Models of Second-Order Zermelo Set Theory (Q4940733)

The author investigates second-order variants of Zermelo set theory and their impact on the cumulative hierarchy of sets (the rank universes \(V_{\alpha}\)). The results of his work allow him to assert that replacement has an impact at suprisingly low levels of the cumulative hierarchy and that without it the second-order Zermelo axioms are not adequate to describe initial segments of the hierarchy indexed by limit \(\lambda>\omega\). \(\text{Z}^{-}\) denotes the core second-order theory where separation has a second-order formulation and the replacement axiom schema and the infinity axiom are removed. The first part of the paper studies \(\text{Z}^{-}\) together with four possible axioms of infinity: \(\text{Inf}_{\text{Z}}\), Zermelo's original axiom of infinity; Inf, the usual textbook axiom of infinity; InfDed, Dedekind's axiom of infinity; and InfNew, \(\exists y\) \((\emptyset\in y\wedge\forall x \forall z\) \((x\in y\wedge z\in y\rightarrow x\cup\{z\}\in y)\). It turns out that \(\text{Z}^{-} + \text{InfNew}\) is the strongest amongst the four possible \(\text{Z}^{-}\) + \textit{an axiom of infinity} variants, and it guarantees the existence of \(V_{\omega}\). On the other hand, \(\text{Z}^{-} + \text{InfDed}\) is the weakest of the four. The author proves that there is no standard \(\subseteq\)-least transitive model of \(\text{Z}^{-} + \text{InfDed}\). Even though \(\text{Z}^{-} + \text{InfNew}\) guarantees the existence of the first \(\omega + \omega\) levels of the cumulative hierarchy, the author shows that it has non-well-founded models and hence does not serve to characterize initial segments of the cumulative hierarchy indexed by limit \(\lambda>\omega\). The author sketches a proof that \(\text{Z}^{-} + \text{InfNew}\) is relatively interpretable in \(\text{Z}^{-} + \text{Inf}\). He also indicates that \(\text{Z}^{-} + \text{InfNew}\), \(\text{Z}^{-} + \text{Inf}_{\text{Z}}\) and \(\text{Z}^{-} + \text{Inf}\) are equi-interpretable. He views this as evidence that both \(\text{Z}^{-} + \text{Inf}\) and \(\text{Z}^{-} + \text{Inf}_{\text{Z}}\) are sufficient for the development of ordinary mathematics. In the last section of the paper the author takes up the question of what can be added to \(\text{Z}^{-} + \text{Inf}\) to characterize the \(V_\lambda\)'s for limit \(\lambda>\omega\).