Thomas Forster wrote:
> I think what Bill was after (his query arose in a conversation we were
> having) was the cutest/simplest (pair of) constructions of two
> nonprincipal ufs over N - using AC of course.
The simplest "construction" that comes to mind is to build the two
ultrafilters simultaneously, by a transfinite induction of length c
(the cardinal of the continuum, regarded as an initial ordinal),
preventing, at each step, one permutation from being an isomorphism.
One uses a family of independent sets to make sure the construction
doesn't end before c steps, so there's enough "time" to handle all
the permutations. This argument is due to Kenneth Kunen,
"Ultrafilters and independent sets," Trans. A. M. S. 172 (1972) pp.
299-306. Kunen actually constructs two free ultrafilters on the
integers with the property that no function (permutation or not) maps
either ultrafilter to the other.
Andreas Blass
|
