Let $p$ be a prime number and $\alpha\in\mathbb{N}$ such that $\forall\beta\in\mathbb{Q} \space\alpha\neq \beta^p $ e.g. $\alpha$ is not a $p$-th power of any rational number. Let $E$ denote the splitting field of the polynomial $x^p-\alpha$.


$Gal(E/\mathbb{Q}) \cong G $ where we define $G$ as the group:

$G =\{\begin{pmatrix} a & b\\ 0 & 1 \end{pmatrix} s.t. \space a,b\in \mathbb{F}_p \space \wedge a\neq0 \}$

My attempt: denote the relevant Galois group as $H$, so it is obvious that $H$ is of order $p(p-1)$ because we can write $E / \mathbb{Q}$ as $ \mathbb{Q}(\zeta_p, \sqrt[p]{\alpha})/ \mathbb{Q} $. This is not a proof but rather intuition why the isomorphism might hold. Then I tried to explicitly define the isomorphism. So, we have to map every $\sigma\in H$ to a matrix in $G$. The mapping that seemed natural to me is the following:

Denote the roots of $x^p-\alpha$ as $\{x_1,...,x_p\}$, then we map an autormphism to a matrix by:

if $\sigma(\zeta_p) = x_i$ and $\sigma(\sqrt[p]{\alpha}) = x_j$ then $\sigma \mapsto M$ where we have for $M: a=i \space\wedge\space b=j\space$ e.g. we map the automorphism (enough to define on the generators of the extension, I think) to a matrix in $G$ correlated to the way the automorphism acts on both generators. However, this does not work and I'm stuck. Any clues, ideas or insight would be greatly appreciated.


I'll write $a$ instead of $\alpha$. The roots of $x^p-a$ are $\sqrt[p]{a}e^{\frac{2\pi ik}{p}}$ where $0\leq k\leq p-1$. Each element $\varphi\in Gal(E/\mathbb{Q})$ is uniquely determined by the images of $\sqrt[p]{a}$ and $e^{\frac{2\pi i}{p}}$ because they generate the field extension. What might the images be? $e^{\frac{2\pi i}{p}}$ is the root of $x^p-1$, hence $\varphi$ must send it to a root of this polynomial, though not to $1$. (since $\varphi$ is injective and $\varphi(1)=1$). So $\varphi(e^{\frac{2\pi i}{p}})=e^\frac{2\pi ij}{p}$ for some $1\leq j\leq p-1$. Now what is the image of $\sqrt[p]{a}$? It is the root of $x^p-a$, hence $\varphi(\sqrt[a]{p})=\sqrt[p]{a}e^{\frac{2\pi ik}{p}}$ for some $0\leq k\leq p-1$. And now define your isomorphism between $Gal(E/\mathbb{Q})$ and $G$ by sending such $\varphi$ to the matrix $\begin{pmatrix} j & k\\ 0 & 1 \end{pmatrix} $.

The map is injective because the images of $\sqrt[p]{a}$ and $e^{\frac{2\pi i}{p}}$ uniquely determine the element of $Gal(E/\mathbb{Q})$. It is onto because it is an injective map between finite sets of the same size. Finally, check that this map is indeed a group homomorphism. I'll leave it to you.

  • $\begingroup$ Thanks! Though I still wonder if this can be proven without explicitly defining the isomorphism.. $\endgroup$ – Adar Gutman May 10 at 19:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.