I'd like to compute the intermediate fields between $\mathbb{Q}$ and a splitting field for $f=X^4-10X^2+4$.

Here's what I've done:

The roots of $f$ are $\pm \sqrt{5\pm \sqrt{21}}$. Multiplying them we see that a splitting field for $f$ over $\mathbb{Q}$ is $\mathbb{Q}(\sqrt{5+\sqrt{21}})$ which has degree 4 over $\mathbb{Q}$.

Hence the order of the Galois group is 4. Its elements are determined by the action on $\sqrt{5+\sqrt{21}}$, which can be sent to either of the four roots of $f$. It's easily checked that the three non-identity elements have order $2$, hence the Galois group is isomorphic to $C_2\times C_2$.

By the Galois correspondence, there are three intermediate fields. One of them is obviously $\mathbb{Q}(\sqrt{21})$, but how do I get the other ones? I tried to manually determine the fixed fields for all three subgroups, but the calculations got too nasty for me.

  • $\begingroup$ How do you know $\sqrt{5-\sqrt{21}}\in\mathbb{Q}(\sqrt{5+\sqrt{21}})$ ? $\endgroup$ – Belgi Nov 14 '12 at 13:43
  • $\begingroup$ Their product is $2$. $\endgroup$ – Berci Nov 14 '12 at 13:51
  • 1
    $\begingroup$ Now that you know that the roots lie in $\mathbb Q(\sqrt 6, \sqrt {14}, \sqrt {21})$, you can compute $((\sqrt 6 + \sqrt{14})/2)^2$ and act surprised. $\endgroup$ – mercio Nov 14 '12 at 14:43
  1. $\Bbb Q\left(\sqrt{5+\sqrt{21}} + \sqrt{5-\sqrt{21}}\right)$
  2. $\Bbb Q\left(\sqrt{5+\sqrt{21}} - \sqrt{5-\sqrt{21}}\right)$

For an involutive automorphism $\phi$, for all elements $s$, we have that $s+\phi(s)$ and $s\cdot \phi(s)$ are always fixed, so they have a chance to generate the subfield..

Update: interestingly, $$ \begin{align} \left(\sqrt{5+\sqrt{21}} + \sqrt{5-\sqrt{21}}\right)^2 &= 14 \\ \left(\sqrt{5+\sqrt{21}} - \sqrt{5-\sqrt{21}}\right)^2 &= 6. \end{align} $$ Hmm.. In other words, the following 3 intermediate fields we have: $$\Bbb Q(\sqrt 6),\ \Bbb Q(\sqrt{14}),\ \Bbb Q(\sqrt{21})$$

| cite | improve this answer | |
  • $\begingroup$ Did you agree with the OP that the degree of the extension is $4$ ? $\endgroup$ – Belgi Nov 14 '12 at 13:51
  • $\begingroup$ Yes, it seems so. $\endgroup$ – Berci Nov 14 '12 at 13:51
  • 1
    $\begingroup$ Very clear. Thanks a lot. $\endgroup$ – user46225 Nov 14 '12 at 14:37

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.