Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

To be more precise (but less snappy): is there an example of a group $G$ with finitely many infinite-index subgroups $H_1,\dots, H_n$ and elements $k_1,\dots, k_n$ such that $G$ is the union of the left cosets $k_1 H_1 , ..., k_n H_n\ ?$ And what if we relax the requirement that these all be left cosets, and ask: can $G$ be the union of finitely many such cosets, some being left cosets, others being right cosets?

If $G$ is amenable then this can't happen, since any coset of an infinite-index subgroup must have measure $0$. So this immediately rules out any abelian group $G$.

I've tried playing around with the only non-amenable groups that I'm comfortable with, the free groups on two or more generators. A few months ago I thought I found a simple counterexample in the free group on $\aleph_0$ generators, but now I've lost my notes and am beginning to doubt I ever had such an example.

(This question was asked to me by a friend who's interested in some kind of application to model theory, but I think it's interesting as a stand-alone puzzle.)

share|improve this question

1 Answer 1

up vote 26 down vote accepted

No. This follows from a beautiful theorem of B.H. Neumann:

Let $G$ be a group. If $\{x_iH_i\}_{i=1}^n$ is a covering of $G$ by cosets of proper subgroups, then $n \geq \min_{i} [G:H_i]$.

Explicitly, this is Lemma 4.1 in


As Neumann remarks, the identity $gH = (g H g^{-1}) g$ shows that it is no loss of generality to restrict to coverings by left cosets.

share|improve this answer
That was quick! Thanks, actually I read about this a long time ago and forgot that it applied to any group G... –  John Goodrick Mar 7 '10 at 17:37
You're quite welcome. I have had a side interest in covering problems of various sorts for a couple of years now, and Neumann's result is one of the nicest in this area. –  Pete L. Clark Mar 7 '10 at 17:42

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.