Let $F$ be a finitely generated free group, $H_1, \dots, H_n$ finitely generated subgroups of infinite index in $F$, and $\epsilon > 0$. Must there be an epimorphism to a finite group $\phi \colon F \to G$ such that $$\frac{|\phi(H_1 \cdots H_n)|}{|G|} < \epsilon ?$$ Where $H_1 \cdots H_n = \{h_1 \cdots h_n : h_1 \in H_1, \dots, h_n \in H_n\}$.

Equivalently: Is the Haar measure of the closure of $H_1 \cdots H_n$ in the profinite completion of $F$ equals $0$ ?

This question asks for a generalization of Is a free group a product of f.g subgroups of infinite index?


1 Answer 1


The answer is yes.

Let's prove this by induction on $n$. If $n=1$, then by M. Hall's theorem there is a finite index subgroup $K \leqslant F$ such that $H_1 \subseteq K$ and $|F:K|>1/\epsilon$. Choose any finite index normal subgroup $N \lhd F$, which is contained in $K$, let $G=F/N$ and let $\phi:F \to G$ be the natural epimorphism. Note that $|G:\phi(K)|=|F:K|$ as $N\subseteq K$. Then $$\frac{|\phi(H_1)|}{|G|} \le \frac{|\phi(K)|}{|G|}=\frac{1}{|G:\phi(K)|}= \frac{1}{|F:K|}<\epsilon .$$

Now, suppose that $n\ge 2$ and the statement has been proved for smaller $n$. By Theorem 1.1 in http://arxiv.org/abs/1308.3192, there is a finitely generated subgroup $H$ of infinite index in $F$, such that $H_{n-1} \subseteq H$ and $|H_n: (H_n \cap H)|=k<\infty$.

Thus $H_n \subseteq \bigcup_{j=1}^k H f_j$ for some $f_1,\dots, f_k \in F$. By the induction hypothesis, we can find an epimorphism $\phi$ from $F$ onto a finite group $G$ such that $$\frac{|\phi(H_1\dots H_{n-2}H)|}{|G|}< \frac\epsilon k.$$ Since $H_1 \dots H_n \subseteq \bigcup_{j=1}^k H_1 \dots H_{n-2}Hf_j$, we have \begin{multline*}\frac{|\phi(H_1\dots H_{n-2}H_{n-1}H_n)|}{|G|}\le \sum_{j=1}^k \frac{|\phi(H_1\dots H_{n-2}H)\phi(f_j)|}{|G|} \\ =\sum_{j=1}^k \frac{|\phi(H_1\dots H_{n-2}H)|}{|G|}<k \, \frac\epsilon k =\epsilon,\end{multline*} as required.

  • $\begingroup$ I asked this question in order to find a new proof for the theorem that you have used. That is, I know how to prove Theorem 1.1 in arxiv.org/abs/1308.3192 using my question (I only need $n = 2$), so I am really looking for a different proof (I knew this proof for $n=2$). $\endgroup$
    – Pablo
    Dec 22, 2014 at 16:37
  • 2
    $\begingroup$ For $n=2$ you don't need the full strength of the theorem from Olshanskii's paper. You can use an easier fact that there is a f.g., subgroup $H$, of infinite index in $F$, containing finite index subgroups of $H_1$ and $H_2$. This is not difficult to prove geometrically (see, for example, Thm. 1 in [Gitik, Rita, "Ping-pong on negatively curved groups". J. Algebra 217 (1999), no. 1, 65–72.] for a more general statement for quasiconvex subgroups in hyperbolic groups). $\endgroup$ Dec 22, 2014 at 16:54
  • $\begingroup$ Now I am interested in the hyperbolic case as a result of your comment and this interesting paper. So, do you know whether these amalgamated products constructed in the paper can be assured to have infinite index? I am ready to assume that the ambient hyperbolic group is residually finite or QCERF... $\endgroup$
    – Pablo
    Jan 18, 2015 at 17:46
  • $\begingroup$ Yes, with amalgamated products together with QCERF it should be easy to ensure that they have infinite index. $\endgroup$ Jan 18, 2015 at 19:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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