Let $\xi$ be a 14-th primitive root of unity. I want to compute its conjugates.
In this question I gave the description of the process that I believed should be followed for computing the conjugates:
Assume $\alpha = \xi + \xi^{-1}$, $\sigma(\xi) = \xi^6$, $\tau(\xi) = \xi^2$
$Gal(\mathbb{Q}(\alpha)/\mathbb{Q}) \cong Gal(\mathbb{Q}(\xi)/\mathbb{Q})/Gal(\mathbb{Q}(\xi)/\mathbb{Q}(\alpha)) \cong Gal(\mathbb{Q}(\xi)/\mathbb{Q})/\langle \sigma \rangle$
I observe that $Gal(\mathbb{Q}(\xi)/\mathbb{Q}(\alpha))$ has degree 2 and by Artin's theorem $|Gal(\mathbb{Q}(\xi)/\mathbb{Q}(\alpha))| = 2$ this is the reason why the quotient is taken with $\langle \sigma \rangle$.
The resulting group has order 3 and therefore is cyclic, its conjugancy classes are $\{1,\sigma\}$, $\{\tau,\sigma\tau \}$, $\{\tau^2,\sigma\tau^2 \}$.
Then I just took one automorphism of each class and computed its image. But the people in the answers claim that the conjugates are not $\xi^4+\xi^{-4},\xi^2+\xi^{-2}$ as I thought but the say they are $\xi^3+\xi^{-3},\xi^5+\xi^{-5}$.
Why is it the case?