Independence of higher Kurepa hypotheses (Q453189)

For infinite cardinals \(\lambda < \kappa\), a \((\kappa,\lambda)\)-Kurepa family is a family \({\mathcal F}\) of subsets of \(\kappa\) of size at least \(\kappa^+\) such that for all subsets \(x\) of \(\kappa\) of size \(\lambda\), the restriction \({\mathcal F} \restriction x = \{ y \cap x : y \in {\mathcal F}\}\) has size at most \(\lambda\). For \(n \geq 1\), the gap-\(n\)-Kurepa hypothesis states that there is a \((\lambda^{+n},\lambda)\)-Kurepa family for all infinite cardinals \(\lambda\). A well-known result of Jensen says that assuming the axiom of constructibility there are \((\kappa,\lambda)\)-Kurepa families for all \(\lambda < \kappa\) with \(\kappa\) regular. The authors prove that for \(n \geq 1\) the statement that GCH holds, the gap-\(m\)-Kurepa hypothesis holds for all \(m \neq n\), yet the gap-\(n\)-Kurepa hypothesis fails is equiconsistent with an inaccessible cardinal. To obtain the consistency of this statement, an inaccessible \(\kappa\) is Lévy-collapsed to \(\omega_{n+1}\), thus yielding the nonexistence of \((\omega_n, \omega)\)-Kurepa families, and then \((\omega_n, \omega_i)\)-Kurepa families are added generically for all \(0 < i <n\). For the converse, the authors show more generally that the nonexistence of \((\kappa,\lambda)\)-Kurepa families for regular \(\kappa\) with \(\kappa^\lambda =\kappa\) implies that \(\kappa^+ \) is inaccessible in the constructible universe \(L\).