While it is true that $\mathcal P(\kappa)$ is a complete Boolean algebra, it is not necessarily true that $\mathcal P(\kappa)/I$ is complete for an ideal $I$. In particular if we consider $I=J_{bd}$ the ideal of bounded subsets of $\kappa$.

For example, Hausdorff showed that there is an $(\omega_1,\omega_1)$ gap in $\mathcal P(\omega)/\textrm{Fin}$. That is, there is a sequence of order type $\omega_1+\omega_1^*$ which is strictly increasing, but there is no $A$ which realizes the cut between the $\omega_1$ and $\omega_1^*$ parts. Consider now the lower $\omega_1$ sets, if there was a supremum to this collection, it would have to be an element realizing the gap.

Assuming $V=L$. How complete is $\mathcal P(\kappa)/J_{bd}$? If the answer is strictly less than $\operatorname{cf}(\kappa)$, what if we limit ourselves to strictly increasing sequences? Can we get $\operatorname{cf}(\kappa)$-completeness?

That is, assuming that $\gamma<\kappa^+$ and $\langle A_\alpha\subseteq\kappa\mid\alpha<\gamma\rangle$ is strictly increasing modulo $J_{bd}$. Is there a least upper bound to the sequence?

I'd be interested to know (at least some partial) answer outside $V=L$, although that is less important.

  • 3
    $\begingroup$ You don't need a gap to show that suprema don't exist in $\mathcal{P}(\omega)/\mathrm{Fin}$; take any sequence $A_n$ ($n < \omega$) of pairwise-disjoint, infinite subsets of $\omega$. If $A$ is a $\subseteq_*$-upper bound for this sequence, then one can find $B\subseteq A$ such that $A\setminus B$ is infinite and $B$ is still an upper bound, by removing one element from each intersection $A\cap A_n$ . Hence the sequence $A_n$ ($n < \omega$) has no least upper bound. $\endgroup$ – Paul McKenney Aug 29 '13 at 18:03
  • $\begingroup$ Thanks Paul, that's quite helpful. This seems to be easily generalized to large cardinalities. Moreover if we take $A'_n=A_1\cup\ldots\cup A_n$ then we get the same result for an increasing sequence, so both my questions have a negative answer. (Nice seeing you here, by the way!) $\endgroup$ – Asaf Karagila Aug 29 '13 at 18:42

If $\kappa$ is regular, then $\mathcal{P}(\kappa)/J_{bd}$ is a $\kappa$-complete boolean algebra. If $\langle A_\alpha : \alpha < \delta < \kappa \rangle$ is a sequence of subsets of $\kappa$, then the union of these is a least upper bound. This uses the $\kappa$-completeness of $J_{bd}$. It is not $\kappa^+$-complete for the same reason Paul mentioned for $\omega$. Take any partition of $\kappa$ into $\kappa$ many unbounded sets, $\langle A_\alpha : \alpha < \kappa \rangle$, let $B$ be an upper bound, and let $C \subseteq B$ have one point from each $A_\alpha \cap B$ removed. So $C$ is an upper bound and strictly below $B$. For singular $\kappa$, the algebra is $cf(\kappa)$-complete but not $cf(\kappa)^+$-complete: Take any unbounded $A$ set of size $cf(\kappa)$ and consider the algebra below $A$. It is isomorphic to $\mathcal{P}(cf(\kappa))/J_{bd}$.

  • $\begingroup$ So basically $\mathcal P(\kappa)/J$ inherits the completeness of $J$? $\endgroup$ – Asaf Karagila Sep 1 '13 at 12:30
  • $\begingroup$ Yes, but the completeness of the algebra can exceed that of the ideal. $\endgroup$ – Monroe Eskew Sep 1 '13 at 21:13

This answer expands on Paul McKenney's comment, and concerns only the gap structure in $\mathcal{P}(\omega)$/fin. I don't know much about what happens for larger cardinals.

The fact that $\mathcal{P}(\omega)$/fin is highly incomplete actually has interesting consequences, and in particular is useful in showing that $\mathcal{P}(\omega)$/fin is universal for Boolean algebras of size at most $\aleph_1$. To give a better sense of the story here, suppose $\langle A_i: i < \omega \rangle$ is a strictly $\subseteq_*$-increasing sequence of subsets of $\omega$. Then if we iterate Paul's construction, we can find a strictly decreasing sequence $\langle B_j: j <\omega \rangle$ of upper bounds for $\langle A_i: i < \omega \rangle$. A natural question is whether the resulting pair of sequences $A_0 \subseteq_* A_1 \subseteq_* \ldots \subseteq_* B_1 \subseteq_* B_0$ is ever a gap. It turns out the answer is no: one can always find a $C$ such that for all $i, j$ we have $A_i \subseteq_* C \subseteq_* B_j$.

This says that there are no $(\omega, \omega)$-gaps in $\mathcal{P}(\omega)$/fin. Rothberger proved that the minimum cardinal $\kappa$ for which there exists an $(\omega, \kappa)$-gap is the bounding number $\mathfrak{b}$. If the CH holds we have $\mathfrak{b} = \omega_1$, and so in this case we can find $(\omega, \omega_1)$-gaps as well as the more celebrated $(\omega_1, \omega_1)$-gaps of Hausdorff.

Even without CH, we have that $\mathcal{P}(\omega)$/fin has the following four properties:

  • it is atomless (as a Boolean algebra)
  • it has no countable limit points
  • it has no countable cofinal sequences
  • it has no $(\omega, \omega)$-gaps

One can prove from these that $\mathcal{P}(\omega)$/fin embeds every Boolean algebra of size at most $\aleph_1$. In the case of CH, any other Boolean algebra of size $\aleph_1$ with these four properties is isomorphic to $\mathcal{P}(\omega)$/fin. The proof for both facts goes by extending countable partial embeddings one element at a time. If there existed countable limit points in $\mathcal{P}(\omega)$/fin, as you asked about in your question, such a proof would not be possible: such points would block certain partial embeddings from being extended.

An encyclopedic reference for these results and the set theory of $\mathcal{P}(\omega)$/fin more generally is the book Hausdorff Gaps and Limits by Frankiewicz and Zbierski.


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.