On the asymptotic behaviour of lengths of factorizations (Q1183351)

Let \(H\) be a monoid i.e. a multiplicative commutative and cancellative semigroup with unit element 1. For a subset \(U\subseteq H\) and an element \(a\in H\), \(a\neq 1\), let \[ \ell_ *(a,U)=\min\{k\in N_ +: a=u_ 1u_ 2\dots u_ k, u_ i\in U\}, \] \[ \ell^*(a,U)=\sup\{k\in N_ +: a=u_ 1u_ 2\dots u_ k, u_ i\in U\}. \] The authors study asymptotic properties of \(\ell_ *(a^ n,U)/n\) and \(\ell^*(a^ n,U)/n\) as \(n\) tends to infinity. They start with the most general setting making no restriction on \(H\) and only a minor restriction on \(a\). Then they specify the situation by taking \(H\) to be the multiplicative semigroup of a ring \(R\) and \(U\) being the set of irreducible elements. In this particular case they prove that the limits \(\lim_{n\to\infty}\ell_ *(a^ n,U)/n\) and \(\lim_{n\to\infty}\ell^*(a^ n,U)/n\) are positive rational numbers provided \(R\) is a Krull domain or Noetherian ring satisfying some additional conditions. This result settles a part of a conjecture put forward by \textit{D. F. Anderson} and \textit{P. Pruis} [Proc. Am. Math. Soc. 113, No. 4, 933-937 (1991; Zbl 0738.13018)]. Moreover, they show by an example that, for Noetherian domains in general, the original Anderson- Pruis conjecture has to be modified.