A field $k$ is rigid if its group of automorphisms $\mathrm{Aut}(k)$ is trivial. For example, the reals $\mathbb R$ and the prime fields $\mathbb Q$ and $\mathbb F_p$.

A field $k$ is algebraically closed if every nonconstant polynomial $p(x) \in k[x]$ has a root in $k$. For instance, the algebraic closures of the above: $\bar{\mathbb R} = \mathbb C$, $\bar{\mathbb Q}$, etc.

Is any field both rigid and algebraically closed?

I am inclined to say no, mostly because I couldn't think of any, but also because it feels like in an algebraically closed field, there should always be finite index subfields which can be fiddled around with. Regardless, I have no idea how to prove or disprove this.

  • $\begingroup$ Finite index subfields of algebraically closed fields are actually quite rare: they never exist in positive characteristic, and in characteristic 0 every finite index subfield of an algebraically closed field has index 2. This is known as the Artin-Schreier theorem (actually, the theorem is a bit stronger and says more about what such subfields look like in the characteristic 0 case). $\endgroup$ – Eric Wofsey Jan 11 '17 at 3:40

Any algebraically closed field has tons of automorphisms. The key fact is that if $K$ and $L$ are algebraically closed and algebraic over subfields $k\subseteq K$ and $\ell\subseteq L$, then any isomorphism $f:k\to \ell$ extends to an isomorphism $\bar{f}:K\to L$. The proof is to just extend $f$ to one element of $K$ at a time: given $\alpha\in K\setminus k$, there exists $\beta\in L$ whose minimal polynomial over $\ell$ is the polynomial obtained by applying $f$ to the coefficients of minimal polynomial of $\alpha$. You can then extend $f$ to an isomorphism $k(\alpha)\to\ell(\beta)$ which sends $\alpha$ to $\beta$. Repeating this by transfinite induction, you get an extension of $f$ to an isomorphism $\bar{f}$ defined on all of $K$. The image of $\bar{f}$ must be all of $L$, since it is an algebraically closed field containing $\ell$ and $L$ is algebraic over $\ell$.

So given an algebraically closed field $K$, let $k_0$ be the prime subfield and let $B$ be a transcendence basis for $K$ over $k_0(B)$, so $K$ is algebraic over $k_0(B)$. Now let $k\subset K$ be some nontrivial finite Galois extension of $k_0(B)$ and let $f:k\to k$ be a nontrivial automorphism. By the key fact above (with $\ell=k$ and $L=K$), $f$ extends to a nontrivial automorphism of $K$.

(Note that this argument uses the axiom of choice, both in the proof of the key fact (to choose $\alpha$ and $\beta$ at every step of the induction) and in getting a transcendence basis for $K$ over $k_0$. I don't know whether it is possible for a rigid algebraically closed field to exist if you don't assume the axiom of choice.)

  • $\begingroup$ Re the last part, we should ask Andreas Blass, which is a user here. :) $\endgroup$ – Pedro Tamaroff Jan 11 '17 at 3:37
  • $\begingroup$ Why does $k$ have nontrivial automorphisms? How does your argument work when, say, $K = \mathbb C$? $\endgroup$ – algorithmshark Jan 11 '17 at 3:40
  • $\begingroup$ $k$ has nontrivial automorphisms since it is a nontrivial Galois extension of $k_0(B)$, and so it has $[k:k_0(B)]$ different automorphisms that fix $k_0(B)$. For a "concrete" example with $K=\mathbb{C}$, let $B$ be a transcendence basis for $\mathbb{C}$ over $\mathbb{Q}$, and let $k=\mathbb{Q}(B)(\sqrt{2})$. Then $k$ has an automorphism that fixes $\mathbb{Q}(B)$ and sends $\sqrt{2}$ to $-\sqrt{2}$. $\endgroup$ – Eric Wofsey Jan 11 '17 at 3:43

This question deals with a similar topic. Specifically, one answer cites a paper that constructs (at least one) real closed field with no non-trivial automorphisms. Unfortunately, often fields of this sort are very difficult to describe, and I'm unable to find a .pdf online.

I encourage you to check out the question, but the paper (which I was able to find at sci hub, but won't link) is:

S. Shelah, Models with second order properties. IV. A general method and eliminating diamonds -- Annals Pure and Applied Logic 25 (1983) 183-212

DOI is http://dx.doi.org/10.1016/0168-0072(83)90013-1


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.