1

Here is my problem: Let $k\subseteq F$ be a finite separable extension. Consider all homomorphisms $\phi:F\rightarrow \bar{k}$ extending the inclusion $k\subseteq \bar{k}$. Prove that $k\subseteq F$ is Galois if and only if the image of $\phi$ is independent of the homomorphism $\phi$.

I'm having a lot of trouble with this problem. I believe I have an outline for an argument, but feel like I'm overlooking some subtleties. I'd appreciate any feedback.

First note that the extension $k\subseteq F$ must be simple, as it is finite and separable. So, we may write $F=k(\alpha)$ for some element $\alpha$. Now, the embeddings $\phi$ described above are determined by where they send $\alpha$, and $\phi(\alpha)$ must be a root of the minimal polynomial $p(x)$ of $\alpha$ over $k$ (a separable polynomial).

For the 'if' direction, suppose that $\phi(F)=\phi'(F)$ for all such embeddings $\phi, \phi'$. Mapping $\alpha$ to itself shows that (perhaps up to some 'identification') $\phi(F)=F$ for all $\phi$. Let $\beta$ be any other root of $p(x)$. Take $\phi$ to be the homomorphism determined by $\alpha \mapsto \beta$. Then $\beta = \phi(\alpha)\in F$, so $F$ contains all roots of $p(x)$. It follows that $F=k(\alpha)$ is a splitting field for the separable polynomial $p(x)$, so $F$ is Galois over $k$, as needed.

For the converse, assume that $k\subseteq F$ is Galois. Then, in particular, this extension is normal. We'll show that $\phi(F)=F$ for all $\phi$ as above. Let $\beta$ be a root of the irreducible polynomial $f(x)\in k[x]$. For any $\phi$, we have

$$f(\beta)=0 \implies \phi(f(\beta))=f(\phi(\beta))=0,$$

which shows that $\phi(\beta)$ is a root of $f$. By normality of the extension, $\phi(\beta)\in F$, so we get $\phi(F)\subseteq F$. Dimension considerations give equality. So, the image of $\phi$ does not depend on $\phi$.

Sorry for the lengthy post; I'd appreciate any help.

  • 1
    This looks good to me! It's definitely the right argument, so nice work. Perhaps the slightly informal reasoning about "mapping $\alpha$ to itself" and "identifying $\phi(F)$ with $F$" could be avoided - just show the unique image is a splitting field for $p(x)$. For the other direction, you could write $F=k(\alpha)$ again and argue that each image is contained in the subfield of $\bar k$ which is a splitting field for $p(x)$, and appeal to uniqueness of splitting fields. These are only nitpicks/style suggestions - I'm aware that talking about identifications like this is common in this field. – Izaak van Dongen Oct 18 '23 at 23:32
  • Thank you! I'll make those adjustments. – ArbitraryElement Oct 18 '23 at 23:39
  • Sorry if this is off-topic, but I was halfway through writing an answer for your question at https://math.stackexchange.com/questions/4883933/inherited-group-law-on-cubic and it was deleted - are you sure you want to delete it? – KReiser Mar 20 '24 at 18:58
  • Sorry about that... I think I mostly solved the problem, although I'm still a bit stuck on the uniqueness part. I'll undelete and edit it to update my progress. Thanks for your time. – ArbitraryElement Mar 21 '24 at 15:30

0 Answers0