# Is there a logarithmic size generating set for some classes of finite semigroups?

Following my question Why is the minimum size of a generating set for a finite group at most $\log_2 n$?, we know that finite groups have generating sets of size at most $\log_2 n$, and a similar statement can be proven for quasigroups. Are there classes of semigroups (other than the class of groups) for which the minimum size of a generating set is at most logarithmic in the number of elements in the semigroup?

More formally, if $\mathcal{G}$ denotes the set of all finite groups and $\mathcal{S}$ denotes the set of all finite semigroups, is there a set $\mathcal{H}$ such that $\mathcal{G} \subsetneq \mathcal{H} \subsetneq \mathcal{S}$ and for each $G$ in $\mathcal{H}$ on $n$ elements, the semigroup $G$ has a generating set of size at most $\log_2 n$ (or even $O(\log_2 n)$)?

• I think the answer is yes, but probably not what you are looking for. Why not set H to be the class of all finite semigroups which have a generating set of size at most log2(n). There are certainly some semigroups in H which are not groups, for example, all the 2-generated semigroups that are not groups, and Jean-Eric's example is in S but not H. – James Mitchell Apr 21 '15 at 21:04

No. Take an $n$-element set $S$ with the multiplication defined by $xy = y$ for all $x, y \in S$. Then the minimal set of generators for $S$ is $S$ itself.
• Okay, this is one semigroup, the zero semigroup, that requires all $n$ elements to generate the semigroup. But there could be another class of semigroups that exclude these, right? – argentpepper Apr 16 '15 at 17:29
• This is not one semigroup, there is one for each $n$. – J.-E. Pin Apr 16 '15 at 17:57