Let $G$ be a nontrivial finite group. Given $n\in\mathbb{Z}_{\geq 1}$, let $G^n$ be the cartesian product of $n$ copies of $G$. Further let $S\subseteq G^n$ be a generating set of $G^n$.

Question: Do we always have $|S|\geq n$?


The answer is no. -- For example, ${{\rm A}_5}^3$ is $2$-generated. -- We have e.g. $$ \langle (2,5)(3,4)(6,7,8,9,10)(11,12,15), (1,3,2,4,5)(6,7,9,10,8)(11,13,12,14,15) \rangle \ \cong \ {{\rm A}_5}^3. $$ Finding such generator pair is easy -- just pick two random elements until you find some which generate.

Actually, according to Hall, even ${\rm A}_5^{19}$ is $2$-generated (while ${\rm A}_5^{20}$ is not). Cf. e.g. Pak's "On probability of generating a finite group", page 4.

  • $\begingroup$ Thanks a lot Stefan for the example. Now I can see why I could not prove it. So what kind of lower bound can you get as a function of $n$. Is $\sqrt{n}$ good enough? $\endgroup$ – Hugo Chapdelaine Nov 16 '14 at 0:40
  • $\begingroup$ @HugoChapdelaine: No, $\sqrt{n}$ is not, see my edit. $\endgroup$ – Stefan Kohl Nov 16 '14 at 0:41
  • $\begingroup$ Hmm, I did not think much about it, but I find it counter-intuitive for the least. $\endgroup$ – Hugo Chapdelaine Nov 16 '14 at 0:42

It is proved in

J. Wiegold, Growth sequences of finite groups, III, J. Austral. Math. Soc. Ser. A 25 (1978) 142–144

that a direct product of $n$ nonabelian simple (or quasisimple) groups can be generated by at most $2 + {\rm Ceiling}( \log_{60} n)$ elements.

On the other hand, given $r$ elements of $G^n$, there are at most $|G|^r$ possible projections on to the components. So, if they generate $G^n$, then no two of the projections can be the same, and hence $n \le |G|^r$ and we have the lower bound $r \ge \log_{|G|} n$.

So the best possible bound is logarithmic in $n$.

  • $\begingroup$ Thanks a lot @Derek for your argument which provides a logarithmic lower bound. I find it amazing that the order of magnitude of this (a priori) crude lower bound is the right one $\endgroup$ – Hugo Chapdelaine Nov 16 '14 at 13:45

For an alternative proof of Wiegold's results, and in particular for finite groups G such that the number of generators of G^n is comparable to log n, see Section 3 in Jacques Thévenaz, Maximal subgroups of direct product groups, Journal of Algebra 198 (1997) 352-361.

Pierre de la Harpe.


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.