Truth definitions without exponentiation and the \(\Sigma _{1}\) collection scheme (Q2892683)

The paper presents three results related to the notoriously difficult open problem: Is the theory \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}+\lnot \mathrm{B}\Sigma_1\) consistent? It is well known that if there is a model of \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}\) with a universal \(\Sigma_1\) formula, then there is also a model of \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}+\lnot \mathrm{B}\Sigma_1\). The first result weakens the assumption a bit. It is shown that if there is a model of \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}\) with cofinal \(\Sigma_1\)-definable elements and a \(\Sigma_1\) truth definition for sentences, then \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}+\lnot \mathrm{B}\Sigma_1\) is consistent. It is not know whether such models exist, but in the second part the authors show that there is a model of \(\mathrm{I}\Delta_0+\Omega_1+\lnot \mathrm{exp}\) in which \(\Sigma_1\) definable elements are cofinal and for which there is a \(\Sigma_2\) truth definitions and \(\Pi_2\) truth definition for \(\Sigma_1\) sentences and for all \(n\geq 2\) there is a \(\Sigma_n\) truth definition for \(\Sigma_n\) sentences. The last section presents a proof of an older, but previously unpublished result, that the assuption ``Lessan's bound for truth definitions is optimal'' implies that \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}+\lnot \mathrm{B}\Sigma_1\) is consistent. The paper concludes with some comments on why the problem of consistency of \(\mathrm{I}\Delta_0+\lnot \mathrm{exp}+\lnot \mathrm{B}\Sigma_1\) is so hard.