The Frobenius coin problem -- a cylindrical approach (Q2204980)

Let \(a\) and \(b\) be two relatively prime positive integers, both greater than \(1\). Let \(S\) be the set of positive integers that are not a nonnegative integer linear combination of \(a\) and \(b\). Using a principle of reciprocal measuring similar to the inclusion-exclusion principle, Sylvester showed the following which characterizes \(S\): \begin{itemize} \item[(i)] The largest element in \(S\) is \(ab-a-b\); \item[(ii)] For an integer \(x\) with \(0\le x\le ab-a-b\) we have the equivalence: \[x\in S \iff (ab-a-b-x)\notin S.\] \end{itemize} This paper describes a beautiful cylindrical interpretation for the set \(S\) and an alternative proof of Sylvester's statement. We define a new partial order \(\le_F\) on the integers, called the Frobenius order, by \[x\le_F y \iff y-x\text{ is a nonnegative integer linear combination of } a\text{ and }b.\] This gives a Frobenius poset \(P=(\mathbb{Z},\le_F)\) on the integers which can be embedded into a suitable cylinder such that the Frobenius order is preserved by shifting. An ideal \(I\) in \(P\) is a subset of \(P\) such that if \(x\le_F y\) and \(y\in I\), then \(x\in I\). Similarly, a filter is a subset \(J\) of \(P\) such that if \(x\le_F y\) and \(x\in J\), then \(y\in J\). Given a subset \(G\) of \(P\), then \(G\) generates the ideal \[\langle G\rangle_{\le_F}=\{x\in P\mid \exists y\in G, x\le_F y\}\] and the filter \[\langle G\rangle_{\ge_F}=\{y\in P\mid \exists x\in G, x\le_F y\}.\] An ideal, respectively a filter, is called principal if it is generated by one element \(x\), written as \(\langle x \rangle_\le\) and \(\langle x \rangle_\ge\). The main result is the following disjoint union of the integers with respect to an integer \(x\): \[\langle x+ab-a-b\rangle_{\le_F}\dot\cup\, \langle x\rangle_{\ge_F}=\mathbb{Z}.\] Now, Sylvester's statement is a straightforward consequence of this result.