Rado equations solved by linear combinations of idempotent ultrafilters (Q2052576)

The paper characterises the solvability of Rado equations inside linear combinations \(a_1\mathcal{U} + \cdots + a_n\mathcal{U}\) of idempotent ultrafilters \(\mathcal{U}\in\beta\mathbb{Z}\) by exploiting known relations between such combinations and strings of integers and generalises a partial characterisation obtained previously by \textit{M. Di Nasso} [Proc. Am. Math. Soc. 143, No. 4, 1749--1761 (2015; Zbl 1386.03057)]. More precisely, a string \((a_1,\ldots , a_n)\in\mathbb{Z}^{<\omega}\) is called reduced if we have: (a) for every \(i\in\{1,\ldots,n - 1\}\), \(a_i\neq a_{i+1}\); and (b) for every \(i\in\{1,\ldots,n\}\), \(a_i\neq 0\). One of the main results in the paper reads as follows: Let \(P \in \mathbb{Z}[x_1,\ldots, x_m]\) be a linear homogeneous polynomial and let \(\sigma \in\mathbb{Z}^{<\omega}\) be a reduced string. Then, under the assumption that there exists a strongly summable ultrafilter on \(\mathbb{Z}\), the following statements are equivalent: 1. For all idempotent ultrafilters \(\mathcal{U}\in\beta\mathbb{Z}\) we have that \[ \sigma\mathcal{U}\models P ( x_1 ,\dots, x_m ) = 0. \] 2. There exists a strongly summable ultrafilter \(\mathcal{U}\in\beta\mathbb{Z}\) such that \[ \sigma\mathcal{U}\models P ( x_1 ,\dots, x_m ) = 0. \] 3. \(P\) has a solution in \(\mathfrak{G}(\sigma)\), where \(\mathfrak{G}(\sigma)\) is the equivalence class of \(\sigma\) under a certain equivalence relation.