Let $F,F'$ be two fields. Is there anything that can be said about the number of isomorphisms that can exist? In particular can there be more than one? What if $F$ is the complex numbers $\mathbb C$?

Context: I am trying to establish a bijection between characters in Banach algebras and maximal ideals in the algebra using the Gelfand Mazur theorem.

  • $\begingroup$ There are lots of automorphisms of $\Bbb C$. If you want to fix $\Bbb R$, then there's only one: $a+bi \mapsto a-bi$, but if you just want it to be an automorphism as a ring there are uncountably many (this requires the axiom of choice, but I'm a firm believer in a set's right to choose). See for instance this article by Paul Yale. $\endgroup$ – user98602 Mar 31 '14 at 9:09

There are $2^{2^{\aleph_0}}$ automorphisms of $\mathbb C$, a mind-boggling cardinal, and of those only the identity and conjugation can be explicitly described.
The others depend on set-theoretical wizardry (axiom of choice).

What is much more interesting is to take a field extension $k\subset K$ and to study the group $Aut(K/k)$ of automorphisms $\sigma: K\to K$ fixing $k$ pointwise: $\sigma(q)=q $ for all $q\in k$.
This study is called Galois theory, although in a strict sense Galois theory is restricted to the case where the extension $k\subset K$ is Galois, i.e. algebraic, separable and normal.
The results obtained in this set-up are then much more reasonable.

For example:
You may see $K$ as a vector space over $k$ (a point of view introduced, I think by Emil Artin).
If the dimension of that vector space $K$ over $k$ is finite and equal to $d$, then $Aut(K/k)$ is a finite set containing $\leq d$ elements, independently of whether the extension is Galois or not.
The simplest illustration is $Aut(\mathbb C/\mathbb R)$, a group consisting of just two elements, namely the conjugation and the identity of $\mathbb C$.
Contrast this with the absurdly large automorphism group of $\mathbb C$ evoked at the beginning!

  • 2
    $\begingroup$ To be fair the first group you describe is $\mathrm{Aut}(\mathbb{C}/\mathbb{Q})$, isn't it? $\endgroup$ – Najib Idrissi Mar 31 '14 at 9:38
  • 1
    $\begingroup$ Dear @nik: yes, that's absolutely correct: +1. In general the absolute automorphism group $Aut(K)$ of a field is equal to the relative automorphism group $Aut(K/k)$ where $k$ is the prime field of $k$, the smallest subfield of $K$ . That prime field $k$ is isomorphic to $\mathbb Q$ or $\mathbb F_p$ according as $K$ has characteristic $0$ or $p$. $\endgroup$ – Georges Elencwajg Mar 31 '14 at 9:50

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.