I was looking at the Galois group of the splitting field of $x^4-7$ over $\mathbb{Q}$. I found it to be $\mathbb{Q}(\sqrt[4]{7},i)$, and the Galois group to be the dihedral group of order $8$. Now $D_8=\langle \sigma,\rho\mid \sigma^4=\rho^2=1, \sigma\rho=\rho\sigma^3\rangle$.

I know $D_8$ has $5$ subgroups of index $4$, there should be $5$ subfields of $\mathbb{Q}(\sqrt[4]{7},i)$ of degree $4$ over $\mathbb{Q}$ by the fundamental theorem of Galois theory.

I found $3$ of them to be $\mathbb{Q}(i\sqrt[4]{7})$, $\mathbb{Q}(\sqrt[4]{7})$, and $\mathbb{Q}(\sqrt{7},i)$.

I can't figure out what the other $2$ are, but I think they're the fixed fields of the subgroups $\{1,\rho\sigma\}$ and $\{1,\rho\sigma^3\}$, but I can't determine those. Does anyone have any idea what the other two subfields of degree $4$ over $\mathbb{Q}$ are? Many thanks.

  • 2
    $\begingroup$ This is just linear algebra: write down the action of $\rho\sigma$ on the splitting field as a matrix $M$ with respect to the $\mathbb{Q}$-basis $\{\sqrt[4]{7}^ni^m\}$ and compute the kernel of $M-I$. $\endgroup$ – Alex B. Nov 30 '11 at 7:27
  • $\begingroup$ @AlexB. The book I was using gave some ad hoc examples only. I wish I had thought of this method before, it works like a charm! Thank you very much. $\endgroup$ – Kiera Nov 30 '11 at 7:56
  • $\begingroup$ You are welcome, Kiera! $\endgroup$ – Alex B. Nov 30 '11 at 15:56

Let $a=\sqrt[4]{7}$, $b=i\sqrt[4]{7}$, $c=-\sqrt[4]{7}$, and $d=-i\sqrt[4]{7}$ be the four roots of $x^4-7$. Then the automorphisms $\rho$ and $\sigma$ permute these roots, with $$ \rho \,=\, (b\;\;d) \qquad\text{and}\qquad \sigma \,=\, (a\;\;b\;\;c\;\;d) $$ Observe then that: $$ \rho\sigma \,=\, (a\;\;d)(b\;\;c), \quad \rho\sigma^2 \,=\, (a\;\;c), \quad \rho\sigma^3\,=\, (a\;\;b)(c\;\;d), \quad\text{and}\quad \sigma^2 \,=\, (a\;\;c)(b\;\;d) $$ As you suggest, the intermediate fields $\mathbb{Q}(\sqrt[4]{7})$, $\mathbb{Q}(i\sqrt[4]{7})$, and $\mathbb{Q}(i,\sqrt{7})$ correspond to three of the subgroups of order two: $$ \mathbb{Q}(\sqrt[4]{7}) \leftrightarrow \{1,\rho\},\qquad \mathbb{Q}(i\sqrt[4]{7})\leftrightarrow \{1,\rho\sigma^2\}, \qquad \mathbb{Q}(i,\sqrt{7})\leftrightarrow\{1,\sigma^2\} $$ The main trick for figuring out the fixed field of $\{1,\rho\sigma\}$ is to consider the symmetrizing function $1+\rho\sigma$. Because of symmetry, anything in the image of $1+\rho\sigma$ must be fixed under $\rho\sigma$: $$ \rho\sigma(1+\rho\sigma)(x) \,=\, (\rho\sigma + (\rho\sigma)^2)(x) = (\rho\sigma + 1)(x) = (1+\rho\sigma)(x) $$ Now, observe that $$ (1+\rho\sigma)(a) \,=\, a+d = (1-i)\sqrt[4]{7} $$ and $$ (1+\rho\sigma)(b) \,=\, b+c = (i-1)\sqrt[4]{7} $$ Therefore, the fixed field of $\{1,\rho\sigma\}$ is $\mathbb{Q}\bigl((1-i)\sqrt[4]{7}\bigr)$. Using a similar calculation, we can see that the fixed field of $\{1,\rho\sigma^3\}$ is $\mathbb{Q}\bigl((1+i)\sqrt[4]{7}\bigr)$.

  • $\begingroup$ My book had mentioned these symmetrizing functions, but the explanation wasn't as clear as yours here. Cheers! $\endgroup$ – Kiera Nov 30 '11 at 8:03
  • $\begingroup$ In the first part of your explanation, mustn't $\left\{1,\sigma^2\right\}$ correspond with $\mathbb{Q}(i)$? If $\sigma^2 = (a c) (b d)$, then the only fixed element under $\sigma^2$ is $i$, isn't it? Also there is a minor typo in both your answer and the question - it seems you have copied this and thus made the same mistake as OP: you both typed $\sqrt{7}$ instead of $\sqrt[4]{7}$ for the subfield corresponding to $\left\{1,\sigma^2\right\}$. $\endgroup$ – Kasper Cools Dec 14 '16 at 22:23
  • $\begingroup$ @KasperCools The subgroup $\{1,\sigma^2\}$ has index four, so it must correspond to a degree four extension of $\mathbb{Q}$. Since $\mathbb{Q}(i)$ only has degree two, there must be more in the fixed field of $\{1,\sigma^2\}$ than just $\mathbb{Q}(i)$. Indeed, since $i = b/a = d/c$ and $\sqrt{7} = a^2 = c^2$, the permutation $\sigma^2 = (ac)(bd)$ fixes both $i$ and $\sqrt{7}$. (The field $\mathbb{Q}(i)$ actually corresponds to the subgroup $\{1,\sigma,\sigma^2,\sigma^3\}$.) I also don't agree that there's a typo. $\endgroup$ – Jim Belk Dec 15 '16 at 4:13

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.