Any bijection from $\Bbb N$ to itself transforms an ultrafilter on $\Bbb N$ to another (isomorphic) ultrafilter. Any two principal ultrafilters are isomorphic in that sense.

For free ultrafilters on $\Bbb N$, there are $2^{2^{\aleph_0}}$ of them. Since there are $2^{\aleph_0}$ bijections of $\Bbb N$ to itself, there are also $2^{2^{\aleph_0}}$ isomorphism classes of free ultrafilters on $\Bbb N$. So lots of free ultrafilters must be nonisomorphic to each other.

Question: Can you give an explicit example or construction of two free ultrafilters on $\Bbb N$ that are not isomorphic? Assume ZFC.

(Added at the suggestion of @bof in the comments below, in case the question proves too difficult to answer directly):

  1. Give an explicit example of two filters such that no free ultrafilter extending one of them can be isomorphic to a free ultrafilter extending the other.
  2. Can you state a property, preserved by isomorphism, possessed by some but not all free ultrafilters?

(1) is as good as the original question as far as I am concerned.

  • 1
    $\begingroup$ @bof . The existence of a free ultrafilter on $\Bbb N$ cannot be done without the Axiom of Choice, so in a sense there is no explicit example, as it is also consistent with ZF that they don't exist. $\endgroup$ Jun 24, 2020 at 7:51
  • 6
    $\begingroup$ I don't know what you mean by an explicit example since even exhibiting a single nonprincipal ultrafilter on $\omega$ is rather nonexplicit, but what you call being isomorphic is usually called being Rudin-Keisler equivalent. The RK-order has been (is being) extensively studied and there are plenty of results from the 70/80s constructing large families of RK-incomparable ultrafilters on $\omega$. In J. van Mill's chapter on The Handbook of Set Theoretic Topology you can find the construction of two RK-incomparable ultrafilters on $\omega$. Would you count it as explicit? $\endgroup$ Jun 24, 2020 at 9:34
  • 1
    $\begingroup$ To add to the comment by @AlessandroCodenotti, you might also want to look at combinatorial properties such as p-points, q-points, etc., as well as rapid filters for (1), While it is consistent that no such ultrafilter exists, it is still a way to find a reasonable condition. (I should add that I know very little about these questions, but these are keyword for you to start your journey with.) $\endgroup$
    – Asaf Karagila
    Jun 24, 2020 at 17:22
  • $\begingroup$ @AlessandroCodenotti and Asaf: I was hoping there would something simple, but that may turn out to be elusive. Thanks for the references and background info. $\endgroup$
    – PatrickR
    Jun 24, 2020 at 17:56
  • $\begingroup$ There is a product operation on ultrafilters. Ultrafilters that have square roots are not isomorphic to those that haven't. $\endgroup$
    – markvs
    Jun 24, 2020 at 18:17

1 Answer 1


The simplest property that I can think of (right now) that provably (in ZFC) distinguishes some non-principal ultrafilters on $\mathbb N$ from others is "weak P-point", which means "not in the closure in $\beta\mathbb N$ of a countable set of other non-principal ultrafilters." The existence of weak P-points is a theorem of Kunen; the existence of non-principal ultrafilters that are not weak P-points is trivial (take any countably infinite set of non-principal ultrafilters and take any other point in their closure).

  • $\begingroup$ Thanks! Would you have a reference for weak P-points? $\endgroup$
    – PatrickR
    Jun 25, 2020 at 4:17
  • 1
    $\begingroup$ ktiml.mff.cuni.cz/~verner/download/… has a good overview $\endgroup$
    – PatrickR
    Jun 25, 2020 at 8:22
  • $\begingroup$ If I understand you correctly, since $\beta \Bbb N$ \ $\Bbb N$ has no isolated points, this result of Kunen is that $\beta \Bbb N$ \ $\Bbb N$ is not a space of countable tightness. $\endgroup$ Jun 26, 2020 at 6:21
  • $\begingroup$ @DanielWainfleet Kunen's result looks a bit stronger than "not of countable tightness" to me. A counterexample to countable tightness would be a point p that is in the closure of a set A but not in the closure of any countable subset of A. Such a point p could still be in the closure of a countable set disjoint from A, so it might fail to be a weak P-point. $\endgroup$ Jun 26, 2020 at 13:05
  • $\begingroup$ Yes I see. Stronger. $\endgroup$ Jun 26, 2020 at 19:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.