4
$\begingroup$

$\DeclareMathOperator{\RK}{\mathrm{RK}}$Let $\beta\omega$ be the Stone-Cech compactification of $\omega$ with the discrete topology. We can endow $\beta\omega$ with an addition operation that extends the addition on $\omega$ in a natural way. (The same statement holds for multiplication.)

The Rudin-Keisler preorder on $\beta\omega$ is defined by $${\bf p} \leq_{\RK} {\bf q} :\;\Leftrightarrow\; (\exists f:\omega\to\omega)\,(\forall U\in{\bf p})\, f^{-1}(U)\in {\bf q}$$ for ${\bf p},{\bf q}\in\beta\omega$.

Question. Given ${\bf a},{\bf b}, {\bf c}\in\beta\omega$, do we have

  1. ${\bf a} \leq_{\RK} {\bf a+b}$, and
  2. if ${\bf a}\leq_{\RK}{\bf b}$, then $({\bf a}+{\bf c}) \leq_{\RK} ({\bf b}+{\bf c})$?
$\endgroup$
6
  • 1
    $\begingroup$ Sanity check: is ${\bf a+b}=\{N\subseteq\omega:\{x:\{y: x+y\in N\}\in{\bf b}\}\in{\bf a}\}$? "The set of subsets of $\omega$ such that for $\bf a$ many $x\in\omega$ there are $\bf b$ many $y\in\omega$ with $x+y\in N$" $\endgroup$ Nov 11, 2020 at 11:00
  • $\begingroup$ That's correct @AlessandroCodenotti $\endgroup$ Nov 11, 2020 at 11:21
  • 1
    $\begingroup$ In other words $x+\mathbf{b}=\lim_{y\in\mathbf{b}}(x+y)$, and $\mathbf{a}+\mathbf{b}=\lim_{x\in\mathbf{a}}(x+\mathbf{b})$. $\endgroup$
    – YCor
    Nov 11, 2020 at 12:03
  • 1
    $\begingroup$ Does the Rudin-Keisler preorder extend the order on $\mathbb N$? Does your question have a positive answer if $a$ or $b$ is in $\mathbb N$? $\endgroup$
    – Tri
    Nov 12, 2020 at 6:52
  • 2
    $\begingroup$ @Tri all principal ultrafilters are RK-isomorphic (look at the bijection of $\omega$ swapping the two points whose singletons are in the ultrafilters) $\endgroup$ Nov 12, 2020 at 9:04

1 Answer 1

9
$\begingroup$

For Question 1: First, there are idempotent ultrafilters, so some ultrafilters satisfy 1 in a very strong form. But 1 does not hold in general. The reason is that the semigroup $\beta\omega-\omega$ has subsemigroups $G$ that are groups of cardinality $2^{\mathfrak c}$ where $\mathfrak c$ is the cardinal of the continuum. (See Hindman and Strauss, "Algebra in the Stone-Cech Compactification", Theorems 2.7(d) and 2,25, and remember that all infinite closed subsets of $\beta\omega-\omega$ have cardinality $2^{\mathfrak c}$.) If the answer to Question 1 were always affirmative, then all elements of $G$ would be RK-below each other (and therefore isomorphic, but I don't need that). That's impossible as any ultrafilter is RK-above only $\mathfrak c$ others.

I expect the answer to Question 2 to be negative also, but I'll need to think some more to prove it. I also expect that there's an easier proof for Question 1. Finally, the answer to Question 1 becomes positive if either a or b is assumed to be a P-point.

Edit: A negative answer to Question 2 follows from the continuum hypothesis. Under CH (or certain weaker hypotheses, but not in ZFC alone), there exist P-point ultrafilters. Let a be one of these, and let b be an idempotent ultrafilter Rudin-Keisler above a. (It's provable in ZFC that any non-principal ultrafilter on $\omega$ is below an idempotent one.) Because a is a P-point, any sum $\mathbf a+\mathbf c$ (with c noprincipal) is isomorphic to the tensor product (also called Fubini product) $$ \mathbf a\otimes\mathbf c=\{S\subseteq\omega\times\omega:\{x:\{y:(x,y)\in S\}\in\mathbf c\}\in\mathbf a\}, $$ and is therefore strictly above both a and c in the Rudin-Keisler order. In particular, $\mathbf a+\mathbf b\not\leq_{\text{RK}}\mathbf b$. But since b is idempotent, this gives $\mathbf a+\mathbf b\not\leq_{\text{RK}}\mathbf b+\mathbf b$ even though $\mathbf a\leq_{\text{RK}}\mathbf b$.

$\endgroup$
1
  • $\begingroup$ Thanks for these results! While your answer is worth accepting, I'll wait for little while before doing so -- maybe an answer to question 2 can be found. $\endgroup$ Nov 11, 2020 at 20:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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