Let $d \geq 2$ be an integer, and let $\mathcal{F}_d$ be the family of finite groups such that $G \in \mathcal{F}_d$ if and only if every subgroup of $G$ can be generated by at most $d$ elements.

Is there a finite nontrivial word $w = w(x_1, \dots, x_n)$ which is trivial on $\mathcal{F}_d$?

That is, can it be that for any choice of $G \in \mathcal{F}_d$ and $y_1, \dots, y_n \in G$ we have $w(y_1, \dots, y_n) = 1$.

Alternatively: Let $F$ be a free group of rank $\aleph_0$. Is it possible that the intersection of all finite index subgroups $N \lhd F$ with $F/N \in \mathcal{F}_d$ is nontrivial?

I am also interested in the analogous question for $p$-groups.

  • 1
    $\begingroup$ This rank was introduced by Malcev and is called special rank. Namely, The special rank of a group $G$ is the minimal $d$ such that every finitely generated subgroup of $G$ can be generated by $d$ elements. $\endgroup$ – Anton Klyachko Dec 6 '14 at 20:04

No, there is not such a word, by the following two facts.

  1. The two elements $$A=\left(\begin{array}{cc}1&2\\0&1\end{array}\right),\quad B=\left(\begin{array}{cc}1&0\\2&1\end{array}\right)$$ of $\text{SL}_2(\mathbf{F}_p)$ satisfy no nontrivial relation of length less than $c\log p$.
  2. Every subgroup of $\text{SL}_2(\mathbf{F}_p)$ is generated by at most 2 elements.

For 1, there is a well known ping-pong argument which proves that $A$ and $B$ generate a free subgroup of $\text{SL}_2(\mathbf{Z})$. The argument can be reproduced in $\text{SL}_2(\mathbf{F}_p)$ for words which are not too long.

For 2, the subgroups of $\text{SL}_2$ are well understood. Refer for instance to Theorem 6.17 in Suzuki's book:

Suzuki, Michio. Group theory. I. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], 247. Springer-Verlag, Berlin-New York, 1982. xiv+434 pp. ISBN: 3-540-10915-3 MR0648772

  • 1
    $\begingroup$ Excellent! For a complete answer you need to argue why treating the case $n=2$ (as you did) is sufficient (it is possible that there are relations involving more than 2 elements). I think that such an argument has been given by @HJRW in the comments to his answer. $\endgroup$ – Pablo Jan 28 '15 at 10:14
  • 1
    $\begingroup$ Great! I didn't know that $SL_2(F_p)\in\mathcal{F}_2$ for all $p$. $\endgroup$ – HJRW Jan 28 '15 at 10:18
  • $\begingroup$ What happens if we fix a prime $p$, and consider only finite $p$-groups and not all finite groups? This was the last comment in my question. $\endgroup$ – Pablo Jan 28 '15 at 10:56
  • 1
    $\begingroup$ @Pablo Possibly you can adapt a proof that $F_2$ is residually $p$. Replace every occurence of $\mathbf{F}_p$ above with $\mathbf{Z}/p^n\mathbf{Z}$, now with $p$ fixed, and replace $A$ and $B$ respectively with $(1,p,0,1)$ and $(1,0,p,1)$. Then for $n$ large enough depending on your word the group $\langle A,B\rangle$ might suffice, but this requires better knowledge of $\text{SL}_2(\mathbf{Z}/p^n\mathbf{Z})$ than I claim to have. Maybe somebody else can comment. $\endgroup$ – Sean Eberhard Jan 28 '15 at 12:05
  • $\begingroup$ Thinking about it again after lunch, I think it's much more likely that if $G\in\mathcal{F}_d$ is nilpotent then its nilpotency class must be bounded in terms of $d$, which would of course imply that there is such a word. $\endgroup$ – Sean Eberhard Jan 28 '15 at 13:28

As I said in a comment on your other questions, there is no such word. Indeed, for any $w\in F_n$, there is $u\in F_{n-1}$ so that $w$ survives under the retraction $F_n\to F_{n-1}$ that sends the last generator to $u$.

  • $\begingroup$ And why there are no such words $w(x,y)\in F_2$? $\endgroup$ – Anton Klyachko Dec 4 '14 at 14:17
  • $\begingroup$ @AntonKlyachko - good point! Because $F_2$ is residually finite. $\endgroup$ – HJRW Dec 4 '14 at 14:37
  • $\begingroup$ @HJRW Have you noticed that I required every subgroup of $G$ to be generated by at most $d$ elements and not just $G$ itself? I just fail to see why does your answer solve my question... $\endgroup$ – Pablo Dec 4 '14 at 14:42
  • $\begingroup$ @Pablo - I'm sorry, you're right - I hadn't noticed this requirement. I guess my answer shows that you can take $n=d=2$. $\endgroup$ – HJRW Dec 4 '14 at 14:47
  • $\begingroup$ @HJRW I don't think I see how, but if this is true I find it very interesting... $\endgroup$ – Pablo Dec 4 '14 at 14:55

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.