Show that $\sigma(\sqrt[7]{2}), \sigma(e^{\frac{2\pi i}{7}})$ determine an automorphism $\sigma$ on $\mathbb{Q}(\sqrt[7]{2}, e^{\frac{2\pi i}{7}})$.
My Attempt: Let $\alpha=\sqrt[7]{2}$ and $\zeta=e^{\frac{2\pi i}{7}}$. Indeed, for every automorphism $\sigma$ on $\mathbb{Q}(\alpha, \zeta)$, we must have $\sigma(\alpha)\in A:=\{\alpha, \zeta\alpha, \zeta^2\alpha, \zeta^3\alpha, \dots, \zeta^6\alpha\}$, and $\sigma(\zeta)\in B:=\{\zeta, \zeta^2, \dots, \zeta^6\}$.
Let $\sigma(\alpha)=a\in A$ and $\sigma(\zeta)=b \in B$, then define $\sigma(\frac{p(\alpha, \zeta)}{q(\alpha, \zeta)})=\frac{p(a, b)}{q(a, b)}$, where $p, q \in \mathbb{Q}[t_1, t_2]$. We claim that $\sigma$ is an automorphism.
- First we must show that $\sigma$ is well-defined. Suppose $\frac{f(\alpha, \zeta)}{g(\alpha, \zeta)}=\frac{p(\alpha, \zeta)}{q(\alpha, \zeta)}$, where $g(\alpha, \zeta)\neq0\neq q(\alpha, \zeta)$. We claim that $\sigma(\frac{f(\alpha, \zeta)}{g(\alpha, \zeta)})=\sigma(\frac{p(\alpha, \zeta)}{q(\alpha, \zeta)})$, i.e., $\frac{f(a, b)}{g(a, b)}=\frac{p(a, b)}{q(a, b)}$.
Proof:
Lemma: Suppose $p, f\in\mathbb{Q}[t_1, t_2]$, then $f(\alpha. \zeta)=p(\alpha, \zeta)\implies f(a, b)=p(a ,b)$.
Proof of Lemma: Note $a=\alpha\zeta^i, b=\zeta^j$ for some $0\leq i\leq6, 1\leq j\leq6$. Note each term in $f, p$ has the form $ct_1^mt_2^n$, where $c\in\mathbb{Q}$; hence for $f(\alpha, \zeta)$, each term has the form $c\alpha^m\zeta^n=c2^k\alpha^{m'}\zeta^{n'}$, where $0\leq m', n'\leq6$, $m'\equiv_7m, n'\equiv_7n$, and $m-m'=7k$. The same term in $\mathbf{f(a, b)}$ becomes $\mathbf{ca^mb^n=c(\alpha\zeta^i)^m(\zeta^j)^n=c\alpha^m\zeta^{im}\zeta^{jn}=c\alpha^m\zeta^{im+jn}}=c2^k\alpha^{m'}\zeta^t$, where $\mathbf{0\leq m', t\leq6}$, $\mathbf{m'\equiv_7m, t\equiv_7im+jn}$. Thus $f(\alpha, \zeta)=c_0+c_1\alpha+c_2\zeta+c_3\alpha\zeta+c_4\alpha^2+\dots+c_s\alpha^6\zeta^6$, where $c_i\in\mathbb{Q}$. Similarly, $p(\alpha, \zeta)=d_0+d_1\alpha+d_2\zeta+d_3\alpha\zeta+d_4\alpha^2+\dots+d_s\alpha^6\zeta^6$, where $d_i\in\mathbb{Q}$. If $f(\alpha, \zeta)=p(\alpha, \zeta)$, then we must have $c_0=d_0, c_1=d_1, \dots, c_s=d_s$.
Now, by the boldfaced text, the only difference between the terms in $f(\alpha, \zeta)$ and $f(a, b)$ corresponding to $ct_1^mt_2^n$ is that $\zeta^{n'}$ becomes $\zeta^t$. Screeching halt.
My Question: Should I define $\sigma$ explicitly in a different way? Or is there a trick that I could use? Any help would be greatly appreciated.