Skip to main content

Section 17.1 Galois Extensions

Subsection Group Actions and Automorphism Groups

“Give me extension and motion and I will construct the universe.”
―René Descartes

Definition 17.1. Field Automorphism Group.

Let \(K\) be a field. The automorphism group of \(K\text{,}\) written \(\Aut(K)\text{,}\) is the collection of field automorphisms of \(K\text{,}\) with the binary operation of composition.
The automorphism group of a field extension \(F \subseteq K\text{,}\) written \(\Aut(K/F)\text{,}\) is the subgroup of \(\Aut(K)\) consisting of those field automorphisms of \(K\) that restrict to the identity on \(F\text{.}\)

Exercise 17.2.

Let \(K / F\) be a field extension. Then \(\operatorname{Aut}(K)\) is a group under composition of maps, and \(\operatorname{Aut}(K / F)\) is a subgroup of \(\operatorname{Aut}(K)\text{.}\)

Example 17.3. \(\Aut(\C/\R)\).

I claim \(\Aut(\C/\R)\) has two elements (and so is a cyclic group of order \(2\)): the identity map on \(\C\) and the element \(\s\) given as complex conjugation. Each of these is an element of \(\Aut(\C/\R)\) — for \(\s\text{,}\) this amounts to the fact that complex conjugation commutes with addition and multiplication of complex numbers (and that it sends \(1\) to \(1\)).
To see these are the only elements of \(\Aut(\C/\R)\text{,}\) suppose \(\tau \in \Aut(\C/\R)\text{.}\) For any \(z = a + ib \in \C\text{,}\) we have \(\tau(z) = a + b \tau(i)\) since \(\tau|_\R = \id_\R\text{.}\) Moreover, \(-1 = \tau(-1) = \tau(i \cdot i) = \tau(i) \cdot \tau(i)\) and so \(\tau(i) = \pm 1\text{.}\) Thus \(\tau = \id\) or \(\tau = \s\text{.}\)

Example 17.4. \(\Aut(\Q(\sqrt d)/\Q)\).

For any square-free integer \(d\text{,}\) \(\Aut(\Q(\sqrt{d})/\Q)\) has order \(2\text{,}\) and its two elements are the identity and the map sending \(a + b \sqrt{d}\) to \(a - b \sqrt{d}\text{.}\) Checking that each really is an element of this group and that there are the only two elements in this group is done similarly to Example 17.3

Proof.

Let \(L\) be a field and let \(\sigma \in \operatorname{Aut}(L)\text{.}\) Then the Theorem 8.59 gives that there is an induced ring homomorphism \((-)^{\sigma}: L[x] \rightarrow L[x]\) such that for each \(q=a_{n} x^{n}+\cdots+a_{0} \in K[x]\text{,}\) we have
\begin{equation*} q^{\sigma}(x)=\sigma\left(a_{n}\right) x^{n}+\cdots+\sigma\left(a_{0}\right). \end{equation*}
If \(\sigma \in \operatorname{Aut}(L / K)\) and \(q \in K[x]\text{,}\) then \(q^{\sigma}=q\text{.}\)
  1. By assumption, \(\sigma\) is a homomorphism and it restricts to the identity on \(F\text{.}\) Thus for any polynomial \(q=a_{n} x^{n}+\cdots+a_{0} \in F[x]\text{,}\) we have
    \begin{equation*} \sigma(q(c))=\sigma\left(a_{n} c^{n}+\cdots+a_{0}\right)=\sigma\left(a_{n}\right) \sigma(c)^{n}+\cdots+\sigma\left(a_{0}\right)=a_{n} \sigma(c)^{n}+\cdots+a_{0}=q(\sigma(c)). \end{equation*}
  2. If \(\alpha\) is a root of \(f\text{,}\) then
    \begin{equation*} \begin{aligned} 0=\sigma(0) & =\sigma(q(\alpha)) \\ & =q(\sigma(\alpha)) \quad \text { by a }) \end{aligned} \end{equation*}
    showing that \(\sigma(\alpha)\) is also a root of \(q\text{.}\)
We now come to the main idea connecting field extensions and groups. It concerns the action of the group of automorphisms of a splitting field of a polynomial on the set of roots of that polynomial.

Proof.

  1. Let \(G=\operatorname{Aut}(L / F)\text{.}\) To see that the the action above is well-defined, notice that if \(b \in S\) then \(\sigma(b) \in S\) by Lemma 6.10. [provisional cross-reference: cite] Now we have
    \begin{equation*} \begin{gathered} \sigma \cdot\left(\sigma^{\prime} \cdot b\right)=\sigma\left(\sigma^{\prime}(b)\right)=\left(\sigma \circ \sigma^{\prime}\right)(b) \quad \text { for all } \sigma, \sigma^{\prime} \in G, b \in S, \\ 1_{G} \cdot b=\operatorname{id}_{K}(b)=b \quad \text { for all } \sigma \in G \text { and } b \in S, \end{gathered} \end{equation*}
    so the given formula does indeed define an action of \(G\) on \(S\text{.}\)
    This action is faithful: if \(\sigma\) fixes all the roots \(\alpha_{1}, \ldots, \alpha_{n}\) of \(f\text{,}\) then it fixes every element of \(F\left(\alpha_{1}, \ldots, \alpha_{n}\right)=L\text{.}\) Thus the corresponding group homomorphism \(\operatorname{Aut}(L / F) \rightarrow \operatorname{Aut}(S)\) is injective. On the other hand, the group of automorphisms on a set of \(n\) elements is isomorphic to \(S_{n}\text{,}\) so we have an inclusion of \(\operatorname{Aut}(L / F)\) into \(S_{n}\text{,}\) and thus \(\operatorname{Aut}(L / F)\) is isomorphic to a subgroup of \(S_{n}\text{.}\)
  2. Let \(\alpha, \beta\) be any two roots of \(f\text{.}\) By Theorem 5.69, there is an isomorphism \(\theta: F(\alpha) \rightarrow F(\beta)\) that fixes \(F\text{.}\)
    Our polynomial factors both as \(f=(x-\alpha) g\) and \(f=(x-\beta) h\text{.}\) Since \(f^{\theta}=f\) and \((x-\alpha)^{\theta}=x-\beta\text{,}\) we must have \(g^{\theta}=h\text{.}\) Theorem 5.69 applies, showing there is an automorphism \(\sigma: L \rightarrow L\) that extends \(\theta\text{.}\) In particular, \(\sigma\) fixes \(F\text{,}\) since \(\sigma\) extends \(\theta\) and \(\left.\theta\right|_{F}=\operatorname{id}_{F}\text{,}\) so \(\sigma \in \operatorname{Aut}(L / F)\text{.}\) Moreover, since \(\sigma\) extends \(\theta\) we have \(\sigma(\alpha)=\theta(\alpha)=\beta\text{.}\) This proves the action is transitive on the set of roots of any irreducible polynomial.
Soon we will show that if \(f \in F[x]\) is separable but not necessarily irreducible, and \(L\) is the splitting field of \(f\text{,}\) then the orbits of the action of \(\operatorname{Aut}(L / F)\) on the set of roots of \(f\) are precisely the sets of roots of the same irreducible factor of \(f\text{.}\) But to do so, we will need a little bit of Galois theory.

Proof.

We showed in Theorem 17.6 that \(\operatorname{Aut}(L / F)\) is isomorphic to a subgroup of \(S_{n}\text{,}\) and thus it as at most \(\left|S_{n}\right|=n!\) elements.
We will give an improved version of this result soon.

Exercise 17.8.

Let \(F\) be a field and \(L=F\left(a_{1}, \ldots, a_{n}\right)\text{,}\) where \(a_{1}, \ldots, a_{n}\) are elements in some extension of \(F\) that are algebraic over \(F\text{.}\) Each element \(\sigma \in \operatorname{Aut}(L / F)\) is uniquely determined by \(\sigma\left(a_{1}\right), \ldots, \sigma\left(a_{n}\right)\text{.}\)
A typical question that arises from Theorem 17.6 is to explicitly identify the automorphisms of a splitting field extension as a subgroup of the symmetric group.

Example 17.9. \(\Aut\) and \((x^3-2)\).

For a more complicated example, let \(K\) be the splitting field of \(x^3 - 2\) over \(\Q\text{.}\) Recall
\begin{equation*} K= \Q(\a_1, \a_2, \a_3) \end{equation*}
\begin{equation*} \a_1 = \sqrt[3]{2}, \a_2 = e^{2 \pi i/3}\sqrt[3]{2}, \a_3 = e^{4 \pi i/3}\sqrt[3]{2}. \end{equation*}
Let us ponder how big \(\Aut(K/\Q)\) could be. Pick any \(\s \in \Aut(K/\Q)\text{.}\) Since \(\s\) is a ring homomorphism, for any \(i\) we have
\begin{equation*} \s(\a_i)^3 = \s(\a_i^3) = \s(2) = 2 \end{equation*}
and thus \(\s(\a_i)\) is also a root of \(x^2 - 2\text{.}\) In other words, for each \(i\) we have \(\s(\a_i) = \a_j\) for some \(j\text{.}\) Moreover, since \(K\) is generated as a field extension of \(\Q\) by \(\a_1, \a_2, \a_3\text{,}\) the action of \(\s\) on the three roots completely determines the action of \(\s\) on all of \(K\text{.}\) In more detail, every element of \(K\) is given taking \(\Q\)-linear combinations of sums and products and quotients of these roots, and any element of \(\Aut(K/\Q)\) preserves sums, products and \(\Q\)-linear combinations.
To summarize, we have proven that there are {} \(3! = 6\) possibilities for \(\s\text{.}\) In fact, more is true: The function
\begin{equation*} \phi: \Aut(K/\Q) \to \Perm(\{\a_1, \a_2, \a_3\}) \end{equation*}
given by sending \(\s \in \Aut(K/\Q)\) to its restriction to the subset \(\{\a_1, \a_2, \a_3\}\) is an injective group homomorphism. Thus \(\Aut(K/\Q)\) is isomorphic to a subgroup of \(S_3\text{.}\) I claim that \(|\Aut(K/\Q)| = 6\) and hence \(\Aut(K/\Q) \cong S_3\text{.}\) I will prove this directly - we will learn of fancier methods to do so later.
First we notice that the field automorphism of \(\C\) given by complex conjugation, namely \(\C \to \C, z \mapsto \overline{z}\text{,}\) permutes the roots of \(x^3 - 2\) and hence it restricts to a field map from \(K\) into \(K\text{.}\) Since this map is \(\Q\)-linear and injective (as are all field maps) and \(K\) is a finite dimensional \(\Q\)-vector space, this map must be onto as well. So, we obtain an element \(\a \in \Aut(K/\Q)\) given by \(\a(z) = \overline{z}\) for all \(z \in K\text{.}\) It corresponds \((2 \, 3) \in S_3\text{.}\)
Next, we apply Corollary 16.76, which gives \(\gamma \in \Aut(K/\Q)\) such that \(\gamma(\sqrt[3]{2})=e^{2 \pi i/3} \sqrt[3]{2}\text{.}\) So, in the numbering above, \(\gamma\) corresponds to either \((1 \, 2)\) or \((1 \, 2 \, 3)\) in \(S_3\text{.}\) We don’t really know which of these it is. (In fact, both will occur — the map \(\gamma\) is not unique.) But either way we have proven the claim: For notice that both subsets \(\{ (2 \, 3), (1 \, 2)\}\) and \(\{ (2 \, 3), (1 \, 2 \, 3)\}\) of \(S_3\) generated all of \(S_3\text{.}\)
In other words, every possible permutation of roots of \(x^3 - 2\) arises as a field automorphism of its splitting field over \(\Q\text{.}\) This is what I meant before when I said that the roots of \(x^3 - 2\) are “as symmetric as possible’’.

Example 17.10. \([L:\Q]\) and \((x^4-2)\).

Let \(L\) be the splitting field of the irreducible polynomial \(q(x) = x^4 - 2 \in \Q[x]\text{.}\) So \(L = \Q(R)\) where \(R = \{z_1 = \sqrt[4]{2}, z_2 = i \sqrt[4]{2}, z_3= - \sqrt[4]{2}, z_4 = - i\sqrt[4]{2}\}\text{.}\) By Theorem 17.6, the action of \(\Aut(L/\Q)\) on \(R\) is faithful so that we have an injective group homomorphism \(\Aut(L/\Q) \to \Perm(R)\text{.}\)
Note that this map cannot possibly be onto: there is no \(\s \in \Aut(L/\Q)\) such that \(\s(z_1) = z_2\text{,}\) \(\s(z_2) = z_1\text{,}\) \(\s(z_3) = z_3\text{,}\) and \(\s(z_4) =z_4\text{;}\) i.e., the permutation \((1 \, 2)\) of these roots is not realizable by a field automorphism. To see this note that if \(\s(z_1) = z_2\) then \(\s(z_3) = \s(-z_1) = - \s(z_1) = - z_2 = z_4\text{.}\) So, any field automorphism that interchanges \(z_1\) and \(z_2\) would have to also interchange \(z_3\) and \(z_4\text{.}\) In fact, as we shall see, \(|\Aut(L/\Q)| = 8\text{,}\) considerably smaller than \(4!\text{.}\)
Let us compute \([L:\Q]\text{.}\) Note that \(i \in L\) since \(i = z_2/z_1\) and in fact \(L = \Q(z_1, i)\text{.}\) In the chain of extensions
\begin{equation*} \Q \subset \Q(z_1) \subset L = \Q(z_1)(i) \end{equation*}
the first one has degree \(4\text{,}\) since \(x^4 - 2\) is irreducible by Eisenstein’s Criterion, and the second has degree at most \(2\) since \(i\) is a root of \(x^2 + 1\text{.}\) It would be less than two if \(x^2+1\) factors in \(\Q(z_2)[x]\text{.}\) But since \(\Q(z_1) \subset \R\) and \(L\) is not contained in \(\R\text{,}\) the second extension cannot be trivial and so must have degree exactly \(2\text{.}\) We conclude \([L:\Q] = 8\text{.}\) It follows from Proposition that \(|\Aut(L/\Q)| \leq 8\text{.}\) (In fact, since \(L\) is the splitting field of a separable polynomial, the Theorem below will tell us that \(|\Aut(L/\Q)| = 8\text{.}\) But we won’t appeal to this fact here.)
We claim \(|\Aut(L/\Q)| = 8\) and \(\Aut(L/\Q)\) is isomorphic to the subgroup \(H\) of \(S_4\) generated by \((2 \, 4)\) and \((1 \,2 \, 3 \,4)\text{.}\) (This is isomorphic to \(D_8\text{.}\))
The map \(\C \to \C\) given by complex conjugation permutes the roots of \(x^4 - 2\) and it restricts to an automorphism of \(L\) — specially, it fixes \(z_1, z_3\) and interchanges \(z_2\) and \(z_4\text{.}\) It follows that complex conjugation determines an element \(\s \in \Aut(L/\Q)\) that corresponds to \((2 \, 4) \in H \leq S_4\text{.}\)
By the The Degree Formula we get \([L: \Q(i)] = 4\text{.}\) Since \(L = \Q(i)(z_1)\text{,}\) the degree of \(m_{z_1, \Q(i)}(x)\) must be \(4\text{.}\) This shows that \(x^4 - 2\) must remain irreducible as a polynomial in \(\Q(i)[x]\text{;}\) this is not obvious, but we have now proven it, and this fact will be useful in what we do next.
To construct another element of \(\Aut(L/\Q)\text{,}\) we use that that \(L\) is the splitting field of the polynomial \(x^4 - 2\) over \(\Q(i)\) and that, as we just showed, \(x^4 - 2\) is irreducible in \(\Q(i)[x]\text{.}\) We may thus apply Porism (also stated in the Corollary) to get that there is an element \(\tau \in \Aut(L/\Q(i))\) such that \(\tau(z_1) = z_2\text{.}\) We may regard \(\tau\) as an element of \(\Aut(L/\Q)\) since, by definition, \(\Aut(L/\Q(i))\) is a subgroup of \(\Aut(L/\Q)\text{.}\) We have
\begin{equation*} \tau(z_2) = \tau(i z_1) = \tau(i) \tau(z_1) = i z_2 = z_3 \end{equation*}
since \(\tau(i) = i\) by construction. A key point here is that if we had merely specified \(\tau\) to be an element of \(\Aut(L/\Q)\) sending \(z_1\) to \(z_2\text{,}\) then we would have no idea what \(\tau\) does to \(z_2\) — it was key to define \(\tau \in \Aut(L/\Q(i))\) as we did. We then also get \(\tau(z_3) = z_4\) and \(\tau(z_4) = z_1\text{.}\) So \(\tau\) corresponds to the permutation \((1 \,2 \, 3 \,4)\text{.}\)
We have proven that \(\Aut(L/\Q)\) is isomorphic to a subgroup of \(S_4\) having order at most \(8\) (by the Proposition above) and that it contains \((2 \, 4)\) and \((1 \,2 \, 3 \,4)\text{.}\) Since the subgroup generated by these two elements has order \(8\text{,}\) the claim follows.

Subsection Finite Extensions and Galois Groups

“Give me extension and motion and I will construct the universe.”
―René Descartes
We will now focus on finite field extensions and their automorphism groups. We start by giving a much better upper bound on the order of the automorphism group of a finite field extension.

Proof.

We proceed by induction on \([L: F]\text{.}\) In the base case, \([L: F]=1\text{,}\) and thus \(L=F\text{,}\) so \(\operatorname{Aut}(L / F)\) is the trivial group, and both statements hold.
Now let \(n \geqslant 1\) and suppose that \(|\operatorname{Aut}(L / F)| \leqslant[L: F]\) holds for all \(L\) and \(F\) such that \([L: F]<n\text{.}\) Let \([L: F]=n\text{.}\) Pick \(\alpha \in L \backslash F\) and let \(m=m_{\alpha, F}\text{,}\) and consider \(F(\alpha) / F\text{.}\)
Note that \(H=\operatorname{Aut}(L / F(\alpha))\) is a proper subgroup of \(G=\operatorname{Aut}(L / F)\text{.}\) By induction, we have \(|H| \leqslant[L: F(\alpha)]\text{.}\) We claim that it suffices to prove \([G: H] \leqslant[F(\alpha): F]\text{.}\) Indeed, using the Degree Formula and the fact that \(|G|=|H| \cdot[G: H]\text{,}\) if \([G: H] \leqslant[F(\alpha): F]\) then
\begin{equation*} |G|=|H| \cdot[G: H] \leqslant[L: F(\alpha)][F(\alpha): F]=[L: F]. \end{equation*}
To show that \([G: H] \leqslant[F(\alpha): F]\text{,}\) consider the function
\begin{equation*} \begin{gathered} G / H=\{\text { cosets of } H \text { in } G\} \stackrel{\Psi}{\longrightarrow}\{\text { roots of } m \text { in } L\} \\ g H \longmapsto g(\alpha) . \end{gathered} \end{equation*}
By Lemma 17.5, for any \(g \in G\) the element \(g(\alpha)\) is also a root of \(m\text{.}\) For any \(h \in H\text{,}\) we have
\begin{equation*} g h(\alpha)=g(h(\alpha))=g(\alpha). \end{equation*}
Thus \(\Psi\) is well-defined. Moreover, for any \(g_{1}, g_{2} \in G\) we have
\begin{equation*} \Psi\left(g_{1} H\right)=\Psi\left(g_{2} H\right) \Longleftrightarrow g_{1}(\alpha)=g_{2}(\alpha) \Longleftrightarrow g_{2}^{-1} g_{1}(\alpha)=\alpha \end{equation*}
which is equivalent to saying that \(g_{2}^{-1} g_{1}\) fixes \(F(\alpha)\text{,}\) and equivalently
\begin{equation*} g_{2}^{-1} g_{1} \in H \Longleftrightarrow g_{1} H=g_{2} H. \end{equation*}
This proves that the function \(\Psi\) is injective.
By Theorem [provisional cross-reference: cite] \(, \operatorname{deg}(m)=[F(\alpha): F]\text{.}\) Thus \(\Psi\) is an inclusion of \(G / H\) into a set with at most \([F(\alpha): F]\) many elements. Therefore,
\begin{equation*} [G: H]=|G / H| \leqslant[F(\alpha): F]. \end{equation*}
Now suppose that \(f\) is separable, so that
\begin{equation*} f=c \prod_{i=1}^{n}\left(x-\alpha_{i}\right) \in L[x]. \end{equation*}
with \(\alpha_{i} \neq \alpha_{j}\) for \(i \neq j\) and \(L=F\left(\alpha_{1}, \ldots, \alpha_{n}\right)\text{.}\)
Set \(\alpha=\alpha_{1}\) and let \(m\) be the irreducible factor of \(f\) that has \(\alpha\) as a root. Notice \(m=m_{\alpha, F}\text{.}\) As before, we consider \(F(\alpha)\) and set \(H=\operatorname{Aut}(L / F(\alpha)) \leq \operatorname{Aut}(L / F)=G\text{.}\) Note that \(L\) is also the splitting field of
\begin{equation*} g=\prod_{i=2}^{n}\left(x-\alpha_{i}\right) \in F(\alpha)[x] \end{equation*}
over \(F(\alpha)\text{,}\) and \(g\) is also separable. By induction \(|H|=[L: F(\alpha)]\text{,}\) and it remains to show that
\begin{equation*} [G: H]=[F(\alpha): F]=\operatorname{deg}(m). \end{equation*}
Since \(f\) is separable, so is \(m\text{,}\) so \(\operatorname{deg}(m)\) is exactly the number of distinct roots of \(m\text{.}\) Showing that \([G: H]=\operatorname{deg}(m)\) amounts to the assertion that the injective map \(\Psi\) is also surjective. This holds since \(G\) acts transitively on the roots of \(m\text{,}\) as shown in Theorem 17.6.
The finite field extensions whose automorphism group is as large as possible are very important, and are the main subject of this final chapter.

Definition 17.12. Galois Extension.

A finite field extension \(F \subseteq L\) is Galois if \(|\operatorname{Aut}(L / F)|=[L: F]\text{.}\) In this case we write \(\operatorname{Gal}(L / F)\) for \(\operatorname{Aut}(L / F)\text{,}\) and say \(\operatorname{Gal}(L / F)\) is the Galois group of \(L\) over \(F\text{.}\)

Example 17.13. \(\Q(\sqrt[3]2)\) not Galois.

I claim \(L = \Q(\sqrt[3]{2})\) is not a Galois extension of \(\Q\text{.}\) Let \(R\) be the set of all roots of \(x^3-2\) in \(L\text{.}\) Since \(L \subseteq \R\text{,}\) \(R\) has just one element: \(R = \{\sqrt[3]{2}\}\text{.}\) Since \(L = \Q(R)\text{,}\) the function \(\phi: \Aut(L/\Q) \to \Perm(R)\) is injective by Proposition CITEX and so since \(|R| = 1\text{,}\) we have \(|\Aut(L/\Q)| = 1\text{.}\) Thus it isn’t Galois. Since it is separable, \(L\) must not be a normal extension of \(\Q\text{.}\)
Theorem 17.11 tells us how to construct Galois extensions:

Example 17.15. Galois Extension.

\(L = \Q(\sqrt[3]{2}, e^{2 \pi i/3})\) is a Galois extension of \(\Q\text{,}\) since it the splitting field of \(x^3 - 2\text{.}\) We proved above that \(\Aut(L/\Q)\) has six elements and \([L: \Q] = 6\text{,}\) as the Theorem predicts.

Example 17.16. Galois and Characteristic \(p\).

Let \(F\) be a field of characteristic \(p\text{,}\) for a prime integer \(p\text{,}\) and assume \(L\) is a finite field extension of \(F\) such that there exists an element \(a\) of \(L\) with \(a \notin F\) but \(a^p \in F\text{.}\) Then \(F \subseteq L\) is not Galois since \(|\Aut(L/K)| < [L:K]\) in this case.
We will need the following notation.

Definition 17.17. Fixed Subfield.

If \(G\) is subgroup of \(\operatorname{Aut}(L)\text{,}\) the subfield of \(L\) fixed by \(G\text{,}\) denoted \(L^{G}\text{,}\) is by definition
\begin{equation*} L^{G}:=\{\alpha \in L \mid s(\alpha)=\alpha, \text { for all } s \in G\}. \end{equation*}

Exercise 17.18.

If \(G\) is subgroup of \(\operatorname{Aut}(L)\text{,}\) show that \(L^{G}\) is a subfield of \(L\text{.}\)

Example 17.19.

Let \(G=\operatorname{Aut}(\mathbb{C} / \mathbb{R})\text{.}\) Then \(\mathbb{C}^{G}\) is the subgroup of complex numbers fixed by all the elements in \(\Aut(\mathbb{C} / \mathbb{R})\text{,}\) which we saw in Example 17.3 has only two elements, the identity and the conjugation map \(s\text{.}\) Therefore, \(\mathbb{C}^{G}\) is the set of complex numbers fixed by conjugation, and thus \(\mathbb{C}^{\operatorname{Aut}(\mathbb{C} / \mathbb{R})}=\mathbb{R}\text{.}\)