Minimal relations and the Diophantine Frobenius problem in embedding dimension three (Q518075): Difference between revisions

Let \(n_1,\dots, n_t\) be a set of positive integers with \(\gcd(n_1,\dots, n_t)=1\). The numerical semigroup \(S=\langle n_1,\dots, n_t\rangle\) generated by \(n_1,\dots,n_t\) is the set of all linear combinations of \(n_1,\dots,n_t\) with nonnegative integer coefficients. The Diophantine Frobenius Problem is to compute the largest positive integer not in \(S\). We call this integer the Frobenius number of \(S\) and denote it by \(F(n_1,\dots,n_t)\). It is well-known that if \(n_1\) and \(n_2\) are relatively prime, then \(F(n_1, n_2)=n_1n_2-n_1-n_2\). There is a huge literature about the Frobenius problem for numerical semigroups with three or more generators. See for example, the book of \textit{J. L. Ramírez Alfonsín} [The Diophantine Frobenius problem. Oxford: Oxford University Press (2005; Zbl 1134.11012)]. This paper focuses on the problem of giving an explicit formula for \(F(n_1, n_2, n_3)\). A result of \textit{F. Curtis} implies that there is no polynomial formula for this quantity [Math. Scand. 67, No. 2, 190--192 (1990; Zbl 0734.11009)]. A theorem of \textit{S. M. Johnson} reduces the problem of computing \(F(n_1, n_2, n_3)\) to the case where the three integers are pairwise relatively prime [Can. J. Math. 12, 390--398 (1960; Zbl 0096.02803)]. Let \(c_1\) be the smallest multiple of \(n_1\) that is in \(\langle n_2, n_3\rangle\), and define \(c_2\) and \(c_3\) analogously. Another result of Johnson, a variation of which is given in the author's earlier paper [Semigroup Forum 90, No. 1, 126--134 (2015; Zbl 1395.11042)], gives an expression for \(F(n_1, n_2, n_3)\) that involves \(c_1, c_2\), and \(c_3\). A main idea of this paper is that there exist positive integers \(\lambda_{ij}, \lambda_{ik}\) such that \[ n_i = n_j n_k - \lambda_{ij} n_j - \lambda_{ik} n_k, \] and that these integers can be expressed in terms of certain remainders. The author gives expressions for \(c_1, c_2, c_3\) that involve \(\lambda_{12}, \lambda_{13}, \lambda_{23}\). For a different recent perspective on this problem, see the paper of \textit{A. Tripathi} [J. Number Theory 170, 368--389; (2017; Zbl 1402.11046)].