10
$\begingroup$

Is it consistent that there exists a partition $P$ of the real number line $\mathbb{R}$ such that $|P|>\aleph_{0}$ but where $\bigcup R$ is Borel whenever $R\subseteq P$?

If $2^{\aleph_{0}}<2^{\aleph_{1}}$, then the answer to this question is $\textbf{no}$ since there would be at least $2^{\aleph_{1}}$ subsets of each uncountable partition $P$ but there are only $2^{\aleph_{0}}$ Borel sets. We therefore need to use a model such that $2^{\aleph_{0}}=2^{\aleph_{1}}$ to construct our counterexample.

Under $MA+\neg CH$ does there exist a partition $P$ of the real number line $\mathbb{R}$ such that $|P|>\aleph_{0}$ but where $\bigcup R$ is Borel whenever $R\subseteq P$?

Perhaps such a partition $P$ is impossible to construct by some fact obvious to descriptive set theorists that I probably have overlooked.

I call a partition $p$ of a Boolean algebra $B$ such that $\bigvee R$ exists whenever $R\subseteq p$ a subcomplete partition and subcomplete partitions came up all the time in my work on Boolean partition algebras, so I would be interested in which partitions of a given Boolean algebra are subcomplete and which ones are not.

$\endgroup$
2
  • 1
    $\begingroup$ Trivial observation: all but countably many $A \in P$ must be uncountable (and hence have $|A| = \mathfrak{c}$). For suppose there is an uncountable $P_c \subset P$ with every $A \in P_c$ countable. Since we are assuming $\lnot \mathsf{CH}$, we can find $P_0 \subset P_c$ with $\aleph_0 < |P_0| < \mathfrak{c}$. Then $B = \bigcup P_0$ has cardinality $|P_0|$, which cannot be if $B$ is Borel. $\endgroup$ Aug 18, 2015 at 4:21
  • $\begingroup$ Another trivial observation: if by "consistent" we mean "consistent with ZF", then it is, because we can work in a model of ZF in which every subset of $\mathbb{R}$ is Borel. (For instance, a model in which $\mathbb{R}$ is a countable union of countable sets.) In that case we can take the trivial partition of $\mathbb{R}$ into singletons. Ashutosh's answer below shows it is not consistent with ZFC. $\endgroup$ Sep 8, 2015 at 16:36

1 Answer 1

16
$\begingroup$

Suppose $\{P_i : i < \kappa\}$ is such a partition. Let $f: R \to R$ be a function satisfying $|f[P_i]| = 1$ and for all $i < j < \kappa$, $f[P_i] \cap f[P_j] = \phi$. Then $f$ is Borel so its image is an uncountable analytic set of size less than continuum: Contradiction.

$\endgroup$
4
  • 2
    $\begingroup$ It took me a moment to understand $f$. Another way to describe it: Using AC, choose one element $x_i$ from each $P_i$, and set $f(x) = x_i$ if $x \in P_i$. For any Borel set $B$, we have $f^{-1}(B) = \bigcup_{i : x_i \in B} P_i$ which by assumption is Borel. So $f$ is Borel. $\endgroup$ Aug 18, 2015 at 4:54
  • $\begingroup$ @NateEldredge Does this only show that the partition must be size $2^{\aleph_0}$? $\endgroup$
    – William
    Aug 18, 2015 at 4:59
  • 4
    $\begingroup$ @William: If the partition has size $2^{\aleph_0}$ then by choosing subsets of the partition we get $2^{2^{\aleph_0}}$ distinct Borel sets, which is impossible. $\endgroup$ Aug 18, 2015 at 5:04
  • $\begingroup$ So, long story short, given any uncountable partition of $\Bbb R$ into Borel sets, the partition is a refinement of a partition into non-Borel sets. $\endgroup$
    – Asaf Karagila
    Aug 18, 2015 at 5:34

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.