Core equality of real sequences (Q6930711)

The paper studies the \textit{core equality problem} for bounded real sequences with respect to ideals on the natural numbers. Given an ideal \(\mathcal{I}\) on \(\omega\), the \(\mathcal{I}\)-core of a bounded real sequence is defined as the smallest closed interval determined by its \(\mathcal{I}\)-limit inferior and superior; this generalizes the classical Knopp core (for \(\mathcal{I}=\mathrm{Fin}\)) and the statistical core (for the density zero ideal).\N\NThe main objective is to characterize infinite real matrices \(A\) such that \(\mathrm{core}_{Ax}(\mathcal{J})=\mathrm{core}_x(\mathcal{I})\) for all bounded real sequences \(x\), where \(\mathcal{I}\) and \(\mathcal{J}\) are ideals on \(\omega\). The first main result provides a complete characterization when an ideal \(\mathcal{J}\) is countably generated and \(A\) maps bounded sequences into bounded ones, i.e. \(A\in(\ell_\infty,\ell_\infty)\). In this case, the core equality holds if and only if \(A\) is \((\mathcal{I},\mathcal{J})\)-regular (that is, it preserves ideal convergence and limits of bounded sequences) and satisfies the condition that \(\mathcal{J}-\limsup_n \sum_{k \in E}\left|a_{n, k}\right|=1\) for every \(\mathcal{I}\)-positive set \(E\). This removes the nonnegativity assumption present in earlier results and extends classical theorems known for the ideal \(\mathrm{Fin}\). It is also shown that this characterization fails in general, notably when both ideals coincide with the density zero ideal.\N\NIt is proved that if \(\mathcal{I}\) is Rudin-Keisler below \(\mathcal{J}\), then there exists a matrix \(A\in(\ell_\infty,\ell_\infty)\) satisfying the core equality for all bounded sequences. This yields explicit examples with distinct ideals \(\mathcal{I}\neq \mathcal{J}\), answering an open question posed in earlier work in the literature.\N\NFinally, the author completely characterizes the ideals \(\mathcal{I}\) for which the equality \(\mathrm{core}_{Ax}(\mathrm{Fin})=\mathrm{core}_x(\mathcal{I})\) holds for some matrix \(A\). This happens if and only if \(\mathcal{I}=\mathrm{Fin}\) or \(\mathcal{I}\) is an isomorphic copy of \(\mathrm{Fin}\oplus\mathcal P(\omega)\). The proof combines techniques from matrix summability, ideal convergence, and the structure theory of analytic \(P\)-ideals and nowhere tall ideals.