The reverse mathematics of non-decreasing subsequences

From MaRDI portal



Abstract: Every function over the natural numbers has an infinite subdomain on which the function is non-decreasing. Motivated by a question of Dzhafarov and Schweber, we study the reverse mathematics of variants of this statement. It turns out that this statement restricted to computably bounded functions is computationally weak and does not imply the existence of the halting set. On the other hand, we prove that it is not a consequence of Ramsey's theorem for pairs. This statement can therefore be seen as an arguably natural principle between the arithmetic comprehension axiom and stable Ramsey's theorem for pairs.


This article analyzes the logical strength of two principles called \(\mathsf{LNS}\) (for \textit{limit non-decreasing subsequence}) and \(\mathsf{CNS}\) (for \textit{computably bounded non-decreasing subsequence}) in the context of reverse mathematics. These principles are both variations of the statement ``every function \(f : \mathbb{N} \rightarrow \mathbb{N}\) admits an infinite non-decreasing subsequence. The principle \(\mathsf{LNS}\) states that if a function \(f : \mathbb{N} \times \mathbb{N} \rightarrow \mathbb{N}\) satisfies \(f(n,s+1) \leq f(n,s)\) for every \(n,s \in \mathbb{N}\), then the limiting function \(\tilde{f}(n) = \lim_s f(n,s)\) admits an infinite non-decreasing sequence. The principle \(\mathsf{CNS}\) states that if \(f : \mathbb{N} \times \mathbb{N} \rightarrow \mathbb{N}\) and \(g : \mathbb{N} \rightarrow \mathbb{N}\) are functions where \(\lim_s f(n,s)\) exists and is \(\leq g(n)\) for every \(n \in \mathbb{N}\), then the limiting function \(\tilde{f}(n) = \lim_s f(n,s)\) admits an infinite non-decreasing sequence. The principles \(\mathsf{LNS}\) and \(\mathsf{CNS}\) are compared to well-known combinatorial principles related to König's lemma and Ramsey's theorem over the base system \(\mathsf{RCA}_0\). Together, the following main results show that \(\mathsf{LNS}\) and \(\mathsf{CNS}\) have interesting positions in the zoo of combinatorial principles below \(\mathsf{ACA}_0\). \(\mathsf{CNS}\) is strictly stronger than \(\mathsf{LNS}\). Indeed, the conjunction of \(\mathsf{LNS}\), the Erdős-Moser principle (stating that every countably infinite tournament has an infinite transitive subtournament), and weak König's lemma (which is König's lemma restricted to infinite \(\{0,1\}\)-branching trees) does not suffice to prove \(\mathsf{CNS}\). \(\mathsf{CNS}\) (and therefore \(\mathsf{LNS}\)) does not imply weak weak König's lemma (which is König's lemma restricted to \(\{0,1\}\)-branching trees of positive measure). The conjunction of Ramsey's theorem for pairs and weak König's lemma does not prove \(\mathsf{CNS}\). As with Ramsey's theorem for pairs, every computable instance of \(\mathsf{CNS}\) admits a low\(_2\) solution. The following questions are raised. Does \(\mathsf{CNS}\) imply Ramsey's theorem for pairs? It is shown that \(\mathsf{CNS}\) at least implies stable Ramsey's theorem for pairs. Does Ramsey's theorem for pairs imply \(\mathsf{LNS}\)?





Describes a project that uses

Uses Software






This page was built for publication: The reverse mathematics of non-decreasing subsequences

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2402955)