4

My primary question is: is there an infinite field $F$ with a finite automorphism group $\text{Aut}(F)$?

So I tried fields with characteristic $2$, say $F_2(\pi)$. But it's still hard to make a positive conclusion since $\pi$'s image is seemingly arbitrary.

So is my conjecture right or not? If it's correct, could you give me any example?

Thanks in advance!

Vim
  • 13,640
  • Sure, $\mathbb Q[\sqrt 2]$ has an automorphism group of size $2$. – Thomas Andrews Dec 23 '15 at 18:06
  • 3
    Wait, are you talking field automorphisms? How is the automorphism group of the rationals infinite? – Thomas Andrews Dec 23 '15 at 18:06
  • @ThomasAndrews Yeah I was talking about field/ring automorphisms. Sorry I seemed to confuse them with group automorphisms when I made that reasoning. I will edit. Thanks for pointing out. – Vim Dec 23 '15 at 18:09
  • What field? Be specific. There is only one field automorphism of $\mathbb Q$, so you must be talking about some other sort of automorphism. – Thomas Andrews Dec 23 '15 at 18:10
  • So, what is wrong with $\mathbb Q$ as an answer to your question? – Thomas Andrews Dec 23 '15 at 18:11
  • @ThomasAndrews Sorry. I was having some confusion about the concepts just now. But now it's clear to me. Yes you are right – Vim Dec 23 '15 at 18:14
  • 3
    Also, it really really doesn't make sense to write $\mathbb F_2(\pi)$. $\pi$ is a real number, and can't be used as a generic indeterminate. As it turns out, $\mathbb Q(\pi)\cong\mathbb Q(x)$, the ring of rational functions if an indeterminate $x$, but that doesn't mean that $F(\pi)$ makes sense for all fields $F$. – Thomas Andrews Dec 23 '15 at 18:15
  • @ThomasAndrews Thanks for your advice. But is there anything wrong to write ${ r(\pi),r(x)\in F_2(x)}$ as $F_2(\pi)$? – Vim Dec 23 '15 at 18:18
  • ${r(\pi)}$ has no meaning in any ring that contains $\mathbb F_2$, so yes, there is something wrong with that. – Thomas Andrews Dec 23 '15 at 18:28
  • Given an element of a field, $a\in k$ and a rational function $q(x)\in K(x)$ for two different fields $k,K$, you can't define $q(a)$ in general, unless you give an embedding $K\to k$ (and even then, it is not always well-defined, if $a$ is a root of the denominator of $q$.) There is no embedding $\mathbb F_2\to \mathbb Q$, so $q(\pi)$ is meaningless. – Thomas Andrews Dec 23 '15 at 18:34
  • So, in $\mathbb F_2(x)$, $x+x=0$, but what is $\pi+\pi$? What does it mean for $2\pi=0$? What if you wanted to evaluate $q(\pi/2)$? Is $\frac{\pi}{2}+\frac{\pi}{2}=\pi=0?$ There's simply no way to define evaluation usefully when the coefficients or your polynomial are in unrelated fields. – Thomas Andrews Dec 23 '15 at 18:36
  • @ThomasAndrews I'll need some time to digest this comment. Thanks ! – Vim Dec 23 '15 at 18:37

3 Answers3

8

Aut$(\mathbb{R})$ is trivial. You can prove by induction that any automorphism $\phi$ fixes the natural numbers. Then using the definition of homomorphism you can show that it fixes the integers, then the rational numbers. Also, $\phi$ is order preserving since if $x>y$, then $x-y=r^2$ for some nonzero $r$, so $\phi(x)-\phi(y)=\phi(r^2)$. Finally, if $\phi(x) \neq x$, then wlog $\phi(x)<x$, so choose a rational $q$ with $\phi(x)<q<x$, and then we have $\phi(q)<\phi(x)$, but $\phi(q)=q>\phi(x)$, a contradiction. So $\phi$ is the identity.

Sorry, I should note that the first part of the argument shows that Aut$\mathbb{Q}$ is also trivial.

Noah Olander
  • 2,514
  • The "order-preserving" argument is spectacular. Thanks and plus one! – Vim Dec 23 '15 at 18:22
  • That's right, thanks! – Noah Olander Dec 23 '15 at 18:38
  • 2
    Note that order is crucial - once we pass to $\mathbb{C}$, the automorphism group becomes huge. – Noah Schweber Dec 23 '15 at 18:43
  • See also http://math.stackexchange.com/questions/449404/is-an-automorphism-of-the-field-of-real-numbers-the-identity-map. – lhf Dec 23 '15 at 19:32
  • 2
    Almost verbatim, if $F$ and $ G$ are isomorphic subfields of $R$, where $F$ contains the square roots of all its positive members, and if $\psi:F\to G$ is a field-isomorphism ,then $\psi = id_F$ and $F=G$. – DanielWainfleet Dec 23 '15 at 20:31
  • @NoahSchweber would you care to give a non-trivial aut of $\Bbb C$? I have trouble finding one.. – Vim Dec 24 '15 at 04:13
  • @Vim Conjugation. Assuming the axiom of choice, there are lots more - in fact $2^{2^{\aleph_0}}$, the maximum possible! - but all except conjugation are very pathological. In particular, it is consistent with ZF that the only nontrivial automorphism of $\mathbb{C}$ is conjugation. Building nontrivial automorphisms via Choice (specifically, Zorn's Lemma) is a good exercise - but also see http://math.stackexchange.com/questions/412010/wild-automorphisms-of-the-complex-numbers. – Noah Schweber Dec 24 '15 at 04:17
5

The group $\mathrm{Aut}(\mathbb Q)$ is actually trivial and therefore finite! Take an automorphism $$\sigma:\mathbb Q\to\mathbb Q$$

Suppose $\frac ab\in \mathbb Q^\times$ with $a,b\in \mathbb N$. Then $$\begin{align} \sigma(\frac ab)&=\frac{\sigma(a)}{\sigma(b)}\\&=\frac{\sigma(\overbrace{1+\cdots +1)}^{a\text{ times}}}{\sigma(\underbrace{1+\cdots +1)}_{b\text{ times}}}\\&=\frac{\overbrace{\sigma(1)+\cdots +\sigma(1)}^{a\text{ times}}}{\underbrace{\sigma(1)+\cdots +\sigma(1)}_{b\text{ times}}}\\&=\frac ab&&\text{since }\sigma(1)= 1. \end{align}$$

Mathmo123
  • 23,018
3

The $p$-adic numbers $\Bbb Q_p$ also have trivial automorphism group. The proof starts out in the same way as for $\Bbb R$, and then you have to show $\varphi(z)=z$ for $z\in\Bbb Q$ implies the same for $z\in\Bbb Q_p$. For this, you have to show that any automorphism $\varphi$ of the field is automatically continuous. I suppose there are many methods of doing this, but one way is to look at the set $S$ defined in the following purely algebraic way: an element $s\in\Bbb Q_p$ is in $S$ if and only if for every $m$ prime to $p$, $1+s$ has an $m$-th root in $\Bbb Q_p$.

Notice that such $s$ must necessarily be in $\Bbb Z_p$, the integers of the field. Furthermore, such $s$ can not be a unit of $\Bbb Z_p$, because either $1+s\in p\Bbb Z_p$, and has only finitely many roots in $\Bbb Q_p$, or $1+s$ is also a unit (necessarily then $p>2$), and doesn’t have a $(p-1)^2$-th root in $\Bbb Q_p$, because its class modulo $p$ doesn’t even have such a root.

On the other hand if $s\in p\Bbb Z_p$, then $1+s$ is a principal unit, and has an $m$-th root in $1+p\Bbb Z_p$ for all $m$ prime to $p$. Thus $S=p\Bbb Z_p$, whose powers are a neighborhood base at zero for the $p$-adic topology. These sets are consequently preserved by the automorphism $\varphi$, so $\varphi$ is continuous.

Lubin
  • 62,818