In the Galois group of $\mathbb C(x)/\mathbb C$ we can identify the dihedral group $D_n=\langle \tau,\sigma \rangle $ where $\tau (x)=\frac{1}{x}$ and $\sigma(x)=\zeta_n x$.

Now, one can easily see that $\tau(x^n+x^{-n})=x^n+x^{-n}$ as well as $\sigma(x^n+x^{-n})=x^n+x^{-n}$, so the fixed field $\mathcal F(D_n)$ surely contains $\mathbb C(x^n+x^{-n})$. I want to show that it is also contained in the given set, so $\mathcal F(D_n)=\mathbb C(x^n+x^{-n})$.

Let's take an element $\frac{p(x)}{q(x)}\in\mathcal F(D_n)$. Since $\mathbb C$ is algebraically closed we can write $p(x)=\prod_{i=1}^n (x-\alpha_i)$ and $q(x)=\prod_{j=1}^m (x-\beta_j)$. If it is in the fixed field, we further obtain the relation $$\frac{(x-\alpha_1)\cdots(x-\alpha_n)}{(x-\beta_1)\cdots (x-\beta_m)}=\tau\left(\frac{(x-\alpha_1)\cdots(x-\alpha_n)}{(x-\beta_1)\cdots (x-\beta_m)}\right)=\frac{(x^{-1}-\alpha_1)\cdots(x^{-1}-\alpha_n)}{(x^{-1}-\beta_1)\cdots (x^{-1}-\beta_m)}$$

and we can do the same for $\sigma$: $$\frac{(x-\alpha_1)\cdots(x-\alpha_n)}{(x-\beta_1)\cdots (x-\beta_m)}=\sigma\left(\frac{(x-\alpha_1)\cdots(x-\alpha_n)}{(x-\beta_1)\cdots (x-\beta_m)}\right)=\frac{(\zeta_n x-\alpha_1)\cdots(\zeta_nx-\alpha_n)}{(\zeta_nx-\beta_1)\cdots (\zeta_nx-\beta_m)}$$

Does this help in any way? How can I show that $\frac{p(x)}{q(x)}$ looks like $x^n+x^{-n}$ for some $n\in\mathbb Z$?


Since $\mathcal{F}(D_n)$ is defined as the fixed subfield of $\mathbb{C}(x)$ by a finite group of automorphism, it is a Galois extension, with Galois group isomorphic to $D_n$ and degree $[\mathbb{C}(x):\mathcal{F}(D_n)]=|D_n|=2n$.

Let $y=x^n+x^{-n}$. Then $x^ny=x^{2n}+1$, so $\mathbb{C}(x)$ is a finite extension of $\mathbb{C}(y)$, with $[\mathbb{C}(x):\mathbb{C}(y)]\leq 2n$. But then considering the tower of extension $\mathbb{C}(y)\leq \mathcal{F}(D_n)\leq \mathbb{C}(x)$, one gets $$2n\geq [\mathbb{C}(x):\mathbb{C}(y)]=[\mathbb{C}(x):\mathcal{F}(D_n)]\cdot [\mathcal{F}(D_n):\mathbb{C}(y)]= 2n [\mathcal{F}(D_n):\mathbb{C}(y)],$$ so that $[\mathcal{F}(D_n):\mathbb{C}(y)]=1$ and thus $\mathcal{F}(D_n)=\mathbb{C}(y)$.

  • $\begingroup$ Thank you for your answer. Can you explain your last $\geq$? Is this really necessary or can we just write $=$? $\endgroup$ – Buh Jun 12 '18 at 12:25
  • $\begingroup$ @Buh Yes, sorry, it's an equality. I don't know why I only put a $\geq$, I've edited it now. $\endgroup$ – Arnaud D. Jun 12 '18 at 12:37
  • $\begingroup$ Actually, do you mind adding a little more explanation as to why $\text{Gal}((\mathbb C(x)/\mathcal F(D_n))\cong D_n$? Maybe it's trivial, but I don't think I understand it entirely. $\endgroup$ – Buh Jun 12 '18 at 17:47
  • $\begingroup$ @Buh It's a general result : for any field $K$ and finite subgroup $G$ of $Aut(K)$, the subfield $K^G$ of elements fixed by $G$ is such that $K/K^G$ is a Galois extension and $Gal(K/K^G)\simeq G$. $\endgroup$ – Arnaud D. Jun 12 '18 at 20:16
  • $\begingroup$ It's a classic theorem of characterization of Galois extensions, so you should be able to find it in most books or lecture notes about Galois theory. $\endgroup$ – Arnaud D. Jun 12 '18 at 21:06

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.