16

I know that if a group $G$ has a finite number of subgroups then the group $G$ is finite.

But if a group $G$ has countable number of subgroups then is the group countable?

Shaun
  • 47,747
Learnmore
  • 31,675
  • 10
  • 110
  • 250
  • Firstly , it is very difficult to count all the subgroups of a given group –  Nov 13 '15 at 15:57
  • 1
    Are you making any set theoretic assumptions, like axiom of choice? Pretty sure you can look at all the cyclic subgroups, each of which can contain at most countable many, so there are uncountable many such groups. But that argument might be trickier without choice (or false). –  Nov 13 '15 at 15:57
  • Perhaps surprisingly, the converse is not true: http://math.stackexchange.com/questions/257175/countable-group-uncountably-many-distinct-subgroup?rq=1 – lhf Nov 13 '15 at 16:20

1 Answers1

15

Let $S$ be the set of all subgroups of $G$.

Consider the map $\phi: G \to S$ given $\phi(g) = \langle g \rangle$.

If $\phi$ were injective, then we'd be done.

Unfortunately, $\phi$ is not injective, but fortunately we can control how not injective it is.

Indeed, every $H \in S$ has a finite number of pre-images since every cyclic group has finite number of generators (including infinite cyclic groups). Therefore $$ G = \bigcup_{H \in S} \phi^{-1}(H) $$ is a countable union of finite sets and so is countable.

lhf
  • 221,500
  • Without choice I am not sure it is necessarily true, I know that countable union of pairs do not have to be countable (this situation would come up in a torsion free group), and I don't see any way to actually choose one or the other. It seems somewhat plausible to construct a group using this idea, if you say all these pairs generate distinct cyclic subgroups, and their unions is not countable. (this is related to the Russel's socks "problem": On the number of Russell’s socks or $2+2+2+. . .= ?$ byHorst Herrlich, Eleftherios Tachtsi. –  Nov 13 '15 at 16:55
  • @Paul: While I'm not saying that it cannot happen, most sets which cannot be well-ordered will also have an uncountable number of finite subsets. So you need to take extra care in which subsets are allowed to span subgroups and which are not. – Asaf Karagila Nov 13 '15 at 19:42
  • @lhf i understand that every cyclic group has finite number of generators but why is that $H$ has a finite number of pre-images? is it because every group is a union of cyclic groups? – Learnmore Nov 14 '15 at 02:24
  • @learnmore, if $H$ is not cyclic, then it has no pre-iimages. The image of $\phi$ is the set of cyclic subgroups of $G$. – lhf Nov 14 '15 at 10:16
  • @PaulPlummer I suspect the OP may not have been bothered about doing this without the axiom of choice, but you make an interesting point. It's not even obvious to me whether, without choice, there could be an uncountable vector space over the field of three elements with countably many subspaces. – Jeremy Rickard Nov 14 '15 at 14:30
  • 1
    @JeremyRickard I agree, I figured I would bring up the point since it is interesting. I decided to ask the question as a separate question here so I don't turn this question into a question it wasn't designed to be. –  Nov 16 '15 at 17:24
  • @PaulPlummer Good. By the way, I was talking about this with a friend, and he came up with a proof that an uncountable vector space over a finite field must have uncountably many one-dimensional subspaces, even without choice, answering the question in my comment. Neither of us has any idea about the more general group question, though. – Jeremy Rickard Nov 16 '15 at 17:33