MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I just ran into this deceptively simple looking question.

Is it always possible to partition $\mathbb{R}$ (or any other standard Borel space) into precisely $\aleph_1$ Borel sets?

On the one hand, this is trivial if the Continuum Hypothesis holds. Less trivially, this also follows from $\mathrm{cov}(\mathcal{M}) = \aleph_1$, $\mathrm{cov}(\mathcal{N}) = \aleph_1$, $\mathfrak{d} = \aleph_1$, and similar hypotheses. However, I can't think of a general argument that allows one to split $\mathbb{R}$ into precisely $\aleph_1$ pairwise disjoint nonempty Borel pieces.

On the other hand, PFA or MM might give a negative answer but I don't see a good handle from that end either.

share|cite|improve this question
I don't have the time right now to check the details, but it looks like you can build such a partition out of a Hausdorff gap. Of course Silver's theorem rules out the partition coming from a Borel equivalence relation (if CH fails). – Clinton Conley Aug 6 '11 at 20:12
Thanks, Clinton. I thought of using a Hausdorff gap on the walk home. I think it works too. – François G. Dorais Aug 6 '11 at 20:29
I think it is not well understood which cardinals $\kappa$ have the property that the real line can be partitioned into $\kappa$ Borel sets. As Andreas points out, this is known in the case $\kappa = \aleph_1$. I think it is open, however, whether the continuum being at least $\aleph_n$ implies that the real line can be partitioned into $\aleph_n$ Borel sets (for $n > 1$). If I remember correctly, some form of this problem is due to Steve Watson. – Justin Moore Aug 6 '11 at 22:37
Just to point out, the question I asked as "A compactness property for Borel sets" (not sure how to make a link - sorry) is (via easy reasoning) the negation of the question asked here. So all the constructions given as negative answers to my question can be imported as positive answers to this one. Similarly, Andreas Blass' very simple (pace the classical theorem) answer to this question, provides yet another counterexample to my question. – Alex Simpson Aug 8 '11 at 9:38
I think Alex is referring to… – Joel Reyes Noche Nov 14 '11 at 4:11
up vote 20 down vote accepted

It suffices to express $\mathbb R$ as the union of $\aleph_1$ (not necessarily disjoint) Borel sets such that no countably many of them cover $\mathbb R$, because then you can list them in an $\omega_1$-sequence and subtract from each one the union of the previous ones. Partition $\mathbb R$ into a non-Borel $\Pi^1_1$ set $A$ (say the set of codes of well-orderings of $\omega$) and its complement. A classical theorem says that any $\Pi^1_1$ set is a union of $\aleph_1$ Borel sets, and so is every $\Sigma^1_1$ set. Apply that to $A$ and to $\mathbb R-A$ to get $\mathbb R$ as a union of $\aleph_1$ Borel sets. No countably many of them cover $\mathbb R$ because $A$ is not Borel and thus not a countable union of Borel sets.

share|cite|improve this answer
Thanks Andreas! This is much more constructive than I thought it would be. – François G. Dorais Aug 6 '11 at 20:31
A slightly more explicit version of Andreas' argument: First, identify reals with trees in some Borel way. Here, trees are subsets of $\omega^{< \omega}$ which are downward closed. Next, each tree $T$ has a derivative $T'$: remove all leaves. This naturally defines a transfinite sequence of iterated derivatives $(T^{(\alpha)}: \alpha< \omega_1)$. Let the ("Cantor-Bendixson") rank of $T$ be the first $\alpha<\omega_1$ such that $T^{(\alpha)}$ has no leaves (possibly, but not necessarily, because it is empty). Now the set of trees with a given rank is certainly Borel. – Goldstern Feb 21 '12 at 20:28

There is a Hausdorff gap, a sequence $\{f_\alpha,g_\alpha:\alpha\lt\omega_1\}$ of $\omega\to\omega$ functions such that $f_\alpha\lt^*f_\beta<^*g_\beta<^*g_\alpha$ hold for $\alpha\lt\beta\lt\omega_1$ (here $f\lt^* g$ denotes eventual dominance, i.e., that $f(n)\lt g(n)$ holds for large $n\lt\omega$) and there is no function $f:\omega\to\omega$ such that $f_\alpha\lt^* f\lt^* g_\alpha$ holds for $\alpha\lt\omega_1$. If we identify the reals with ${}^\omega\omega$ and set $H_\alpha=\{f:f_\alpha\lt^* f \lt^* g_\alpha\}$ then $\{H_\alpha:\alpha\lt\omega_1\}$ is a decreasing sequence of nonempty $F_\sigma$ with empty intersection.

share|cite|improve this answer
Let me emphasize the "$F_\sigma$" in the last sentence of Peter's answer. His argument, unlike mine, bounds the Borel complexity of the pieces of the partition. – Andreas Blass Aug 7 '11 at 13:41

Here is another example from recursion theory:

Take a chain $\{x_{\alpha}\}_{\alpha<\omega_1}$ from Turing degrees.

For each $\alpha<\omega_1$, let $A_{\alpha}$ be the collection of the reals neither in $\bigcup_{\beta<\alpha}A_{\beta}$ nor Turing-computing $x_{\alpha}$.

Then $\{A_{\alpha}\}_{\alpha<\omega_1}$ is a Borel partition of $\mathbb{R}$.

share|cite|improve this answer
Another nice one, thank you! – François G. Dorais Nov 14 '11 at 2:12

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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