Worked solution: Gödel and Cohen: the continuum hypothesis is independent of ZFC (1963)
Every subset of that lives in was named by some finite logical recipe using ingredients from an earlier stage. A counting trick — take a small countable "sample" of the whole construction that still contains the recipe, then collapse that sample down to its true, honest size — shows that recipe could not have needed to wait past the countable-length stage to appear.
Since there are only -many stages before , and only countably many recipes get used up at each one, at most -many different subsets of can ever exist inside — exactly matching what demands.
The condensation lemma: if is an elementary substructure of (for a sufficiently large limit ordinal), its Mostowski collapse is again exactly for some ordinal . Combined with the downward Löwenheim–Skolem theorem, one can, for any with , find a countable elementary substructure (for large enough that ) with ; collapsing gives some with (since is countable) and .
Hence every real of appears already in , so , and since , this gives . Cantor's theorem already gives , so inside , : that is, (the same argument, run at every cardinal instead of just , in fact gives the stronger generalized continuum hypothesis in ).
Since is built inside any model of using only the axioms themselves, this shows : can never be refuted from alone. This is only half of independence; the remaining three steps build a universe with strictly more reals than allows, to show is equally consistent.
- Condensation lemma
- The technical fact that collapsing an elementary substructure of a level of the constructible hierarchy always lands you back inside the hierarchy itself, at some earlier level ; it is the key rigidity property that makes so easy to analyze.