Maximal denumerant of a numerical semigroup with embedding dimension less than four. (Q1947250)

A numerical semigroup is a submonoid of \(\mathbb N\) with finite complement in \(\mathbb N\), where \(\mathbb N\) denotes the set of nonnegative integers. Every numerical semigroup \(S\) admits a unique minimal generating system, that is, a subset \(A\) of \(S\) such that every element is a linear combination of elements in \(A\) with coefficients in \(\mathbb N\), and such that no proper subset of \(A\) has this property. Moreover, this set has finitely many elements, say \(A=\{a_1,\dots,a_t\}\), and \(t\) is known as the embedding dimension of \(S\). Thus given \(s\in S\), there exists \((c_1,\dots, c_t)\in\mathbb N^t\) such that \(s=\sum_{i=1}^tc_ia_i\). We say that \((c_1,\dots,c_t)\) is a factorization of \(s\), that its length is \(c_1+\cdots+c_t\). Every \(s\in S\) has finitely many factorizations, and its cardinality is known as the denumerant of \(s\in S\). The maximal denumerant of \(s\in S\) is the number of factorizations of \(s\) having maximal length, denoted by \(\mathrm{d}_{\max}(s;S)\). The authors prove that the maximum of the set \(\{\mathrm{d}_{\max}(s;S)\mid s\in S\}\) exists, and they call it the maximal denumerant of \(S\), \(\mathrm{d}_{\max}(S)\). In order to prove this, they show that this maximum is reached in the set of the elements in \(S\) that are maximally reduced. An element is said to be maximally reduced if for every \(i\in\{1,\dots,t\}\) there exists a factorization with maximal length with its \(i\)-th coordinate equal to zero. Then the authors prove that every maximally reduced element is less than or equal to \((a_1-1)\sum_{i=2}^t a_i\), and thus there are only finitely many maximally reduced elements. This also gives a procedure to compute \(\mathrm{d}_{\max}(S)\). The authors explicitly compute \(\mathrm{d}_{\max}(\langle a_1,a_2\rangle)=1\). By determining the set of all maximal reduced elements in \(\langle a_1,a_2,a_3\rangle\), they also find formulas for \(\mathrm{d}_{\max}(\langle a_1,a_2,a_3\rangle)\) in terms of \(a_1\), \(m=(a_2-a_1)/g\) and \(n=(a_3-a_1)/g\), with \(g=\gcd(a_2-a_1,a_3-a_1)\). Special families of embedding dimension three numerical semigroups are then considered, and many examples are given to illustrate the results obtained.