Using Ramsey's theorem once

From MaRDI portal
Publication:2274133



Abstract: We show that RT(2,4) cannot be proved with one typical application of RT(2,2) in an intuitionistic extension of RCA0 to higher types, but that this does not remain true when the law of the excluded middle is added. The argument uses Kohlenbach's axiomatization of higher order reverse mathematics, results related to modified reducibility, and a formalization of Weihrauch reducibility.


The paper studies the question of how many applications of one principle are needed to prove another in the context of higher-order reverse mathematics, and its connections to Weihrauch reducibility. The base theory is taken to be \(i\mathsf{RCA}^\omega_0\), consisting of Feferman's system \(\widehat{\mathsf{E\text-HA}}^\omega_\restriction\) of intuitionistic arithmetic of all finite types and the quantifier-free choice schema \(\mathsf{QF\text-AC}^{1,0}\); the extension of this theory to classical logic is the theory \(\mathsf{RCA}^\omega_0\) of \textit{U. Kohlenbach} [Lect. Notes Log. 21, 281--295 (2005; Zbl 1097.03053)]. Considering problems \(P\) of the form \(\forall x\,(p_1(x)\to\exists y\,p_2(x,y))\), the authors define a natural notion that a theory \(T\supseteq i\mathsf{RCA}^\omega_0\) proves a problem \(Q\) with ``one typical use of \(P\). They also consider a formal version \(T\vdash Q\le_WP\) of Weihrauch reducibility between such problems (which differs from actual Weihrauch reducibility \(Q\le_WP\) in that the reduction functionals are not a priori required to be computable). The main general results are that under suitable syntactic restrictions on \(P\) and \(Q\), \(i\mathsf{RCA}^\omega_0\) proves \(Q\) with one typical use of \(P\) iff \(i\mathsf{RCA}^\omega_0\vdash Q\le_WP\), and that this implies \(Q\le_WP\). As an application, the paper looks on dependencies between the Ramsey principles \(\mathsf{RT}(n,k)\), asserting that any \(k\)-colouring of \([\mathbb N]^n\) has an infinite monochromatic set. They show that even though \(i\mathsf{RCA}^\omega_0\vdash\mathsf{RT}(2,2)\to\mathsf{RT}(2,4)\), \(i\mathsf{RCA}^\omega_0\) cannot prove \(\mathsf{RT}(2,4)\) with only one typical use of \(\mathsf{RT}(2,2)\), using known results on Weihrauch reducibility between these problems. More generally, for any \(n\ge1\) and \(k>j\ge2\), \(i\mathsf{RCA}^\omega_0\) does not prove \(\mathsf{RT}(n,k)\) with only one typical use of \(\mathsf{RT}(n,j)\). Surprisingly, the classical theory \(\mathsf{RCA}^\omega_0\) \textit{does} prove \(\mathsf{RT}(2,4)\) with one typical use of \(\mathsf{RT}(2,2)\): the authors show this by a clever use of excluded middle to distinguish whether there exists an infinite bichromatic set or not. More generally, \(\mathsf{RCA}^\omega_0\) proves \(\mathsf{RT}(n,k)\) with one typical use of \(\mathsf{RT}(n,2)\).











This page was built for publication: Using Ramsey's theorem once

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2274133)