Regularity of ultrafilters and fixed points of elementary embeddings (Q1818018)

Let \(\kappa\) be a compact cardinal, \(\lambda\) a cardinal \(\geq\kappa\) and \(U\) a fine ultrafilter on \({\mathcal P}_\kappa\lambda\). As is well known, the canonical embedding \(j_U:V\to \text{Ult} (V,U)\) is the identity below \(\kappa\) and \(j_U(\kappa)\geq 2^{\lambda^{<\kappa}}\) holds. What about the action of \(j_U\) on higher cardinals? \textit{J. B. Barbanel} [J. Symb. Log. 47, 84-88 (1982; Zbl 0503.03026)] determined the class of fixed cardinals when \(U\) is normal. \textit{Y. Abe} [J. Symb. Log. 49, 808-812 (1984; Zbl 0582.03037)] solved the general case: Theorem 1. Let \(\kappa\), \(\lambda\) and \(U\) be as above. Then \(j_U\) moves a cardinal \(\mu>2^{\lambda^{<\kappa}}\) exactly when \(\kappa\leq\text{cf}(\mu)\leq \lambda^{<\kappa}\) or \(\mu=\nu^+\) with \(\kappa\leq \text{cf}(\nu)\leq\lambda^{<\kappa}\). The main advantage of Abe's [loc. cit.] over Barbanel's [loc. cit.] result lies in the assertion that the embedding moves such cardinals. In this paper we pursue this subject in somewhat greater generality, or in terms of regularity instead of fineness. In particular we give two alternative proofs of the main claim of Theorem 1, which are both much simpler than the original one. On the other hand Abe [loc. cit.] noted that some of his arguments work even for a uniform ultrafilter. We also show that such extension may fail in other cases.