In mathematics, extendible cardinals are large cardinals introduced by , who was partly motivated by reflection principles. Intuitively, such a cardinal represents a point beyond which initial pieces of the universe of sets start to look similar, in the sense that each is elementarily embeddable into a later one.
For every ordinal ÷, a cardinal ú is called ÷-extendible if for some ordinal û there is a nontrivial elementary embedding j of V<sub>ú+÷</sub> into V<sub>û</sub>, where ú is the critical point of j, and as usual V<sub>ñ</sub> denotes the ñth level of the von Neumann hierarchy. A cardinal ú is called an extendible cardinal if it is ÷-extendible for every nonzero ordinal ÷ (Kanamori 2003).
For a cardinal , say that a logic is -compact if for every set of -sentences, if every subset of or cardinality has a model, then has a model. (The usual compactness theorem shows -compactness of first-order logic.) Let be the infinitary logic for second-order set theory, permitting infinitary conjunctions and disjunctions of length . is extendible iff is -compact.
A cardinal ú is called ÷-C<sup>(n)</sup>-extendible if there is an elementary embedding j witnessing that ú is ÷-extendible (that is, j is elementary from V<sub>ú+÷</sub> to some V<sub>û</sub> with critical point ú) such that furthermore, V<sub>j(ú)</sub> is ã<sub>n</sub>-correct in V. That is, for every ã<sub>n</sub> formula ÃÂ, àholds in V<sub>j(ú)</sub> if and only if àholds in V. A cardinal ú is said to be C<sup>(n)</sup>-extendible if it is ÷-C<sup>(n)</sup>-extendible for every ordinal ÷. Every extendible cardinal is C<sup>(1)</sup>-extendible, but for nâÂÂ¥1, the least C<sup>(n)</sup>-extendible cardinal is never C<sup>(n+1)</sup>-extendible (Bagaria 2011).
VopÃÂnka's principle implies the existence of extendible cardinals; in fact, VopÃÂnka's principle (for definable classes) is equivalent to the existence of C<sup>(n)</sup>-extendible cardinals for all n (Bagaria 2011). All extendible cardinals are supercompact cardinals (Kanamori 2003).