Semigroup structures on \(\beta\mathbb{N}\) (Q1185777)

The paper contains some remarkable algebraic results about \(\beta\mathbb{N}\). Any binary operation \((m,n)\mapsto m\circ n\) on \(\mathbb{N}\) can be extended to \(\beta\mathbb{N}\) by taking, for \(x,y\in\beta\mathbb{N}\), nets \(m_ i\to x\), \(n_ j\to y\) and writing \(x\circ y=\lim_ i\lim_ j m_ i\circ n_ j\) (and here the order of limits is crucial). When \(\circ\) is associative on \(\mathbb{N}\), it is also associative on \(\beta\mathbb{N}\), but the same is not true of commutativity (not even for +), as was well-known. This paper contains the very strong non-commutativity result that, provided that for each fixed \(m\in\mathbb{N}\), \(m\circ n\to\infty\) as \(n\to\infty\) holds, for each \(\xi\in\mathbb{N}^*=\beta\mathbb{N}\backslash\mathbb{N}\), the set \(\{\eta\in\mathbb{N}^*: (\mathbb{N}^*\circ\xi)\cap(\mathbb{N}^*\circ\eta)\neq\emptyset\}\) is nowhere dense in \(\mathbb{N}^*\). (Indeed, the first \(\circ\) can be replaced by a different binary operation and the conclusion still holds!) For the special operations + and \(\times\), other striking results are obtained. An element of \(\mathbb{N}^*\) which is not right cancellable must factorize. On the other hand, the elements of \(\mathbb{N}^*\) which do not factorize in \(\mathbb{N}^*\) generate (modulo \(\mathbb{Z})\) a free semigroup. The proofs, in comparison with others in this area, are transparent.