Proof check/help for minimal polynomial using Galois theory

29 Views Asked by At

I'm working on the following proof, and think that what I've got so far is okay, but was hoping someone could help give a hint with the remainder.

Question: let $E/F$ be a Galois extension with Galois group $G = G(E/F)$ and $\alpha \in E$. Prove that $E = F(\alpha)$ if and only if for all $g\in G$, $g \neq$ id we have $g(\alpha)\neq \alpha$, and hence prove the minimal polynomial for $\alpha$ is given by: $$ (x-g_1(\alpha))\ldots(x-g_n(\alpha)) $$ where $G = \{g_1, \ldots, g_n\}$.

Proof (so far): We first prove the $\Rightarrow$ direction. Suppose $E = F(\alpha)$, $g\in G$ with $g\neq$ id. Then note that for any $g \in G$ we have $g(x)=x$ for all $x\in F$, so $g$ must always be entirely determined by $g(\alpha)$, and hence if $g(\alpha)=\alpha$ we must have $g = $id, a contradiction. Thus we conclude $g(\alpha)\neq \alpha$ for all $g\in G$, $g\neq$ id. We now prove the $\Leftarrow$ direction. Suppose that for all $g\in G$ with $g\neq$ id we have that $g(\alpha) \neq \alpha$. Note that since $F \subseteq E$ and $\alpha \in E$, clearly $F(\alpha) \subseteq E$, and so it remains to show that $[F(\alpha):F] = [E:F]$ (or equivalently $[E:F(\alpha)]=1$), since then we would have a subspace of equal dimension, thus implying that $E = F(\alpha)$. (Here is where I am stuck for this part of the proof)

Now let $f$ be a polynomial, say of degree $n$, with $\alpha \in E$ as a root, and let $\alpha = \alpha_1, \ldots, \alpha_n \in E$ denote the roots of $f$, possibly not distinct. Now let $\alpha_1, \ldots, \alpha_k$ denote the $k$ distinct roots, and suppose $g \in G = \{g_1, \ldots, g_n\}$ with $g \neq$ id. Then since $G$ forms a group we know that there is a bijective correspondence between the sets $\{\alpha_1, \ldots, \alpha_n\} \leftrightarrow \{g(\alpha_1), \ldots, g(\alpha_n)\}$, that is $g$ permutes the roots. Thus restricting any $g \in G$ to the set of distinct roots $\{\alpha_1, \ldots, \alpha_k\}$ will also be a permutation, since $g$ is injective (it is an automorphism) so its restriction to a finite set is a bijection. Let $p$ be the following polynomial: \begin{align} p(x) = (x-g(\alpha_1))\ldots(x-g(\alpha_k)) \end{align} Then $p$ has coefficients fixed by any element $g \in G$ since these permute the factors by the homomorphism property and the fact that $g$ would just permute the $\alpha_i$, and so the coefficients must lie in the fixed field of $G$, that is, $p \in F[x]$. Note that when applying $g$ to $p$ we do not "gain" any repeated factors, since $g(\alpha)\neq \alpha$ for all $g \in G$ with $g \neq$ id, by our earlier proof. Clearly $p$ is monic, and we claim that it must also be irreducible. To see this, note that if it factored non-trivially, then we must have some other polynomial $h \in F[x]$ with $\alpha$ as a root, and $g(\alpha)$ not a root for some $g \in G$. But then this $h$ will not have all coefficients in $F$, since the automorphisms in $G$ will not simply permute the factors as they did for $p$, and so we conclude no such $h$ can exist. Hence we deduce that in fact $p$ must be the minimal polynomial for $\alpha$.

I'm having trouble with the $\Leftarrow$ direction of the if and only if statement, but think the rest is okay. Any help or feedback would be great!