fail.
(b) Show that if $n$ is prime such $g$ always exists. (Suggestion: Show first that a transtive subgroup of $S_n$ has order divisible by $n$. Show then that a transitive subgroup of $S_p$ has an element of order $p$. Use the coset representation to finish the proof. You may use Cauchy'theorem : a finite subgroup whose order is divisible by a prime $p$ has an element of order $p$.)
In (a), I can't find a simple counterexample.
the solution is:
By considering the (transitive) action of $G$ on $G/H$, we have a homomorphism $$\psi:G\longrightarrow S_p$$ and thus $\psi(G)$ is a transitive subgroup of $S_p$. It follows that $p$ divides $\psi(G)$ and thus, by Cauchy's theorem, there is an element $\psi(g)\in\psi(G)$ of order $p$.
With the element $\psi(g)$ of order $p$, let us first verify that $g \notin H$. Suppose you had $g\in H$. Then $\psi(g^k)$ would fix $H$ for all $k$, so $\psi(g)$ would be a permutation of the $p-1$ other cosets of $H$, of order $p$. But $p \nmid (p-1)!$, so that can't be.
Hence we have $\psi(g)H \neq H$. Now consider the order of the orbit of $H$ under $\psi(g)^k$. Let that order be $n$. We saw $n > 1$. But $\psi(g)^p = \operatorname{id}$, hence $n \mid p$, so $n = p$, since $p$ is prime.
I don't see that:
(i) so $\psi(g)$ would be a permutation of the $p-1$ other cosets of $H$, of order $p$. But $p \nmid (p-1)!$, so that can't be.
My doubt: $\psi(g)$ would be a permutation of the $p-1$ other cosets of $H$, of order $p$. Then $p \mid (p-1)!$?